Introduction

It is well established from paleoecological studies that climatic changes cause displacement of plant distributions. Future global warming is therefore expected to cause large vegetation shifts and long-distance plant migration towards higher latitudes and altitudes (Christensen et al., 2007). Plant migration requires good colonization capacity, and there is a growing concern that plant species that in the present landscape demonstrate poor colonization capacity will not be able to track their suitable habitats at a sufficient speed and hence will suffer major decline in distribution area (Honnay et al., 2002; Higgins, Lavorel & Revilla, 2003; Skov & Svenning, 2004). Many common herb species growing in temperate forests in Europe and North America are known to have poor colonization capacity in the present landscape (Hermy et al., 1999; Flinn & Vellend, 2005). Studying forest herbs that we know vary in colonization capacities may be useful in revealing patterns important for understanding climate impacts on plant distribution both in the past and the future.

Long-distance migration requires successful colonization by seeds for most plant species. The colonization process by seeds requires successful seed production, seed dispersal, break of seed dormancy, germination, and seedling survival to establish new individuals and populations. If any of these processes is limited the species will decrease its colonization capacity (cf. Hermy & Verheyen, 2007). The poor colonization capacity of many forest herb species has been ascribed to poor seed dispersal ability (Hermy et al., 1999; Graae, 2000; Verheyen et al., 2003; Brunet, 2007). Three major arguments have been used: 1) many of the poorly colonizing species lack adaptation to long-distance dispersal (Matlack, 1994; Graae, 2000; 2002; Matlack, 2005), 2) long distances between forest habitats in the temperate agricultural landscape limit colonization (Brunet, von Oheimb & Diekmann, 2000); and 3) seed sowing experiments showed that recruitment often takes place when seeds are added (Eriksson & Ehrlen, 1992; Graae, Hansen & Sunde, 2004; Van der Veken et al., 2007). These are all good indications of limited seed dispersal, but there are indications that limited seed dispersal cannot solely account for the poor colonization capacity.

More than one hundred years ago Clement Reid was already wondering how oak could migrate so rapidly during the early Holocene considering its slow spread at present. This question has been termed Reid's Paradox (Clark et al., 1998). As the colonization processes in the early Holocene seem to have been much more efficient compared to present observed colonization rates, we may expect that some environmental difference(s) between the past and the present affect the colonization process. Based on the assumed importance of seed dispersal, former higher density of large herbivores may be suggested. However, even species without adaptations to zoochorous dispersal and with poor colonization capacity in the present landscape spread over large distances in Northern Europe during the early Holocene. Furthermore, while isolation in the fragmented landscape seems to limit current colonization, even migration rates within forest stands seem to be very slow (Brunet & von Oheimb, 1998; Bossuyt, Hermy & Deckers, 1999). Hence, more spatial continuity, such as dispersal corridors, would most likely not increase dispersal distances and colonization capacity to that observed in the early Holocene. Finally, the seed sowing experiments carried out with forest herbs demonstrated that when seeds are present some species will recruit, but these experiments also demonstrated notable lack of recruitment in many species, low numbers of recruited seedlings in other species (Eriksson & Ehrlen, 1992; Ehrlén & Eriksson, 2000; Graae, Hansen & Sunde, 2004), and low establishment of viable adult plants and populations (Ehrlen et al., 2006; see however Van der Veken et al., 2007). As Clark et al. (2007) pointed out, recruitment of a few seedlings does not guarantee individual or population establishment and spread. Recruitment limitation can either be ascribed to the number and quality of the seeds or the microsites. Studies on soil variables in occupied and non-occupied forest stands have tested whether poor colonization capacity could be due to microsite quality and have found little support for this explanation (Graae, Sunde & Fritzbøger, 2003; Verheyen et al., 2003).

As the climate in the early Holocene was warmer than today and climate is known to affect reproduction of many species, we suggest acknowledging that climate affects not only the distribution of habitats but also the intrinsic colonization capacity of forest herbs (Ibanez et al., 2007).

Few studies have tested the impact of climate on reproduction of forest herb species, despite the fact that laboratory studies have demonstrated that several processes in the reproductive pathway are temperature related (Baskin & Baskin, 1998; Gutterman, 2000; Probert, 2000; Finch-Savage & Leubner-Metzger, 2006). If climate is an important variable in colonization capacity, climate change may contribute to explaining Reid's Paradox; it is also important to incorporate the impact of climate on recruitment processes of field species when modeling plant distribution in future warmer climates.

