1Classifying species by shared functional characteristics is important if common functional response groups are to be identified among different taxa.
2We investigated plant traits that determine the response of forest plant species to land use changes using literature data. Sources from eight European countries and four North-eastern American states, comprising 20 field studies yielded information on 216 forest plant species. For these species, data on 13 life history traits were collected.
3Trait correlation structure was similar in the European and American data-sets and corresponded well to those described in the literature. The European and American herbs and the European graminoids were clustered into distinct emergent groups based on their plant traits. The profiles of the European and American emergent groups were similar.
4Herb species belonging to emergent groups characterized by low dispersability (i.e. large seeds, low fecundity, unassisted dispersal) were relatively slow colonizers. Dispersability (and not recruitment) seems to be a key factor limiting the colonization of some forest plant species. The relationship between dispersability and colonizing capacity was less clear for graminoids.
5A life history trait-based approach offers good opportunities to gain insight into the mechanisms behind species response to land-use change.
Classifying species by shared biological characteristics that relate to function, rather than by phylogeny, is increasingly recognized as a tool in identifying common functional response groups amongst differing taxa (Grime et al. 1988; Lavorel et al. 1997; Diaz & Cabido 2001). These classifications may be grouped into different categories in order of increasing specificity of objective (Lavorel et al. 1997): (i) emergent groups (e.g. Leishman et al. 1995), (ii) strategies (e.g. Grime 1977 and Grime et al. 1988), (iii) functional types (e.g. Thompson et al. 1996) and (iv) specific response groups (e.g. Noy-Meir et al. 1989; McIntyre & Lavorel 2001). Emergent groups are groups of species exhibiting correlations between a set of plant traits. Plant strategies and functional types, respectively, group species that are similar in resource use and in their role in ecosystem functioning. However, both Noble & Gitay (1996) and Lavorel et al. (1997) stressed that such classifications are unlikely to be feasible due to their generality and lack of relation with function. They advocated focusing on the identification of more specific response groups, which are context-specific functional classifications determined by a particular response to disturbance such as grazing or land use change.
We attempted to define trait-based groups of forest plant species related to their response to land use changes. Such species are evolutionarily adapted to stable environments, and the highly dynamic urban and agricultural landscapes of Europe and N-America therefore pose a major conservation problem (e.g. Hermy et al. 1999; Honnay et al. 2002). Deeper insight into the traits determining their responses may also allow prediction of vegetation changes in response to future changes in land use.
We studied secondary colonization following a period of agricultural land-use. This approach offers good opportunities to determine how common life history traits among diverse species control the response to land-use change due to (i) the presence of a well-defined disturbance event (i.e. the complete removal of forest species during agricultural use), (ii) the ability to observe the response of the forest species by monitoring recolonization, and (iii) the empirical observation of widely varying colonizing capacities between species. Thus, it may take tens to hundreds of years for some species to recolonize secondary forests while others colonize within a few years (e.g. Hermy & Stieperaere 1981; Peterken & Game 1984; Whitney & Foster 1988; Motzkin et al. 1996). Previous observational studies (e.g. Matlack 1994; Brunet & von Oheimb 1998; Bossuyt et al. 1999; Butaye et al. 2001; Singleton et al. 2001; Verheyen & Hermy 2001) suggest that seed availability and recruitment ability are major bottle-necks during recolonization. This was partly confirmed by experimental studies in Sweden (Eriksson & Ehrlén 1992; Ehrlén & Eriksson 2000), but it is impossible to screen large sets of species within a reasonable time interval. The potentially more powerful trait-based approach has previously been largely unexplored (but see Graae & Sunde 2000). However, the availability of similar data-sets in North-eastern America and North-western Europe allows us to compile two independent data-sets and thus to generalize across temperate forest regions.
Different methodologies to delineate response groups are possible (McIntyre & Lavorel 2001). Our methodology consists of (i) the identification of slow and fast colonizing forest plant species (i.e. the response groups) based on an extensive literature review, (ii) the identification of emergent groups (sensuLavorel et al. 1997) based on plant trait correlations, (iii) investigation of the relation between plant traits and emergent groups on the one hand and plant response to land-use change on the other.
the species sets
We compiled the available literature comparing the vegetation of ancient and recent forests for the cool temperate regions of North-eastern America (the area covered by Gleason & Cronquist 1991) and North-western Europe. Forests situated on land that was once cleared for agriculture, pasture or habitation, are defined as recent. In North-western Europe, forest that has existed continuously since the date of the oldest available historical site information, typically 150–400 years before present (cf. Hermy et al. 1999), is defined as ancient. However, given the prolonged and intensive human impact on the European landscape, one cannot assume that these forests had previously been continuously forested. In North-eastern America, ancient forests are those that have never been cleared since European settlement (∼350 bp). It should however, be noted that most ancient forests on both continents have been managed and are therefore rarely virgin or old-growth forests.
