Seed bank dynamics in abandoned meadows during a 20-year period in the Białowieża National Park



1 The relationship between size and floristic composition of the seed bank and vegetation dynamics was studied between 1976 and 1996 in 0.75 ha of an abandoned Cirsietum rivularis meadow. The plot was divided into 100-m2 (10 × 10 m) quadrats and sampled 5-yearly to map the vegetation and determine the soil seed bank.

2 Densities of seeds in the soil fluctuated as succession proceeded. The initially small seed bank trebled by 15 years after abandonment, before falling, after 20 years, to approximately the same as in the initial stage.

3 The floristic richness of the seed bank decreased during succession, with the number of species falling from 38 to 25. The diversity of life forms, however, increased in later periods, with tall herbs, shrubs and trees appearing after 10 years.

4 Seed bank floristic composition is apparently both a product of the species composition of the current vegetation and a record of the long-term substitution of species. Other factors, including the structure of the vegetation, also influence the accumulation of seeds in the soil.

5 Although changes in number of species show a directional pattern, the seed bank size fluctuated in the course of succession on these fertile wet meadows.


Mechanisms of secondary succession are assumed to involve the bank of seeds in the soil, and thus to depend on the different patterns observed for seed banks. For example, succession on fallow land (a habitat markedly transformed by humankind) has been described in terms of directional change, in which the floristic richness of the seed bank declines between the initial and terminal stages, in parallel with the density of seeds (Pickett & McDonnell 1989; Roberts & Vankat 1991). However, succession in semi-natural grassy or meadow communities tends to show other types of change (Donelan & Thompson 1980; Patridge 1989; Milberg 1995) in which the greatest density of seeds in the soil occurs during the temporary (herb-dominant) stage (Oosting & Humphreys 1940).

Many of the patterns described for seed banks fail to reflect the species composition of the above-ground vegetation (Leck et al. 1989). This is often linked with a failure of the seeds of many species to persist in the seed bank (Thompson 1992) but may also be due to some species present in a seed bank failing to be detected by the particular method used (Gross 1990; Bakker et al. 1996; Ter Heerdt et al. 1996; Thompson et al. 1997). Long-term simultaneous and multi-faceted research on the dynamics of both the vegetation and the seed bank are therefore needed to understand the relationships between the species compositions. The influence of the seed bank on the course of succession is also dependent on the relative importance of the bank of seeds present in the soil as the process is initiated, compared with the subsequent influx of seeds from the exterior.

The aim of the present research was thus to determine the pattern of the seed bank during secondary succession from meadow to forest, as part of a series of studies on vegetational change in abandoned meadows carried out in Polanłowieża Primeval Forest between 1972 and 1996. Correlation with data on the dynamics of the vegetation and the demography of species participating in succession (Falińska 1989, 1991) should allow for a fuller recognition of the role of the seed bank in secondary succession, by determining the degree to which changes in floristic composition are reflected in the species composition of the seed bank. In particular, we considered whether seed banks differ in particular successional stages or include the same species throughout development from meadow to willow scrub.

We also determined whether species that play a significant role in vegetational development can be absent from or only weakly represented in the seed bank, and if species occurring only sporadically in the vegetation, or altogether absent from it, can account for a significant proportion of the seed bank.

In essence, we investigated whether the composition of the seed bank is wholly or partially determined by the floristic composition of the vegetation or is dependent on random events.

Similarities between vegetational floristic composition and the composition of a seed bank have usually been restricted to natural vegetations (Leck et al. 1989), including łowieża Primeval Forest communities (Falińska 1981, Falińska (1997; Pirożnikow 1983; Jankowska et al. 1998). It is more common for the correlation between the two elements to be weak (Hall & Swaine 1980; Pratt et al. 1984; Hakan et al. 1987; Roberts & Vankat 1991; Milberg 1992), and this is often ascribed to disturbances, with closer relationships found for more stable communities (Williams 1984).

Materials and methods

Study area

The seed bank was studied from 1976 to 1996 at a site within the Reski Range, a 15-ha area of abandoned meadow land that has been the subject of studies of plant demography and succession for 25 years (Falińska 1991, 1995, 1997). This site, which is located at the centre of thłowieża Primeval Forest, in the valley of the Narewka River, was deforested 200 years ago and, since 1996, the area has been part of łowieża National Park (52°42′32′′, 23°50′20′′E). The floristically rich meadows of the Cirsietum rivularis community occupy fertile and wet habitats that once supported floodplain forest (Circaeo–Alnetum). Prior to abandonment the meadows had not received applications of fertilizer or been grazed, and were only mown once a year (in June or July) because of the persistence of floodwater in spring and autumn.