In the present study we focus on the first element in the colonization process expected to relate to climate: seed formation. Seed formation results in seeds of varying quality, which can be measured as the germination percentage of the seeds. Germination percentages are time consuming to obtain, however, and it is not always known how to treat the seeds to get the highest germination percentage. Therefore, an easier measurable trait reflecting germination percentage would be convenient as a surrogate trait. Seed mass could be used as such a soft trait (sensu Jeltsch et al., 2008) if germination percentage and seed mass correlate closely. From a climate change perspective, the germination percentage obtained under optimal temperature treatment is relevant, but the climatic requirements for breaking seed dormancy are crucial. Many forest species have demanding requirements for break of seed dormancy and low germination percentages when tested in incubators (Grime et al., 1981), and Verheyen et al. (2003) demonstrated that slow-colonizing species have more restrictive germination requirements than fast-colonizing ones. As the requirements for break of seed dormancy vary with the environment affecting the maternal plant during seed formation (Baskin & Baskin, 1998; Cavieres & Arroyo, 2000; Ellison, 2001), we looked at this trait as a third seed quality measure.

In this study, we investigate how germination percentage, seed mass, and dormancy-breaking requirements change for seeds collected along a 1400-km-long latitudinal (10°) gradient from northern France to central Sweden. Mean annual temperature decreases 3.7 °C from north to south along this transect, thereby presenting a clear climatic gradient. Using a set of 11 long-lived forest herbs that according to literature differ in their colonization abilities, we specifically explored the following:

Methods

STUDY SPECIES

Eleven herbaceous species common in closed forests and representing a broad range of colonization abilities were selected for this study: Anemone nemorosa, Lamiastrum galeobdolon, Melica uniflora, Mercurialis perennis, Oxalis acetosella, Stachys sylvatica, Stellaria holostea, Brachypodium sylvaticum, Carex sylvatica, Circaea lutetiana, and Geum urbanum. All are common in northwestern and central Europe. The first 7 have adaptations to dispersal by ants or do not have any attachments that aid dispersal, while Brachypodium sylvaticum, Carex sylvatica, and especially Circaea lutetiana and Geum urbanum possess structures for epizoochorous seed dispersal (Hodgson et al., 1995).

Seeds of Brachypodium sylvaticum, Carex sylvatica, Geum urbanum, Lamiastrum galeobdolon, Mercurialis perennis, Oxalis acetosella, and Stachys sylvatica have physiological dormancy, whereas Anemone nemorosa seeds have morphophysiological dormancy, in which both growth of the embryo and break of the physiological inhibiting mechanism are needed before germination can occur (Baskin & Baskin, 1998). We found no information on dormancy in seeds of Circaea lutetiana, Melica uniflora, and Stellaria holostea, but other members of these genera have physiological dormancy (Baskin & Baskin, 1998).

For each species, a colonization capacity index (CCI) was calculated following Verheyen et al. (2003). In addition to data from the studies in Verheyen et al. (2003), data from 7 recent publications (Heinken, 1998; Graae, 2000; Graae & Sunde, 2000; Jacquemyn, Butaye & Hermy, 2001; Wulf, 2003; Brunet, 2004; Kolb & Diekmann, 2004) were used in the calculation of the CCI. The data met the same criteria as given in Verheyen et al. (2003), and the same methods were used for calculating the CCI (but with opposite sign, cf. Hermy & Verheyen, 2007). The formula therefore becomes where

R^* is the number of studies in which species X is significantly more frequent in recent forest; R is the number of studies in which species X is equally frequent or tends to be more frequent in recent forest; A^* represents the number of studies in which species X is significantly more frequent in ancient forest; and A is the number of studies in which species X is more frequent, but not statistically significant, in ancient forest. For studies lacking statistical analyses of species affinity with ancient forest, we calculated Pearson Chi² association measures. The index ranges from 100 (strongly associated with recent forest) to -100 (strongly associated with ancient forest). CCI varied from 29 to -84. Geum urbanum is the best colonizer (having the positive value), while the other species have values between -15 and -84 (Appendix I).

DATA ANALYSIS

An Immediate Germination Index (IGI) was constructed as follows:

where W is the germination percentage of fresh seeds, CW is the germination percentage after cold stratification followed by warm incubation, and WCW is the germination percentage after warm incubation followed by cold stratification and warm incubation (W is subtracted to exclude the seeds germinated as fresh seeds). Hence, a high IGI indicates high germination percentage without any stratification treatment, and a low value indicates that most seeds needed both warm and cold stratification before germination occurred.

For the among-species analyses, mean values for the seed quality variables of each of the 11 species across all samples were calculated. As the number of species was low and data were not normally distributed, Spearman Rank correlations were used to test for relationships between the CCI, seed mass, germinability, and the IGI.