We searched a large variety of data sources, including peer-reviewed journal papers, unpublished theses and reports, but only those studies that met the following criteria were used for further analyses: (i) the sampled forests are situated on mesic soils in lowlands or lower montane areas; (ii) a clear description of the land-use history is given; (iii) frequency data are available for herbaceous vascular plants in both forest types and (iv) in recent forests canopy closure had already taken place.
For each species that was cited at least twice (i.e. 139 herbs + 33 graminoids = 172 European species and 44 herbs + 0 graminoids = 44 American species), a ‘Colonizing Capacity Index’ (CCI) was calculated using a similar index to that employed by Bekker et al. (1998) to assess the longevity of seeds in soil.
CCI = [((1.5 × A* + A) − (1.5 × R* + R))/(A* + A + R* + R)]× 100/1.5,
where A* represents the number of studies in which species X is significantly more frequent in ancient forest; A the number of studies in which species X is more frequent, but not statistically significant, in ancient forest; R the number of studies in which species X is equally frequent or tends to be more frequent in recent forest and R* the number of studies in which species X is significantly more frequent in recent forest. For studies lacking statistical analyses of species affinity with ancient forest, we calculated Pearson Chi2 association measures (Table 1). The index ranges from −100 (strongly associated with recent forest) to +100 (strongly associated with ancient forest). An overview of the species and their CCI-values is provided in Appendix 1 (Supplementary Material).
For the same 172 European and 44 American species, data on 13 traits were compiled based on a list of core traits suggested by Weiher et al. (1999) supplemented with traits believed to affect plant species colonization (e.g. dispersule production, age of first reproduction; Table 2). Compilations of autecological information of European species (The Biological Flora of the British Isles; Ellenberg et al. 1992; Hodgson et al. 1995; Kleyer 1995; Thompson et al. 1997; Bouman et al. 2000) and American species (Montgomery 1977; Gleason & Cronquist 1991; Cullina 2000; Mabry et al. 2000; Kern unpubl.) were used in addition to species descriptions found in the ecological literature (a full list of references is available from the authors). As we are not aware of any suitable American seed bank compilation, this variable was omitted from the American data-set. On average, trait data were lacking for 30% of the European herb species, 27% of the European graminoids and 16% of the American herbs (Table 2).
Table 2. Overview of the plant traits incorporated in the database
Five classes according to Mabry et al. (2000): 1: spring flowering; 2: mid summer flowering; 3: spring-mid summer flowering; 4: fall flowering; 5: mid-summer-fall flowering.
Separate statistical analyses were performed for the European and American data-sets as well as for the different growth forms (graminoids and herbs; cf. McIntyre & Lavorel 2001), although no American graminoids were included in the data-set. Except where mentioned, all analyses were performed with SPSS 10.0 (SPPS 1999).
In an exploratory phase, correlations between the traits were examined using Spearman rank correlation coefficients calculated between the ordinal and ratio scaled traits. The relationships between ordinal and nominal traits were assessed by means of a Kruskal–Wallis test and the association between nominal traits was determined with the Pearson Chi2 test-statistic (Siegel & Castellan 1988). A Mantel test using PopTools 2.1 (Hood 2001) was used to determine whether a similar correlation structure was found between the traits in Europe vs. North America. For this analysis, proximity matrices based on Spearman rank correlations were used (eight traits) and 999 random permutations were calculated. The intercontinental congruence of the relationships between nominal traits was assessed using the average rank of the ordinal and continuous traits in the nominal trait classes (these ranks are generated by the Kruskal–Wallis test; see above). We then counted the number of traits with a similar ranking of nominal trait classes in Europe and America. Next, emergent groups based on plant trait correlations were identified. This was done by calculating Gower's Similarity Coefficients (GSC; Gower 1971) among a subset of species for which information was available on more than 50% of the studied traits (102 and 44 European and American herbs and 25 European graminoids). GSC is a general measure of proximity that can cope with mixed data-types (nominal, ordinal, interval and ratio) as well as with missing values. The calculated Gower proximity matrices were then used to (i) ordinate the species by non-metric multidimensional scaling (NMDS) and (ii) cluster the species into emergent groups using Ward's method (Clustangraphics 5.08; Clustan Ltd. 2001). The optimal cluster number was determined through visual inspection of the cluster dendrogram and the ordered proximity matrices. NMDS-ordination plots were made to visualize the relationships between the groups. NMDS-ordinations were performed with the ALSCAL algorithm implemented in SPSS 10.0 (SPSS 1999).