Between 1972 & 1974 most of the Reski Range was divided into 10 × 10-m quadrats to allow mapping of vegetation at 5-yearly intervals, as well as observation of the spatial dynamics of plant populations. Detailed demographic studies have been based on 5 × 5-m plots marked out within these 100-m2 quadrats (Falińska 1991). Earlier research on plant population dynamics in meadows had shown that plots of this size meet the condition of representativeness in homogeneous communities of fine-grained structure (Falińska 1978). The meadows had not all been abandoned at the same time, and a final 50 × 150-m (0.75-ha) fragment, some 300 m from the forest edge, was not left until 1976, when it was divided into quadrats in line with the scheme adopted previously. This became the site for seed bank study.

To determine changes in vegetation over time, the 0.75-ha study area was mapped every 5 years at the beginning (March or April) and end (September or October) of the growing season, and the presence of species in each of the 25-m2 plots was also recorded. These data served as a basis for assessing the frequency of occurrence of species and their substitution over the 20 years.

Soil samples were collected during the same seasons for seed bank assessment using three different methods. An area 6 × 100 m2 was selected as representative of the characteristic floristic composition in the patch of meadow and divided into three 2 × 100-m plots (A1, A2 and A3), each comprising eight subplots of 25 m2 (5 × 5 m).

Material from A1 was analysed by seed separation methods, while A2 was used to follow seedling emergence in the greenhouse. Plots established within A3 were used to determine seedling emergence in the field, following removal of existing vegetation.

Plot a1: seed separation methods

Sampling to determine the total reserve of seeds in the soil was done after the end of every fifth growing season, so that seeds could be selected over the winter. Soil samples of surface area 100 cm2 and depth 3 cm were taken with metal containers, five from each of the eight subplots (total 40). Sampling sites were selected randomly so that the 25 samples taken over the course of the study came from all parts of the subplot. After removal of roots and plant fragments with the aid of special sieves, soil brought to the laboratory was examined for seeds under a microscope. Seeds were then identified using a collection formed between 1972 and 1980 as part of earlier work in the same area. Identifications were confirmed by germinating selected seeds in Petri dishes.

Plot a2: seedling emergence from soil samples in the greenhouse

Greenhouse work was based on 40 10 × 10 × 3-cm soil monoliths that were taken in spring, following a period of stratification. As in plot A1, samples were positioned at random and taken at 5-year intervals. After removal of roots and plant fragments, soil was placed in cuvettes in a greenhouse at a temperature of 18–22 °C, with soil humidity maintained using distilled water. Earlier studies on the germination of meadow and forest seeds (Falińska 1978, 1981) had shown that most seeds germinated at this temperature, which was close to natural conditions during the growing season. Samples were checked two to three times a week over 3 consecutive years for emerging seedlings. The fourth season was given over to preparing the greenhouse for the receipt of the next set of samples, as earlier seed bank studies had found germination in the fourth year to be little more than sporadic (Falińska 1981; Pirożnikow 1983).

Plot a3: seedling emergence in the field

In the year of meadow abandonment, and every 5 years thereafter, 10 1-m2 plots were marked off and in autumn (October) all plants were removed. Each time the plots were located in a different part of the 200-m2 area in order to avoid total destruction of the vegetation. Seedling emergence was monitored at 3-day intervals from the disappearance of snow cover in the following spring until the end of that growing season.

Statistical analysis

One-way analysis of variance was used to assess the relationship between the sizes of seed banks and different stage of succession. The significance of differences between different stages was tested using the Sandecora F-test.

The density of seeds in the soil was assessed on the basis of the mean number of seeds per 100 cm2 calculated from all samples, and then transformed into the value per m2. Densities per m2 for particular species were estimated from the mean value per m2 and the frequency with which the species occurred in the samples.

t-tests were used to assess the significance of differences between mean densities of seeds in different phases of succession, while the relationships between the species compositions of the vegetation and the seed bank were assessed using Spearman rank correlations. In the tables species are ordered according to increasing frequency or number of seeds (seedlings) in the first sample (i.e. the year of abandonment).


