STUDY SITES AND SAMPLING

The study was carried out in the subarctic Kilpisjärvi area (69°01′N, 20°50′E), in northernmost Finland, during 1995–1998. The mean annual precipitation of the region is 422 mm, annual temperature is −2.6°C, and the length of the growing season is 100 d (Järvinen, 1987). Treeline is formed by mountain birch (Betula pubescens ssp. czerepanovii) at 600 m a.s.l. Eight common subarctic plant communities, at least 50 m apart, were chosen for the study (Table 1). In each plant community, 12 seed bank samples were randomly taken with a soil core (15 cm diameter × 5 cm height) close to the transects in the first week of September 1997. At that time, most current-year seeds had not yet dispersed. Hence, we assumed most seeds to be at least 1 yr old and to represent persistent seed bank (Thompson et al., 1997). The samples were stratified and germinated twice after spreading them as a 0.5-cm-thick layer over autoclaved soil of mixed sand and gravel. The first moist stratification was done for 2 mo at 5°C in a dark, cold room (Staniforth et al., 1998), and the first germination was done in the greenhouse at 16 h of daylight (400-W lamps, 1 m apart, placed 1 m above the samples) and in 8 h of twilight (60% of the lamps switched off) for 8 mo. The twilight was used to simulate night conditions during subarctic summer. The temperatures varied diurnally from 15°C to 24°C, which ensured maximum germination during the experiment. Such temperature fluctuations facilitate the seed germination of many species (Sayers and Ward, 1966; Thompson and Grime, 1983). The second moist stratification occurred at 5°C in a dark, cold room for 4 mo, after which the second seed bank germination was carried out for 3.5 mo. The light regimen and the daytime temperature of the second germination period were similar to the first germination period, whereas the minimum temperature was now 10°C. The samples were watered regularly during the germination periods. Seedlings were counted every second day and identified according to species.

In each plant community, 3 parallel 7-m-long transects were established 3 m apart. At each transect, 4 0.8 × 0.8–m squares were placed 1.3 m apart. Field seedlings of 0–3 yr old were counted, and the cover of adult vascular plants was measured in the squares during the pediod from mid-July to August in 1995 and 1996. The developmental stage of field seedlings was estimated as proposed by Wager (1938) and Freedman et al. (1982). The seedlings and adult plants were identified according to species except for 5 genera: Alchemilla spp., Antennaria spp., Cerastium spp., Pyrola spp., and Luzula spp., the seedlings of which were classified to genus. Current-year and older seedlings were not distinguished from each other. Hence, the term “recruitment” indicates all seedlings counted in the field. Vegetation cover was estimated at 5% intervals except for the cover values under 10%, which were classified in the following categories: 0.5%, 1%, 5%, and 7%.

The species were divided into 4 regenerative groups based on their abilities to reproduce vegetatively (Söyrinki, 1938; Table 2). Diaspore morphology and seed weight classes were determined from seeds collected mostly from Kilpisjärvi during 1996–1998 (Table 2). For the determination of seed weight classes, 100 seeds per species were dried at 80°C for 3 d. Seed weights were measured after removal of the appendages.

STATISTICS

The seed bank densities were compared between plant communities using the one-way ANOVA and the Dunn-Šidak post hoc test. Square-root transformation was used in the analyses. Within each plant community, the seed bank density was compared with the field seedling density using the t-test. Three correlations were calculated using Spearman's rank correlation: (1) between the seed bank densities and the altitude; (2) between the seed bank densities and the relative proportions of plants of different regenerative groups, seed sizes, and morphological characters in standing vegetation; and (3) between the field seedling densities and the relative proportions of plants of different regenerative groups, seed sizes, and morphological characters in the seed banks. The SPSS 8.0 Package was used for statistical analyses.

The floristic similarity between the seed banks and the standing vegetation and between the seed banks and the field seedlings across the plant communities was tested using a nonmetric multidimensional scaling (PC-ORD Package 3.0). The empty samples were deleted before the analyses. Jaccard's similarity coefficient was chosen as a distance coefficient. To analyze the floristic similarity between seed bank, field seedling flora, and standing vegetation, qualitative Jaccard's similarity coefficient (Jongman et al., 1987; Welling and Laine, 2000b) and Spearman's correlation analysis (Chambers, 1993; Booth and Larson, 1998, 2000; Welling and Laine, 2000b) were used. The equation for the qualitative Jaccard's similarity coefficient is SJ = c/(A + B − c), where A = the total number of species in the seed bank, B = the total number of species in the field seedlings or in the standing vegetation, and c = the number of species shared by the seed bank and the standing vegetation or the field seedlings. Only the species occurring in both the seed bank and the field seedling flora were included in the Spearman's correlation analysis (Chambers, 1993; Booth and Larson, 2000). To examine whether the similarity between seed bank and field seedlings differed between plant communities, the one-way ANOVA and the Tamhane's post hoc test were used. The Tamhane's test is appropriate when the variances are unequal. To compare the floristic similarity between the seed bank and the standing vegetation and between the seed bank and the field seedlings, the t-test with the Dunn-Šidak correction was used. Spearman's correlation analysis was used to investigate correlations between Jaccard's similarity coefficients and the relative proportions of the plant regenerative groups in the standing vegetation and in the seed banks.