Finally, the relationship between the species’ colonization capacity, the individual traits, and the emergent groups was quantified by means of Spearman rank correlation coefficients and Kruskal–Wallis tests (Siegel & Castellan 1988). Using the same methods, the relationships between the CCI and the traits within the emergent groups of European herb species was assessed. Due to sample size limitations, these analyses could not be performed for the American herbs or the European graminoids.
species response to land-use change
On both continents, the CCI spans the whole range between −100 and +100 but the index scores for most species are intermediate (see Appendix 1). Species with extreme CCI scores exhibit a consistent response across the studies, while species with intermediate scores often exhibit divergent responses in different studies.
relationships among the traits and delineation of the emergent groups
Many of the traits investigated are significantly correlated for both European and American herb species (Tables 3 & 4, respectively). Mantel analysis for the ordinal and continuous traits (rM = 0.64; P = 0.002) and similar rankings of dispersal and flowering phenology classes for four out of eight traits indicate that the correlation structures of both data-sets are similar. Note that all of the American species are perennials and therefore, ‘life cycle’ was not included as a variable in the analyses. Fewer significant intertrait correlations were found for the European graminoids, all of which were perennials (Table 5). This is at least partly a result of the small sample size.
Table 3. Relationships among traits and with the CCI for European herb species. Values are Spearman rank correlations, Kruskall Wallis test statistics (italic) and Pearson Chi2-measures of association (underlined)
Table 4. Relationships among traits and with the CCI for American herb species. Values are Spearman rank correlations, Kruskall Wallis test statistics (italic) and Pearson Chi2-measures of association (underlined)
Table 5. Relationships among traits and with the CCI for European graminoids. Values are Spearman rank correlations, Kruskall Wallis test statistics (italic) and Pearson Chi2-measures of association (underlined)
The optimal cluster numbers for the emergent groups were four (European herbs), three (American herbs) and three (European graminoids). The most characteristic traits were used to label the groups (Table 6). Most groups separated out well in the ordination space (Fig. 1), although some overlap exists with the European herbs.
Table 6. Overview of the traits associated with the emergent groups. (a) European herbs; (b) American herbs; (c) European graminoids
, Kruskall–Wallis test. Values are group averages and groups that are significantly different are indicated with different letters (a, b, c). A Bonferroni correction with alfa = 0.05 was performed for the multiple comparisons between the groups (Siegel & Castellan 1988: 213).
, Pearson Chi2 association test. Values are the differences between the observed and expected values for the different classes (see Table 2).
Not unexpectedly, the European short lived herbs (mostly annuals and biennials, e.g. Melampyrum pratense) differ from the other three emergent groups by their short generation time and limited vegetative spread (Table 6). The species belonging to this group do not flower early and produce relatively few but heavy seeds. No predominant dispersal strategy is present and the seed bank is relatively persistent. Species grouped as tall perennials with light seeds (e.g. Solidago virgaurea) also have a distinct profile (Table 6 and Fig. 1), producing many light seeds that are mostly wind dispersed and form a persistent seed bank. These species are typically late-flowering, have a long generation time and exhibit intermediate vegetative spread. There are relatively few differences between the two emergent groups characterized by heavy seeds (Table 6 and Fig. 1). Both groups contain species that produce relatively few heavy seeds that do not form a persistent seed bank and that exhibit strong vegetative spread. These two emergent groups differ by the size, flowering time and dispersal strategy of the species. The tall perennials with heavy seeds (e.g. Stachys sylvatica) primarily flower during mid-summer and have varying dispersal strategies. By contrast, the small perennials with heavy seeds (e.g. Anemone nemorosa) typically flower in the spring and are passively dispersed.
There are some striking similarities between the three groups that emerged from the American herb data-set and the three groups with European perennials (Table 6). The American perennials with light seeds– group contains species that produce many small wind-dispersed seeds and flower late (e.g. Solidago caesia), while the two other American groups contain species with heavy seeds. The latter groups are primarily discriminated by dispersal type (animal vs. unassisted). Exemplary species are, respectively, Smilacina racemosa and Cardamine diphylla.