Change in the spatial pattern of the abandoned meadow vegetation

During the 20-year period the species composition of the original meadow community (Cirsietum rivularis) changed considerably. Meadow species retreated gradually, although many of them still occurred sporadically in the willow scrub after 20 years (Table 1). Five years after the abandonment of the meadow, macroforbs and sedges formed aggregations, and after 10 years many of these included forest herbs and young trees (Populus, Alnus, Betula). After 15 years swamp species were beginning to appear as dominant and subdominant species, e.g. Carex acutiformis, Carex cespitosa and willows such as Salix cinerea. After 20 years almost half of the 0.75 ha that was originally meadow was occupied by willows, while the remaining part was overgrown with the tall herbs Filipendula ulmaria, Lythrum salicaria and Lysimachia vulgaris and the sedges Carex acutiformis and Carex cespitosa. The meadow community, which at the initial stage had a fine-grain spatial structure, had been transformed into a mosaic of sedge, tall herb and willow patches (Fig. 1). Three stages could be distinguished in the dynamics of the abandoned meadow vegetation.

Table 1.  Comparison of the floristic composition of the vegetation and seed bank in the soil during succession. Frequency ● < 10; ● 11–50; ● 51–100%. * Frequency represents total seed bank for genera (Cirsium, Viola, Carex, Salix) or family (Poaceae). Data are from the seed separation method
 VegetationSeed bank
After years
Species Phases 0123401234 
Meadow species
Lychnis flos-cuculi  
Ranunculus acris   
Rumex acetosa     
Lathyrus pratensis    
Viola palustris  *
Ranunculus repens 
Caltha palustris    
Myosotis scorpioides     
Cirsium rivulare *
Cirsium palustre      
Geum rivale    
Polygonum bistorta   
Potentilla palustris    
Trifolium repens       
Galium palustre   
Cerastium holosteoides      
Filipendula ulmaria 
Lythrum salicaria 
Lysimachia vulgaris 
Juncus effusus 
Mentha arvensis  
Epilobium palustre     
Galeopsis tetrahit      
Ranunculus flammula       
Lotus corniculatus        
Polemonium caeruleum       
Cardamine pratensis      
Campanula patula        
Potentilla anserina     
Plantago lanceolata       
Geranium palustre      
Vicia cracca        
Phragmites australis         
Taraxacum officinale        
Scirpus sylvaticus   
Senecio jacobaea        
Viola epipsila         
Crepis paludosa       
Symphytum officinale      
Urtica dioica   
Valeriana officinalis       
Festuca pratensis      
Poa pratensis         
Agrostis stolonifera         
Holcus lanatus         
Poa palustris         
Poa trivialis         
Deschampsia cespitosa         
Alopecurus pratensis         
Anthoxanthum odoratum         
Festuca gigantea         
Glyceria fluitans         
Phleum pratense         
Lolium perenne         
Agrostis tenuis         
Dactylis glomerata         
Phalaris arundinacea         
Molinia caerulea        
Poaceae (total of seeds)      *
Carex flava         
Carex nigra         
Carex acutiformis       
Carex cespitosa       
Carex appropinquata       
Carex sp. (total of seeds)      *
Forest herbs
Scutellaria galericulata   
Solanum dulcamara       
Angelica sylvestris        
Peucedanum palustre     
Lycopus europaeus     
Anemone nemorosa      
Aegopodium podagraria       
Geum urbanum         
Geranium robertianum         
Cardamine amara         
Impatiens noli-tangere         
Woody species
Salix cinerea    *
Salix pentandra        
Salix aurita        
Betula pendula      
Frangula alnus         
Populus tremula         
Alnus glutinosa       
Figure 1.

Vegetation dynamics in an abandoned meadow (0.75 ha) in the period 1976–96. The first map was drawn in 1981, i.e. 5 years after the cessation of mowing.

The initial stage (0–5 years) shows considerable changes in the spatial relations between the meadow species populations and could be divided into: phase 0 – onset of succession, consisting of the year (1976) of abandonment of meadow; and phase 1 (1–5 years) – the proportion of the tall herbs Filipendula ulmaria, Lysimachia vulgaris and Lythrum salicaria in the vegetation increases.

The temporary stage consists of: phase 2 (5–10 years) – dominance of sedges and tall herbs increases, clumps of willows develop, meadow species occur sporadically in gaps between sedges, a macroforb community develops (Lysimachio vulgaris–Filipenduletum); and phase 3 (10–15 years) – limited aggregations of willows grow up and create willow scrub with a high proportion of sedges.

The terminal (early forest) stage is represented by phase 4 (15–20 years) – willows dominate, with birches and alders among them, tall herbs and sedges grow along the edges of willow clumps; and a Salicetum pentandra cinereae community forms.