THE SEED BANK AND FIELD SEEDLING DENSITIES

The seed bank density (number of seedlings emerging from the seed bank samples) ranged from 99.1 to 1108.8 seedlings m²⁻¹ (Fig. 1). The density was larger in the meadows (EUME, TALL), two snowbeds (EUBED, SAL), and the heaths of the lower slopes (MYRT) compared with RANU and the heaths of the upper slopes (CASS, GRHE). Six species, Carex bigelowii, Cassiope tetragona, Gnaphalium supinum, Phyllodoce caerulea, Sibbaldia procumbens, and Veronica alpina, formed about half, 49 ± 11%, of the seed banks (Appendix 1). Most seeds of the seed bank samples germinated during the first germination period. However, several hundred seeds m^2–1, mostly sedges and rushes, also germinated during the second germination period. As expected, there was a negative correlation between the seed bank densities and altitude (R_s = −0.810, n = 8, P < 0.05). The correlation between the seed bank densities and the relative proportion of the smallest seeds (weight class 1) in the standing vegetation was also negative (R_s = −0.734, P < 0.05). The main reason for this pattern was the occurrence of Cassiope hypnoides, a species with small seeds and no germinated seeds in the seed bank. When this species was deleted from the analysis, the correlation disappeared (R_s = –0.048, n = 8, ns). The correlations between the seed bank densities and seed weight classes and diaspore morphology classes were statistically negligible. Except for RANU, the difference between seed bank and field seedling densities was significant (P < 0.05, Fig. 1).

An increase in the relative proportions of species with an ability for ineffective vegetative reproduction (VI) and diaspores with pappus (diaspore morphology class 2) increased the field seedling densities (Table 3). In contrast, an increase in the relative proportions of species with an ability for effective vegetative reproduction (VE) and heavy seeds (seed weight class 4) decreased the field seedling densities (Table 3).

FLORISTIC SIMILARITY ACROSS THE PLANT COMMUNITIES

Two dimensions were appropriate to characterize the data. In the ordination of the seed bank, the first axis separated the meadows (EUME, TALL) and two snowbeds (EUBED, SAL) from the heaths (MYRT, CASS, GRHE, Fig. 2a), and the second axis separated TALL from EUME and the snowbeds. In the ordination of the field seedlings, the first axis separated the dwarf shrub heaths (CASS, MYRT) from GRHE and RANU, TALL from the low herb communities (EUME, EUBED), and the low herb communities from SAL (Fig. 2b). The second axis separated the meadows, EUBED and SAL, from the heaths and RANU. In contrast to the pattern of the seed banks, there were several distinctly separated groups in the field seedlings: the group of MYRT and CASS; the group of TALL; the group of GRHE and RANU; and the group of EUBED, EUME, and SAL. In the ordination of the standing vegetation, the first axis clearly separated the dwarf shrub heaths from the meadows, snowbeds, and GRHE (Fig. 2c). There were three distinctly separated distant groups in the ordination: the group of MYRT and CASS; the group of RANU and SAL; and the group of GRHE, EUBED, EUME, and TALL. The species richness was generally low (range from 5 to 18 species, Appendix 1).

FLORISTIC SIMILARITY WITHIN THE PLANT COMMUNITIES

The correlation between the seed bank and the field seedlings was significant only in EUBED (R_s = 0.829, n = 6, P < 0.05). The floristic similarity was generally low between the seed bank and the field seedlings, ranging from 0.13 to 0.39. The difference of the floristic similarities between the plant communities was statistically significant (one-way ANOVA, F = 3.029, df = 7, P < 0.01), but only the pairwise difference between EUBED (0.39) and TALL (0.13) was significant.

There was no correlation between the standing vegetation and the seed bank. The floristic similarity between the seed bank and the standing vegetation was generally low, ranging from 0.05 to 0.24. Despite the difference between the plant communities (one-way ANOVA, F = 4.105, df = 7, P < 0.001), only EUBED (0.22) differed significantly from TALL (0.09) and RANU (0.05).