The three emergent groups in the European graminoids exhibit different trait combinations compared to the herb species. The small, summer flowering vegetatively spreading graminoids (e.g. Deschampsia flexuosa) have relatively small, more spherical seeds that germinate easily and do not form a persistent seed bank. The large, summer flowering graminoids are generally taller and characterized by limited vegetative spread (e.g. Festuca gigantea). The seeds tend to be larger, more elongated and germinate less easily. Except for flowering time and dispersal and seed bank type, most traits of the early flowering graminoids have values intermediate between those of the two other groups. Species of the early flowering graminoids like Milium effusum tend to flower early, are more passively dispersed, and produce a persistent seed bank.
relationships between traits, emergent groups and species response to land-use change
The strongest relationships between individual traits and the CCI were found for the European herb species (Table 3). The American herb species exhibit fewer significant correlations with the CCI (Table 4), although similar trends to the European species were observed (e.g. seed production, age of first reproduction). Slow colonizing European graminoids tend to be early flowering species, species with relatively heavy seeds, or species requiring specific germination conditions (Table 5).
The emergent groups are also closely related with the species response to land-use change (Table 6). The short-lived herbs and the small seeded perennial herbs tend to have good colonizing capacities, while large seeded perennials have relatively poor colonizing capacities. The colonizing capacity of the graminoids decreases from the small, summer flowering vegetatively spreading species to the large, summer flowering species to the early flowering graminoids (Table 6).
For European herbs, the varied responses to land-use change within emergent groups was due to differential responses to one or more traits (Table 7).
Table 7. Spearman rank correlations between the CCI and the traits within the emergent groups of the European herbs. Using the Kruskall–Wallis test, no significant differences were found for dispersal type and flowering phenology (results not presented). Life cycle was not included in these analyses since it was nearly constant within the groups
Short lived herbs
Tall perennials with light seeds
Tall perennials with heavy seeds
Small perennials with heavy seeds
, P ≤ 0.001;
, 0.001 < P ≤ 0.01;
, 0.01 < P ≤ 0.05;
, 0.05 < P ≤ 0.1; NS, not significant; /, less than 5 observations. No correlations calculated.
Predicting the effects of land-use changes on plant community composition is a major challenge for ecologists. Using readily available data on relevant plant traits, the trait-based methodology enabled us to identify patterns of species response to land-use change, irrespective of factors such as local environmental conditions and region, and may be used to predict the response of species not included in this data-set.
the cci and species response to land-use change
The colonizing capacity index (CCI) proved to be a useful, synthetic measure to objectively combine species responses from different studies. The weighting given to studies in which species are significantly more frequently in either ancient or recent forest is debatable. However, changing this weight is unlikely to affect the results drastically.
Although not formally analysed, some closely related species (e.g. Allium ursinum vs. A. tricoccum; Asarum europaeum vs. A. canadense; Paris quadrifolia vs. Trillium spp.; Polygonatum multiflorum vs. Polygonatum spp.) exhibit similar responses on both continents. Similarly, Kornas (1972) noted an ecological correspondence between disjunct pairs of species in Eurasia and North America, a pattern that is striking given the fact that the disjunct patterns in temperate forests of the northern hemisphere had established by the Miocene and further diversification of species on one or both continents took place afterwards (e.g. Wen 1999). Some species or related taxa (e.g. Arum maculatum vs. Arisaema triphyllum; Circaea lutetiana; Oxalis acetosella) have different colonizing abilities in Europe vs. North-eastern America, perhaps as a result of differences in landscape characteristics. As convincingly shown by Vellend (2003), colonization success is strongly dependant on the proportion of ancient forests in the landscape and on the overall degree of forest fragmentation. In general, far less ancient forest exists and actual forest cover is much lower in North-Western Europe than North-Eastern America.
relationships among the traits and delineation of emergent groups
Our results are consistent with well-established correlation patterns among plant traits (e.g. Harper 1977; Grime et al. 1988) and, in particular, with the characterization of life-history attributes of temperate forest herbs by Bierzychudek (1982). A formal analysis of trade-offs, which is beyond the scope of this paper, should account for phylogenetic relationships and possibly confounding effects on some measures (e.g. size-related effects on fecundity; cf. Jakobsson & Eriksson 2000). Our results are, however, consistent with those previously reported. The negative correlation between seed number and seed size (e.g. Harper et al. 1970; Jakobsson & Eriksson 2000; Leishman 2001), the association between fecundity and dispersal type (e.g. Eriksson & Jakobsson 1999) and the association between flowering time on the one hand and seed size and germination requirements on the other (e.g. Grime et al. 1981; Primack 1987) are all known. In addition, it appears that larger seeds have less persistent seed banks (see also Thompson et al. 1993; Bekker et al. 1998). However, similar to Leishman & Westoby (1998), we did not find correlations between seed shape and seed bank persistence for forest herb species.