Floristic composition of vegetation and seed banks

A total of 81 species occurred in the vegetation of the abandoned meadow over the 20 years of study. The number of species at any one time ranged between 31 and 40, while the number identified in the seed bank ranged from 38 to 25. Species appearing in the vegetation were generally also present in the seed bank. Twenty species occurred at all successional stages in both the vegetation and the seed bank, although not always at similar frequencies. The frequency of occurrence of some species (e.g. Lychnis flos-cuculi, Galium palustre, Lythrum salicaria, Plantago lanceolata and Urtica dioica) was higher in the seed bank than in the vegetation in later phases of succession, while the reverse relationship was true of sedges.

Comparisons between the seed bank and vegetation also revealed species such as Lathyrus pratensis, Geum rivale and Polygonum bistorta, which were present in the vegetation in all successional stages but in the seed bank at some stages only. The reverse situation (continued presence in the seed bank but absence from the vegetation during some stages) characterized Epilobium palustre, Urtica dioica and Potentilla anserina.

Correlations between the species composition of the seed bank and of the vegetation decreased as succession proceeded, although they were high in absolute terms throughout the early phases (r = 0.95, r = 0.79 and r = 0.65 for phases 0, 1 and 2, respectively). Only in the third and fourth phases did significant differences between the species compositions of the seed bank and the vegetation appear (r = 0.35, r = 0.28, respectively).

The size of the seed bank in the course of succession

The reserve of seeds in the soil

The density (abundance) of seeds in the meadow soil varied significantly over the 20 years (Fig. 2). Low densities were noted both under closed-canopy scrub (2468 m−2) and in the initial stage of succession under the meadow (2975 m−2), while the highest density was 9170 m−2, 15 years after the cessation of mowing on the meadow, i.e. in the transitional stage (phase 3), when the vegetation consisted of meadow, scrub and forest species (Table 1). Analysis of variance showed that the different successional stages differed significantly in the size of the seed bank (F = 9.6 > FM = 2.42). Significant differences were observed between phase 0 (meadow) and all intermediate stages, but not the terminal willow scrub phase (phase 4), and not between phases 1 and 2 (F = 1.18 < FM = 1.60) and phases 2 and 3 (Fig. 2).

Figure 2.

Comparisons of seed bank size in consecutive phases of succession; data from (a) the separation method; (b) seedling emergence in a greenhouse; (c) seedling emergence following clearing in the field.

Seedling emergence in the greenhouse

The lowest density of seedlings was noted for monoliths from under willow scrub (1250 m−2) and meadow (1890 m−2). The highest densities were noted 5 and 10 years into succession (4052 m−2 and 4182 m−2, respectively).

The density of seeds estimated by this method was significantly lower than the total seed bank recorded directly from the soil. However, the trend was the same in both cases, with the lowest densities characterizing the early and late phases and the highest the temporary stage (after 10–15 years) (Fig. 2).

The estimates of seed bank size based on greenhouse emergence of seedlings differed significantly between the successional phases (F = 32.76 > FM = 2.42). The size of the seed bank from the willow scrub phase (phase 4) differed significantly from those in phases 1, 2 and 3, but not from that in the meadow (phase 0).

Seedling emergence on plots in the field

Estimates of seed bank size based on emergence following the removal of vegetation were significantly lower (up to 10-fold) than those using other methods. Analysis of variance again revealed significant differences in seedling abundance between consecutive successional phases (F = 4.34 > FM = 3.48), with similar trends to those inferred from the other methods (Fig. 2). Thus seedling densities were lowest at phases 0 and 4 (132 m−2 and 156 m−2) and highest (432 m−2) when the vegetation was removed 5 years after the last mowing.

Proportions of species

The pattern shown in the relative proportions of species in vegetation throughout succession allowed identification of groups of plants that showed significant differentiation from one another. Meadow and forest species could be distinguished and meadow species further divided into herbs, grasses, sedges and so-called ‘other’ species (which included Juncus effusus, Scirpus sylvaticus and Urtica dioica, i.e. species whose presence is often associated with disturbance). Tall herbs (Filipendula ulmaria, Lysimachia vulgaris and Lythrum salicaria) were also regarded as a separate group, as they form a distinct macroforb community which increases markedly after mowing ceases.

The proportions of the seed bank represented by these different groups were seen to change significantly over the 20 years (Fig. 3). When total seeds were analysed in the first year following the abandonment of the meadow, some 64% of the seed bank came from meadow herbs. This figure declined throughout the study to 17% after 20 years, while the share of seeds of forest species increased significantly (from 0.5% to 23%). Significant changes in the floristic composition of the seed bank occurred during the temporary phase of succession. After 10 years had elapsed no less than 40% of the pool of seeds were of the ‘other’ species, although they had increased only slightly in the vegetation: the remaining fraction included seeds of all other categories. By the time willow scrub had developed, a quarter of the seed bank consisted of seeds of forest species, a quarter of seeds of tall herbs, and the remaining half of seeds from meadow, grass and sedge species that had been present in the earlier successional stages.