To test for effects of climate and latitude on seed mass within species, a General Linear Model (GLM) was applied. Therefore, GDH and latitude were included as covariates, and possible differences in effects of GDH and latitude between faster- and slower-colonizing species were tested through inclusion of the interactions with the factor colonization group. The random factor colonization group was defined to have 2 levels, slow-colonizing species (CCI < -50; 6 species) and faster-colonizing species (CCI > -50, 5 species). To allow a joint statistical analysis of species, seed mass values were standardized for each species by dividing each individual seed mass by the mean seed mass of that species across all sites. Similarly, GDHs were standardized per species by dividing site-specific values by the overall mean to be able to compare differences in GDHs over all sites and all species. Finally, log₂+1-transformation of the seed mass was necessary to meet the statistical assumptions.

To test for effects of climate and latitude on germinability, a similar GLM was performed on the log2+1-transformed standardized germinability (cf. above). To test whether patterns in germinability were caused by correlation with seed mass, a third model was run, but with seed mass included as a covariate.

The effects of climate and latitude on the requirements to break dormancy were again tested using a similar GLM, but this time with IGI as response variable.

Appendix I shows all data obtained in the experiment and used in the data analysis listed for the species in the order of increasing colonization index CCI. Due to a shortage of seeds and a mistake in the laboratory procedures during the germination trials the resulting sample size was decreased slightly in the different analyses. All analyses were performed with SPSS 15.0.

Results

In the among-species analyses, the CCI was not significantly correlated with any of the seed quality measures (r_s < 0.45, P > 0.17, n = 11 for all 3 variables). Germinability decreased significantly with seed mass (r_s = -0.75, P < 0.01, n = 11) and increased with IGI (r_s = 0.66, P < 0.05, n = 11), whereas the correlation between IGI and seed mass was only marginally significant (r_s = -0.53, P = 0.10, n = 11). Hence, large-seeded species tended to have lower germinability and stronger requirements to break dormancy.

Standardized seed mass within species increased significantly by latitude but not significantly by standardized GDH (Table I). Interactions with CCI groups were not significant. Standardized germinability was marginally significantly correlated with latitude and standardized GDH (Table II). Interactions between the CCI groups and both the standardized GDH and latitude were significant. Only the slow-colonizing species increased standardized germinability significantly with standardized GDH (R² = 0.17, P = 0.016, see Figure 1a). The back-transformed germination percentage indicates that an increase in the GDH of 20% increased the germination percentage by 30 to 33% for the group of slow-colonizing species. Only the fast-colonizing species increased standardized germinability with latitude (R² = 0.14, P = 0.049, see Figure 1b). The back-transformed germination percentage indicates that an increase in latitude of 1° increased the germination percentage by 6 to 9% for the group of better-colonizing species.

Standardized seed mass and germinability were significantly positively correlated (r = 0.35, P < 0.01, n = 61), with the variability in seed mass (range 40–160% of the average mass) being smaller than the variability in germinability (range ∼0–300% of the average germination percentage). If the GLM is run with standardized germinability as dependent variable and standardized seed mass included as explanatory variable, standardized seed mass also explains a significant part of the variation (F_{1, 53} = 5.80, P < 0.05), leaving the latitude (F_{1, 53} = 1.87, P = 0.18) and GDH (F_{1, 53} = 1.68, P = 0.2) insignificant in the model, whereas the interactions with CCI remain significant in the model (F_{1, 53} = 4.58 and 4.48 for standardized GDH and latitude, respectively, and P < 0.05 for both).

There was no impact of latitude or standardized GDH on dormancy breaking requirements (Table III).

Discussion

In the present study, slow colonizers did not have particularly poor germinability or complicated break of dormancy patterns compared to fast colonizers, but the results demonstrated that the slow-colonizing species increased their number of germinable seeds when growing in a warmer climate. In contrast, the better colonizers increased their germinability with latitude rather than with temperature. Although germinability was correlated with seed mass within species, germinability seemed more strongly affected by temperature than seed mass did. The implications of these results are discussed below.

The small number of species in the present study obviously limited the chances of observing interspecific differences. We saw no significant pattern of poor colonizers having low germinability compared to that of the faster colonizers. However, large-seeded species tended to have lower germinability and stronger requirements to break dormancy. Large seed size has been demonstrated to be associated with poor colonization capacity in forest herb species (Graae, Hansen & Sunde, 2004), and poor colonizers among forest species have been demonstrated to have higher requirements to break of dormancy (Verheyen et al., 2003). Our results do not support these findings, but we suggest that studies with more species to be carried out to reveal the interspecific patterns of reproductive capacity of poor compared to fast colonizers.

TABLE I.