The emergent groups identified illustrate the different ways that understorey species have adapted to forest environments. The similarity between the three groups of perennial herbs on both continents suggests that the ecological correspondence extends beyond disjunct taxa (see above), but pertains to the entire forest flora. In that respect, it is likely that an analogue of the European group of short lived forest species also exists in North-Eastern America, but that it was not well-represented in our analyses due to the limited sample size. These observations also raise the question of whether the corresponding suites of species assemble into forest communities in similar ways on both continents.
relationships between traits, emergent groups and species response to land-use change
Our results demonstrating correlations between life-history attributes and colonization capacity match closely the results of Rejmanek & Richardson (1996) who compared life-history traits of invasive vs. non-invasive pine species. In Danish forests, Graae & Sunde (2000) found that species with heavy seeds, transient seed banks, ant dispersed seeds, early and short flowering time, low stature and high extent of lateral spread were more common in ancient forests, while the proportion of species with small, long-lived seeds, epizoochorous dispersal, little lateral spread and longer flowering period was higher in recent forests. However, Graae & Sunde (2000) restricted their analyses to bivariate correlations between traits and the species response to land-use change and did not take the correlations among the traits explicitly into account. This limitation was resolved in this study by first delimiting emergent groups based on trait correlations and by then assessing the match between these emergent groups and the response to land use change (cf. McIntyre & Lavorel 2001).
Nevertheless, there is still considerable variation in response within the emergent groups (cf. Table 7). So, while the range of possible colonization capacities is strongly related to the group the species belongs to, trait-variation within the group determines species’ variation in colonization capacity. On average, small perennials with large seeds species are relatively poor colonizers (CCI = 23, Table 6). However, colonization becomes even more problematic for species which have a delayed age of first reproduction (Table 7), confirming the need to analyse responses to land-use change (and disturbance in general) in a hierarchical way (Lavorel et al. 1997).
Recent attempts to incorporate such trait variation into mathematical models describing population spread (e.g. Clark et al. 2001; Higgins & Cain 2002) also indicate that colonization ability cannot be predicted from information on a single trait like disperal strategy or fecundity. Interactions among traits and with landscape characteristics (cf. Vellend 2003) greatly affect colonization rates and must therefore be explicitly taken into account.
Finally, it is noteworthy that some species that readily colonize mesic sites are much slower to colonize xeric sites (e.g. Kalmia angustifolia, Vaccinium angustifolium; Motzkin et al. 1999 and Eberhardt 2001). It seems that the production of large quantities of good dispersing seeds is no guarantee for the colonization of sites with more extreme environmental conditions. This may indicate that recruitment, and not dispersal, is the prime factor limiting colonization of some types of sites.
We conclude that the trait-based approach offers insight into the mechanisms responsible for variation in forest plant colonization. The results of this study suggest that the response of forest plant species to land-use coincides with a clustering of species based on a combination of (mostly regenerative) life-history traits and that slow colonizing species are typically characterized by traits that result in low dispersability. Hence, conservation measures should focus on the existing populations of slow colonizing species, many of which occur in ancient forests, and on forest expansion in the immediate vicinity of source populations. In highly fragmented agricultural and urban landscapes, introductions in potentially suitable but unoccupied sites may also be appropriate (cf. van Groenendael et al. 1998).
The first author wants to thank Bea Bossuyt and Glenn Matlack for the fruitful discussions and Cathy Mabry for putting her database at our disposal. Also many thanks to the people at the Harvard Forest and to Jesse Bellemare and Rob Eberhardt, in particular, for their hospitality and interest in my work. The paper benefited from the comments of Sandra Lavorel and anonymous referees and was written while KV held a grant from the Flemish Institute for the encouragement of Scientific and Technological Research (I.W.T.). OH has a postdoctoral fellowship from the Flemish Fund for Scientific Research (F.W.O.).