Figure 3.

Changes in the contribution (%) of particular species groups in the seed bank during succession, as assessed by three methods: (a) the total number of seeds from all samples of soil; (b) the total number of seedling emergence in a greenhouse; (c) the total number of seedlings emerging on plots, 0–20 years after abandonment of the meadow.

When seedlings emerging on cleared plots were analysed meadow species remained dominant for 10 years but decreased from 92% to 47%. Only after 15 or 20 years did their share become less notable, at between 24% and 16%. The proportion of seedlings of tall herb species had already increased by 5 years after abandonment, and was still greater (51%) after 15 or 20 years (Fig. 3c). The so-called ‘other species’ (Urtica, Juncus and Scirpus) were represented to only a limited extent throughout the study period, accounting for between 2% and 10% of seedlings, when this method was analysed. Similar patterns were seen when total seed emergence in the greenhouse was analysed (Fig. 3b).

Comparison of the composition of the vegetation and seed bank over the course of succession showed a number of discrepancies. First, while meadow species in the vegetation declined steadily, this was not fully reflected in the seed bank, with the seeds of many such species being retained into the beginning of the forest phase, by which time they were only sporadically found in the vegetation (cover <10%). The densities in the seed bank of some species varied significantly between the phases (e.g. Cirsium from 320 to 50 seeds m−2, Lychnis 200–63, Ranunculus 255–37; Table 2). Forest species became significant at about the same time in both the vegetation and the seed bank (i.e. after 15 or 20 years), although the seeds of some forest species were present sporadically even in the initial stage of succession (e.g. Betula, Frangula and Salix). The cover of the sedge group increased, especially in the temporary stage (to 40%), although this was not reflected in the seed bank, where proportions remained low throughout the succession (at 5–13%). However, the densities of Carex in the seed bank varied significantly between the phases, e.g. from 1172 to 36 seeds m−2. Finally, the proportions of the seed bank represented by Juncus, Scirpus and Urtica seeds rose markedly, especially in the temporary phase (to 40%), but this was not reflected in the vegetation, where these species remained infrequent (at below 10% cover).

Table 2.  Comparison of the seed bank size between different succession phases. Data are from all soil samples (seed separation method)
 Density of seeds per m2
After years05101520
Phases 01234
Meadow species
Cirsium palustre32039321714450
Ranunculus acris25571599626537
Lythrum salicaria22092755021678
Lychnis flos-cuculi20083861663
Caltha palustris140.815.
Carex sp.140117236987936
Viola sp.130281390.20
Plantago lanceolata112..5.
Juncus effusus100858388392.
Trifolium repens99....
Myosotis scorpioides801325.82
Lotus corniculatus65....
Lysimachia vulgaris6253651722
Galium palustre601043947234
Potentilla anserina508177530.
Rumex acetosa45458..
Epilobium palustre44.13.35
Mentha arvensis4013523451.
Cardamine pratensis36108..
Campanula patula33....
Lathyrus pratensis32..3.
Polemonium caeruleum31....
Potentilla palustris30367...
Geranium palustre30645..
Filipendula ulmaria298619015981
Urtica dioica272311475225
Scirpus sylvaticus25411843.
Cerastium holosteoides21108..
Geum rivale15..30.
Crepis paludosa10....
Vicia cracca5.5..
Symphytum officinale5..10.
Valeriana officinalis5....
Taraxacum officinale5....
Senecio jacobaea5....
Galeopsis tetrahit5....
Polygonum bistorta.68106.30
Poaceae (remaining species)26325941940137
Forest species
Peucedanum palustre10..7415
Aegopodium podagraria.13.8.
Scutellaria galericulata.10237345
Lycopus europaeus..23922550
Anemone nemorosa..93.
Impatiens noli-tangere....90
Cardamine amara....60
Geum urbanum....40
Geranium robertianum....15
Solanum dulcamara.....
Angelica sylvestris.....
Betula pendula.6.47570
Salix sp...9335088
Alnus glutinosa..8.90
Rubus sp.....40
Frangula alnus.....
Populus tremula.....


The results indicate that, in the course of meadow to forest succession, seed bank size fluctuates: a low density was found initially but rose over threefold after 15 years before falling back to a level similar to that recorded at the initial stage. However, according to earlier studies (Falińska 1981, 1996) the density of seeds in a stable (climax) floodplain forest community adjacent to the abandoned meadow was twice as high as in a willow scrub similar to the end-point community here.