Results of the GLM analysis with the log₂+1-transformed standardized seed mass as response variable and latitude, the standardized growing degree hours (Stand. GDH), and their interactions with the colonization group (2 levels: slow- and fast-colonizing species) as explanatory variables (n = 60; R² = 0.09).

TABLE II.

Results of the GLM analysis with the log₂+1-transformed standardized germinability as response variable and latitude, the standardized growing degree hours (Stand. GDH), and their interactions with the colonization group (2 levels: slow- and faster-colonizing species) as explanatory variables (n = 61; R² = 0.17).

FIGURE1.

Relationships between the log₂+1-transformed standardized germinability and a) the standardized growing degree hours and b) the latitude for both the slow-colonizing species (filled circles, full line; R²_a = 0.17 and R²_b = 0.01) and the fast-colonizing species (empty triangles, dashed line; R²_a = 0.01 and R²_b = 0.14).

TABLE III.

Results of the GLM analysis with the immediate germination index (IGI) as response variable and latitude, the standardized growing degree hours (Stand. GDH), and their interactions with the colonization group (2 levels: slow- and fast-colonizing species) as explanatory variables (n = 59; R² = 0.09).

Analyzing the intraspecific variations in all species, latitude was more important for seed mass than temperature was. For many species, seed mass decreases with latitude (Moles & Westoby, 2003), but we found a positive correlation between seed mass and latitude within the species. The many variables that have been demonstrated to correlate with latitude (Jansson & Davies, 2007; Moles et al., 2007) leave various options for interpretation of the results. Seed mass may be determined by the environment during seed development. One may argue that colder temperatures towards the poles mean shorter time for seed development, and hence smaller seeds, compared to lower latitudes. On the other hand, seeds are set later at higher than at lower latitudes, and the increased light regime during the time of their development may be more important than temperature and the length of the growing season for seed development. Increased seed mass with increased latitude may also be related to selective forces such as requirements for rapid seedling establishment in less favourable climates with shorter growth seasons. Increased seed mass for species in increasingly physiologically stressful environments has also been reported by Caddick and Linder (2002).

The most intriguing result of the present study was that slow-colonizing species increased their germinability with warmer climate. This was unrelated to the effect of latitudinal gradient on seed size. These results indicate that species that are today's poor colonizers could become better colonizers in warmer climates, because increased germinability may favour recruitment ability of their seeds. This result also adds interesting aspects to the old debate on how species that are today poor colonizers showed fast migration in the early Holocene. An increased recruitment rate in the warmer early Holocene climate (Kullman, 1998) through increased germination may have added to the fast population spread (Ibanez et al., 2007). The relative importance of such processes for migration rates of slow colonizers has yet to be tested.

Another very interesting pattern in the data was that fast colonizers showed significantly increased germinability with latitude, whereas no such trend was obvious for the slow colonizers. As this pattern was not caused by the GDH, other variables related to latitude may drive the pattern. The observed pattern could be related to the increasing day length during seed set or to the openness of forest canopy where the species occur, which also may increase with latitude. The fast-colonizing species (Geum urbanum, Brachypodium sylvaticum, Mercurialis perennis, Stachys sylvatica, and Circaea lutetiana) differ from the slow colonizers in several ways. Most of the fast-colonizing species have late seed set, and they are often found in more open forests compared to slow colonizers (Anemone nemorosa, Oxalis acetosella, Lamiastrum galeobdolon, Melica uniflora, Stellaria holostea, and Carex sylvatica). Perhaps the fast-colonizing species are more favoured by or dependent on long days with plenty of light for seed development compared to the slow ones, which are adapted to early seed set where at least part of the flowering is before the canopy closes.

When evaluating intraspecific differences we found that low seed size for a species is associated with low germinability. This supports the notion that seed size can be used as a good soft trait for germinability. On the other hand, the relationships between germinability on the one side and GDH and latitude on the other did not change when seed mass was entered as covariate. This indicates that although seed mass and germinability were correlated, the patterns in germinability with respect to colonization capacity could only be revealed by evaluating germination of the seeds. Hence, studies on germinability of seeds may reveal information not revealed in studies where only seed mass is analyzed, and there is certainly a need to know more about the relationship between seed mass and germinability if we are to use seed mass as a valuable soft trait in analyses, such as migration and population expansion modelling.

Numerous other climate-related variables are important to reproductive success in addition to seed formation. Density of established vegetation, herbivory, litter depth, soil nutrients, and microclimate during dormancy breaking, germination, and seedling recruitment may affect seedling emergence. There is an urgent need to increase our knowledge of climate impacts on recruitment processes to understand migration processes of different plant functional types.