There were few statistically significant differences between the Jaccard's floristic similarity coefficients between seed bank, field seedling flora, and standing vegetation (Fig. 3). The similarity between the seed bank and the standing vegetation was significantly lower than that between the seed bank and the field seedlings in EUME, EUBED, SAL, and RANU. Furthermore, the standing vegetation and the field seedlings were significantly more similar than the standing vegetation and the seed bank only in EUME and RANU. There was a correlation in the floristic similarity between the seed bank and the field seedlings and the relative proportions of the plants with an ability for ineffective vegetative reproduction (VI) in the seed bank (Rs = 0.810, n = 8, P < 0.05, Table 3).

Most taxa that occurred only in the standing vegetation were trailing dwarf shrubs with effective vegetative reproduction (Salix spp., C. hypnoides), grasses with a low ability for vegetative reproduction and appendaged diaspores (e.g., Poa alpina, Phleum alpinum), herbs characterized by appendaged diaspores (Antennaria spp., Hieracium spp., Petasites frigidus, Oxyria digyna), and herbs characterized by heavy seeds (e.g., Pedicularis lapponica, Solidago virgaurea, Trientalis europaea; Table 4, Appendix 2). Polygonum viviparum, which reproduces primarily by bulbils, was also in this group. The taxa that were abundant in the standing vegetation were usually those with an ability for effective vegetative reproduction. Appendaged diaspores (Betula spp., Taraxacum spp., Vaccinium spp.) and small seeds (Luzula spp., Juncus trifidus) were common traits in the taxa that occurred only in the seed bank, and most of the taxa that occurred only among the field seedlings had appendaged diaspores (e.g., Antennaria spp., Betula spp., Vaccinium spp., Saussurea alpina). Agrostis mertensii is a narrow-leafed grass that could be overlooked in the cover estimations at TALL. Most taxa that occurred in the seed bank but not among the field seedlings were sedges, erect dwarf shrubs, a rush characterized by small seeds (Juncus trifidus), and herbs characterized by appendaged diaspores or by both large seeds and appendages (e.g., Taraxacum spp., Antennaria spp., Ranunculus spp., and Oxyria digyna). Finally, most taxa that occurred among the field seedlings but not in the seed bank were herbs characterized by large seeds (commonly above 0.5 mg) or both large seeds and appendages (e.g., Astragalus alpinus, Saussurea alpina, Potentilla crantzii, Rumex acetosa, Solidago virgaurea, Trollius europaeus) and grasses with appendages and small seeds (Deschampsia flexuosa, Festuca ovina, Anthoxanthum odoratum).

SEED BANK DENSITIES

Before this investigation, very low seed bank densities were found in the alpine area of the Fennoscandian subarctic region (Diemer and Prock, 1993; Molau and Larson, 2000). The seed bank densities in Kilpisjärvi are more like those in alpine areas of more southern regions (Hatt, 1991; Diemer and Prock, 1993; Chambers, 1993), subarctic forests of northern Finland (Vieno et al., 1993) or arctic regions (e.g., Fox, 1983; Leck, 1980; Freedman et al., 1982; Roach, 1983). Although not investigated in our study, one reason for the surprisingly high densities in Kilpisjärvi may be intensive grazing by reindeer and rodents. Grazing facilitates the colonization of low-statured herbs and graminoids (Virtanen et al., 1997), which formed the majority of the seed banks, e.g., Veronica alpina, Sibbaldia procumbens, Gnaphalium supinum, and Luzula spp. These species are mostly absent in the seed banks of northern Sweden (Molau and Larson, 2000). Our observations supported earlier findings of decreasing seed bank densities toward higher altitudes (Thompson, 1978; Molau and Larson, 2000). The reason may be in the decreasing seed production and seed quality in the higher altitudes (Laine et al., 1995). On the other hand, the denser vegetation on the lower slopes limits seedling recruitment and seemed to negate the influence of the altitude on the field seedlings (see also Welling and Laine, 2000a).