Other studies conducted in semi-natural communities and on arable land (Graham & Hutchings 1988; Cavers & Benoit 1989; Rice 1989; Levassor et al. 1990; Hester et al. 1991; Milberg 1992; Bakker et al. 1996; Mitchell et al. 1997, 1998) show different patterns from those observed here. For instance, although the highest density of seeds was found at the temporary (macroforb) stage (Oosting & Humphreys 1940) it came as late as in the forest community in another (Patridge 1989). Directional changes in the density and species richness of the seed bank during succession have very rarely been described (Pickett & McDonnell 1989; Roberts & Vankat 1991), perhaps because the numerous vegetation changes that occur during succession may not be linear in nature. The floristic composition of the seed bank is partly determined by the current species composition of communities but depends also upon the vegetation history (Grandin & Rydin 1998) and on the biological properties of plants (Harper 1977; Silvertown 1980, 1981). Factors such as fecundity, seed germination ability (Cook 1980; Fenner 1985; Baskin & Baskin 1989), longevity (Baker 1989; Rice 1989; Milberg 1990; McDonald 1993; Thompson et al. 1997), migration, mode of seed dispersal (Rabinowitz 1981), life history (Grime et al. 1988) and size and shape of seed (Thompson et al. 1993; Hutchings & Booth 1996a, 1996b), may be important and can sometimes be regarded as random events. The low number of seeds at the beginning of this succession, in spite of the richness of species, probably results from the fact that the meadows had been mowed during the flowering period (Falińska 1991) so that few plants would have set fruit. Five years after the last mowing of the meadow the number of seeds had doubled. The cessation of mowing would allow the majority of plants to set seed, thus increasing the supply to the soil, and clonal plants to multiply the number of generative ramets, as documented in studies on the demography of species during succession carried out in the same area (Falińska 1986, 1991). The highest seed density in the soil 15 years after the cessation of mowing may have resulted from seed accumulation in the soil because of abundant production combined with limited germination due to the increasing density of clonal plants (Falińska 1995, 1997). Soil from under the canopy of willows contained seeds of meadow species, e.g. Cirsium palustre, Ranunculus acris and Lychnis flos-cuculi (Table 3), that were no longer present in the vegetation, which confirms earlier reports that seeds of certain species are long-lived (Cook 1980; Roberts 1981; Fenner 1985; Milberg 1990; McDonald 1993; Thompson et al. 1997). It is not clear why the seeds of forest species did not begin to appear in the soil until several years after meadow abandonment, as the meadow was adjacent to a forest. The rich meadow vegetation may have limited the inflow of migrating seeds because penetration into the soil may occur only when gaps appear in the vegetation as a result of the death of clones of the dominant species (see data for Filipendula ulmaria;Falińska 1995). Vegetative reproduction may significantly affect the role of the seed bank in succession as it leads to formation of a compact vegetation cover. The appearance of gaps may be needed to facilitate the emergence of seedlings and vegetation regeneration (Pakeman & Hay 1996). Such limitation may also be responsible for the decreasing density of seeds as willow scrub formed. Similar results have been obtained by other authors, for example Davies & Waite (1998), who found a negative correlation between the age of the scrub and the size of the seed bank.

Table 3.  Presence of species in the seed bank at the initial (meadow) and early terminal (willow brushwood) succession stages. The assessment was carried out for the total number of seeds in the soil (SB), seedling emergence in the in greenhouse (GSB) and seedling emergence under field conditions (F)
 Meadow (0–5 years)Willow brushwood (15–20 years)
Cirsium palustre * * * * * *
Cirsium rivulare * * * * * *
Lychnis flos-cuculi * * * * * *
Ranunculus sp. * * * * * *
Myosotis scorpioides * * * * * *
Galium palustre * * * * * *
Viola epipsila * * * * * *
Viola palustris * * * * * *
Filipendula ulmaria * * * * * *
Lythrum salicaria * * * * * *
Lysimachia vulgaris * * * * * *
Urtica dioica * * * * * *
Scutellaria galericulata * * * * * *
Peucedanum palustre * * * * * *
Caltha palustris * * *   * *
Crepis paludosa * * *   * *
Juncus effusus * * *   * *
Rumex acetosa * * *   * *
Lotus corniculatus * * *    *
Cerastium holosteoides * * *   *  
Epilobium palustre * * * *   
Potentilla anserina * * *   *  
Polygonum bistorta * * * *   *
Geum rivale * * *    *
Aegopodium podagraria * *     *
Plantago lanceolata * *     *
Betula pendula *    *   *
Scirpus sylvaticus *      *
Lathyrus pratensis * * *    
Vicia cracca * * *    
Trifolium repens * * *    
Cardamine pratensis * * *    
Campanula patula * * *    
Potentilla palustris * * *    
Geranium palustre * * *    
Mentha arvensis * * *    
Polemonium caeruleum * *     
Galeopsis tetrahit * *     
Symphytum officinale *      
Taraxacum officinale *      
Senecio jacobaea *      
Anemone nemorosa      *
Populus tremula      *
Solanum dulcamara    * * *
Lycopus europaeus    * * *
Rubus sp.    * *  
Geum urbanum    * *  
Geranium robertianum    * *  
Cardamine amara    * *  
Impatiens noli-tangere    * *  
Salix sp.    *   *
Alnus glutinosa      *
Frangula alnus      *

Data from all three methods indicate that the density of forest plant seeds in the soil increased as succession progressed, as observed for seed banks in other forest communities (Falińska 1981; Pirożnikow 1983; Patridge 1989; Pickett & McDonnell 1989; Nakagoshi 1996). However, complete return of the abandoned meadows to forest is expected to take about 150–200 years (Faliński 1986) and the total seed bank size will probably increase and decline several times during this period.

The number of species in the seed bank decreased over the course of succession from 38 to 25. While there might be increasing difficulties in identifying seeds as groups such as grasses, sedges and willows become dominant, this is unlikely to explain the reduction in species as only two species of sedge (Carex acutiformis and Carex cespitosa) and one of willow (Salix cinerea) were common in the vegetation.

It is also worth emphasizing that as well as supplying the recruits needed for the development of new species compositions during succession, the seed bank will, if seeds of meadow species persist, allow regeneration of the meadow if succession is arrested. Emergence in the field after mowing shows that although their proportion decreased gradually in successive years in favour of the forest species, meadow species constituted a considerable part of the pool of seedlings for at least 20 years (Fig. 3c). Juncus effusus, Urtica dioica and Scirpus sylvaticus seedlings were very rare in these plots, and thus it can be assumed that moderate disturbance during succession (by removing vegetation) preserves the floristic richness, although environmental screening leads to the continued absence of some species. The seed reservoir in the soil is therefore of great significance for the regeneration of the vegetation following disturbance, as has often been proposed (Grubb 1988; Hutchings & Russell 1989; Silvertown & Tremlett 1989; Thompson 1992; McDonald 1993; McDonald et al. 1996; Pakeman & Hay 1996; Mitchell et al. 1997, 1998).

Estimates of seed bank size made according to three different methods gave different results (Fig. 2). The total number of seeds picked out of the soil was 2–3 times higher than the size of the seed bank assessed by seedling emergence under laboratory conditions, and 10–20 times higher than suggested by seedling emergence in experimental plots in the meadow. These differences were to some extent expected, if only due to the varying types of seed dormancy and different temperature and humidity requirements (Thompson & Grime 1979; Cook 1980; Fenner 1987; Thompson 1987). No single method can determine the full species composition of a seed bank, because a researcher picking out the seeds of the soil may overlook very small seeds, while counting deeply dormant and non-viable seeds, and the period of observation of seedling emergence in the laboratory may be too short to allow full germination of many species. Our results are consistent with those of Jensen (1969), who showed that seedling emergence distinguished only 40% of the pool of seeds while for the pick-out method the proportion ranged from 70% to 80%.

The role of the seed bank depends on the period over which seeds remain viable in the soil, although this is very difficult to determine. Observation of long-term changes in species composition of seed bank and vegetation in permanent plots (Table 1) allowed the longevity of the seeds of some species to be assessed. Comparison of the species in the seed bank and vegetation of meadows unmown for 20 years showed that seeds of some meadow species may persist in the soil until early phase forest has developed. However, some of these species become infrequent as succession proceeds (Falińska 1991) and continued production cannot explain the presence of their seeds, often in large numbers, after 20 years, suggesting that seeds must remain viable for at least 5 or 10 years. All three methods for determining the seed bank suggest that Ranunculus acris, Lychnis flos-cuculi and Myosotis scorpioides show such behaviour, as seeds were present in the soil after 15 or 20 years, in spite of an almost complete disappearance from the vegetation 10 years after mowing ceased. The longevity of seeds of these species has previously been noted (Thompson et al. 1997). The viability of Juncus effusus seeds has been estimated at several hundred years (Milberg 1990; Milberg & Hansson 1993), and many studies draw attention to the frequent appearance of this species after disturbance, even in natural communities where its presence or proximity had not been noted previously (Pirożnikow 1983). Several other species, e.g. Taraxacum officinale, Trifolium repens and Campanula patula, which have been reported to have long-term seed bank (tens of years) (Thompson et al. 1997), did not persist in the seed bank in this study (Table 3).   The hypothesis that durable seed banks may be a survival strategy typical of light-demanding species assumes that the germination of seeds in the soil is obstructed by a lack of incident light (Pickett & McDonnell 1989), and that it is only when plant cover is disturbed and microhabitat conditions change that exposure and germination take place (Thompson & Grime 1979; Thompson 1992). The ecological consequence of this hypothesis is disturbance-dependent emergence of seedlings from the seed bank. While virtually all the species of genera such as Epilobium and Trifolium are capable of forming durable seed banks in some circumstances (Thompson et al. 1997), their seeds were not present here 20 years after the cessation of mowing, and as the species had disappeared from the vegetation within the first 10 years the viability of their seeds would seem not to have exceeded 10 years in this case.