Plant traits, i.e., the dominant seed size and diaspore morphology and the ability of vegetation to produce vegetatively, should have an influence on seed bank densities (Freedman et al., 1982; Chambers, 1993). However, the relationships were not clear in this study, probably because none of these traits clearly dominated in standing vegetation. Small- and intermediate-seeded taxa can be relicts from the period when they occurred in the standing vegetation. Luzula spp. and C. bigelowii have a seed bank that may persist over 200 yr in the soil (McGraw et al., 1991; Thompson et al., 1997). However, small seed size alone does not guarantee accumation in the seed bank if the seeds are physiologically short-lived, as in C. tetragona and Vaccinium spp. (Thompson et al., 1997) or Salix spp. (Beerling, 1998). Also, larger seeds may sometimes live longer than smaller seeds in the seed bank, as in certain Carex spp., where species with large seeds remain dormant, while germination percentage is significantly higher for smaller seeds (Schütz and Rave, 1999; Schütz, 2000). Considering diaspore morphology, appendages are assumed to facilitate dispersal but have a negative influence on the seed bank accumulation ability (Rabinowitz, 1981). For example, in Gnaphalium supinum, which has diaspores with pappus, the positive influence of small seed size on seed burial seemed to overwhelm the negative influence of pappus. The appendaged diaspores of this species were common in the seed bank, although the species was uncommmon in standing vegetation. On the other hand, the high number of diaspores of Gnaphalium supinum may have partially been an outcome of long-distance dispersal.

FLORISTIC SIMILARITY ACROSS THE PLANT COMMUNITIES

According to nonmetric multidimensional scaling, the seed banks were not as clearly separated into plant communities as the field seedlings and the standing vegetation. One reason was the low species richness of the seed banks. Morgan (1998) and Holmes and Cowling (1997) found the same phenomenon in grasslands and fynbos shrublands. In this study, the poverty and similarity of the seed banks between plant communities may be based on the temporally dynamic nature of seed banks (Simpson et al., 1989) and the importance of Carex spp. in all seed banks. At the time of sampling, the seed banks were probably at their most depleted state since most current-year seeds had not yet dispersed and many previous-year seeds had germinated in the field or lost their viability. This is why the species richness in the seed banks was considerably lower than in the standing vegetation. Carex bigelowii, a common species of almost all plant communities, forms a persistent seed bank (McGraw et al., 1991; Thompson et al., 1997), which facilitates the accumulation of large seed banks also in the communities where the coverage of C. bigelowii is low.

FLORISTIC SIMILARITY WITHIN THE PLANT COMMUNITIES

A majority of the species that occurred in standing vegetation were not found in the seed bank or among the field seedlings. Seeds of these species may have been consumed by herbivores or otherwise destroyed. Also, the small size of the first leaves may have limited the seedling recruitment of graminoids. Large seeds have a contrasting performance at seed bank, field seedling, and standing vegetation stages. Large seeds accumulate poorly in the seed bank due in part to seed predation (Thompson et al., 1998; Eriksson and Jakobsson, 2000), but their abundant nutrient resources increase their germination rate and seedling survival, especially in a stressed environment (Kiviniemi, 1999; Eriksson and Jakobsson, 2000). Despite a relatively high density of large seeds in the seed bank in this study, seedling recruitment was low. A thick moss cover may have reduced the germination of large seeds in the heaths where they were common.

Many clonal plants accumulate a seed bank, but seedling recruitment is rare (Eriksson, 1992). This is probably why there was a negative correlation between the relative proportion of plants with an ability for effective vegetative reproduction in the seed banks and the field seedling densities. On the other hand, species characterized by an ineffective vegetative reproduction ability were effective in both seed bank accumulation and seedling recruitment. Li et al. (1999) demonstrated the deviation in these abilities within the Rhus genus. Both weakly clonal and strongly clonal species have persistent and large seed banks. However, seedling recruitment is effective in the former, but not in the latter species.

Grazing may also have reduced the floristic similarity between the seed bank and the standing vegetation, since the low-statured species that benefit from grazing (Virtanen et al., 1997) were abundant in the seed banks but had a low coverage in the standing vegetation.

The changes in the floristic similarity from seed bank to field seedlings and from field seedlings to standing vegetation were usually equally critical. There was a significant correlation between the seed bank and the field seedlings only in EUBED. Thus, there are restrictions at every studied phase along the regeneration pathway. Eutrophic and moist conditions provide suitable conditions for field seedlings in EUBED. In this community, seed number (seed availability) determines the structure of the field seedlings; studies in the Patagonian steppe have led to similar findings (Aguiar and Sala, 1997). In TALL, tall and closed standing vegetation resulted in the poor regeneration via the field seedling recruitment and therefore the low floristic similarity between the standing vegetation and the seed bank. In RANU, the transient seed bank of Ranunculus glacialis (Thompson et al., 1997) and very low species richness are probably the main factors that resulted in the low floristic similarity between the standing vegetation and the seed bank. These factors were reasons for the clearly lower floristic similarity between the standing vegetation and the seed bank in TALL and RANU than in EUBED, as well as for the clearly lower floristic similarity between the seed bank and the field seedlings of TALL than in EUBED.