The seed bank may be determined by the species composition of the vegetation, but may also depend to some extent on other, possibly random, events (Harper 1977). Significant similarities between the floristic composition of the vegetation and the seed bank in relatively stable forest communities suggest the former pattern (Johnson 1975; Falińska 1981, 1991; Pirożnikow 1983; Nakagoshi 1996; Jankowska-B?aszczuk et al. 1998), but low correlations (Vankat 1991; Hakan et al. 1987; Roberts & Vankat 1991; Milberg 1992, 1995; Ungar & Woodell 1993; Looney & Gibson 1995) are frequent and may be related to disturbance (Williams 1984; Patridge 1989). Our results suggest that factors other than current vegetation are important: the seed bank contains not only species from the existing vegetation but also species that were present in the vegetation of former stages and other species. The initial seed bank does not appear to influence the course of succession observed, contrary to the suggestion of Vankat & Carson (1991). The persistence of meadow species in the seed bank up to the formation of the forest community (Table 3), although many of them had retreated from the vegetation much earlier, confirms other reports of the formation of persistent seed banks by meadow species (Cook 1980; Fenner 1985; Thompson 1992; McDonald 1993).

Thus, the floristic composition of the seed bank at a particular time is not a mirror reflection of the vegetation at that stage of succession, but rather a record of the history of changes in the vegetation. Succession, however, depends on the varying proportions of particular groups of species in the seed bank, which play different roles at different stages in the sequence, leading to the formation of necromass, herb canopy or tree canopy.

Descriptions of seed banks in various studies show that there are at least three patterns during succession (Fig. 4). Unidirectional changes are observed in both density and species richness, corresponding with Clement’s succession model of unidirectional changes in vegetation. This pattern was described for seed banks where succession started in a weed community (Roberts & Vankat 1991). On the other hand, data from semi-natural meadows indicate that during succession the seed bank fluctuates either regularly or with variable peaks (a fluctuation–lottery pattern) as a result of both the long-term turnover of species and of various other processes that significantly affect the seed bank size as well as random events (Lavorel & Lebreton 1992). Just as there is more than one succession model (Glenn-Lewin et al. 1992) various seed bank models may be possible (Leck et al. 1989).

Figure 4.

Seed bank models for succession under various conditions.

Many studies have indicated that a given succession may follow several different paths (Prentice 1986; Van der Maarel 1988). For instance, in various parts of a 15-ha abandoned meadow (Falińska 1991) (i) trees appeared directly without a long-term species turnover; (ii) patches of two to three herb dominants developed and after 15 years were colonized by willow and alder trees; and (iii) the floristic compositions changed according to the classical Clement’s model. The data on the seed bank described in the present paper are representative of the second path of forest regeneration in meadows. Both the course of succession and the seed bank pattern often depend on the initial situation, i.e. on the type of vegetation in which succession begins. Directional changes in species richness and density of seeds in the soil are most likely to occur in disturbed environments. On the other hand, if, as here, succession starts in semi-natural vegetation, the seed bank will probably follow a fluctuation–lottery pattern.


I would like to thank K. Thompson and the anonymous referees for constructive comments on the manuscript. My special thanks go to the former and present technicians of łowieża Geobotanical Station of Warsaw University: Walentyna Maciejewska, Alicja Wiktoruk and Roman Wo?kowycki, who helped me to collect the material over many years. I am grateful to Halina Koícielecka for typing and preparing the list of references.

This work was partly supported by grant KBN.

revision accepted 28 November 1998

Received 3 November 1997