Importance of regional species pools and functional traits in colonization processes: predicting re-colonization after large-scale destruction of ecosystems

Authors


*Correspondence author. E-mail: kirmer@loel.hs-anhalt.de

Summary

  • 1Large-scale destruction of ecosystems caused by surface mining provides an opportunity for the study of colonization processes starting with primary succession. Surprisingly, over several decades and without any restoration measures, most of these sites spontaneously developed into valuable biotope mosaics with many endangered plant species.
  • 2Investigations were carried out in 10 open-cast mined sites. Data from extensive floristic mapping of mined sites and their surroundings were combined with functional traits.
  • 3Using binary logistic regression, we showed that eight variables have a significant influence on the probability of the occurrence of particular plant species in mined sites. The regional species pool explained the highest proportion of variance. Further significant variables were distance of source populations, abundance in the habitat species pool, capacity for long-distance dispersal by wind and birds, terminal velocity of seeds and requirements for light and nitrogen.
  • 4In the 10 study mine sites, 143 establishing events of Red List species have been recorded. For 40% of these establishing events, the nearest recorded seed source is 3–10 km away, while for 19%, the distance to the nearest seed source exceeds 10 km.
  • 5This study showed for the first time that the abundance of species up to a distance of at least 17 km plays an important role in colonization processes following large-scale destruction of ecosystems. In large-scale, nutrient-deficient, open sites, an accumulation of plant species, including rare species, can be expected in time frames amenable to planning (several decades), because the sites acted as huge seed traps in the landscape.
  • 6Synthesis and applications. The floristic colonization probability of restoration sites is higher if large-scale, open and nutrient-poor habitats are available. In regions where such habitats have become highly fragmented, restoration planning of derelict land (e.g. surface mines, quarries, landfills) that supports the creation of such conditions can contribute to the regional survival of rare pioneer species. In future restoration planning, a new protocol must be established that combines the utilization of site potential with spontaneous colonization processes.

Introduction

Population dynamics of plant species are not only determined by local niche-based processes, but also by regional dispersal processes. It is increasingly acknowledged that the availability of seeds can also be a major limiting factor (‘seed limitation’) in ecological restoration projects (Zobel 1997; Bullock & Clarke 2000; Ozinga et al. 2004). The degree of seed limitation is likely to depend on the abundance of species in the regional species pool and specific dispersal traits of the plant species (Zobel 1997; Ozinga et al. 2005). The importance of both factors is inherently difficult to study experimentally, because spatial and temporal scales are often too small for tracing rare long-distance dispersal (LDD) events (Nathan et al. 2002).

In our study, we used a complementary approach in which we studied long-distance dispersal indirectly by regarding large-scale surface-mining areas as huge seed traps in the landscape. In the last century, surface mining of lignite exceeded the mass turnover of the last ice age in the German federal states of Brandenburg, Saxony and Saxony-Anhalt (Müller & Eissmann 1991). This resulted in the destruction of many valuable ecosystems such as flood plains, woodlands and grasslands. After German reunification in 1990, 32 active surface mines were shut down. Until 1990, only 55% of the mined sites had been restored due to economic reasons. Surprisingly, without restoration measures, the sites developed spontaneously over decades from ‘lunar landscapes’ to valuable biotope mosaics with different pioneer and grassland communities, pioneer scrubs and woodlands, reeds and swamps (e.g. Schulze & Wiegleb 2000; Tischew & Kirmer 2007). Beginning in 1993, several research projects in surface-mined land have collected comprehensive floristic data in mined areas of Saxony-Anhalt, Germany (see Acknowledgements). The main objective was to gain knowledge about processes that are important for trajectories in colonization and to derive guidelines for further restoration planning. Simultaneously, large-scale surface mining offered a unique chance to observe primary succession in Central Europe over a longer period of time (e.g. Prach & Pyšek 1999) because, otherwise, large-scale disturbances have become very rare.

In preliminary examinations, it became evident that plant species could bridge distances up to several kilometres (Brändle et al. 2003; Tischew & Kirmer 2003; Tischew & Lorenz 2005). In their study of non-standard long-distance dispersal, Higgins, Nathan & Cain (2003) emphasized the fact that many seeds utilize both standard and non-standard dispersal vectors, with the latter being responsible for rare long-distance dispersal events. In effect, these rare, but ecologically important events may decrease the influence of trait-based processes relative to the importance of trait-neutral stochastic processes. In his provocative ‘neutral theory’, Hubbell (2001) even claimed that traits do not matter at all. The probability of colonizing a site would then merely become a lottery in which chances are weighted by the availability of seed sources in the surroundings (Ozinga et al. 2005). In Hypothesis 1, we tested if the probability of colonization is only determined by propagule availability by analysing how much of the variance in our data set can be explained by distance and the abundance of species in the surrounding area.

An alternative possibility is that species-specific dispersal traits are decisive for the re-colonization of large disturbed areas. Stöcklin & Bäumler (1996), Fort & Richards (1998) as well as Wood & del Moral (2000) emphasized that wind dispersal is the most important LDD vector in primary succession. In particular, species with morphological traits that support low falling velocity (plumed, winged, balloon-like or very small and light seeds) are prone to LDD (Nathan et al. 2002; Tackenberg, Poschlod & Bonn 2003; Soons et al. 2004). Other studies have stressed the importance of dispersal by birds (e.g. Fridriksson 1975; Wilkinson 1997) or focused on seed related traits such as seed weight or seed longevity. Therefore, we tested if inter-specific differences in dispersal traits influenced immigration probability (Hypothesis 2). Based on existing studies (e.g. Brändle et al. 2003; Ozinga et al. 2004, 2005), we selected species traits that already had been proven to be important in colonization processes or in the prediction of species assemblages.

Small-scale restoration projects have shown that immigration via LDD is of limited success (e.g. Verhagen et al. 2001; Bischoff 2002); hence, the special features of large-scale mined areas seem to influence colonization probability. After mining stops, the areas consist predominantly of vegetation-free raw soil which favours the establishment of pioneer species or competition-poor species in general (e.g. Rehounková & Prach 2006). Another special feature of non-reclaimed sites in mined areas is an open vegetation structure in all successional stages caused by nutrient-deficient site conditions. Since several studies have shown that environmental conditions can be important determinants of species composition (e.g. Ozinga et al. 2004; Rehounková & Prach 2006), we tested if adaptations to environmental conditions such as light and nitrogen availability or moisture also can facilitate colonization (Hypothesis 3).

Overall, we sought to establish how much of the variance concerning species composition in mined sites can be explained by mass effects (Hypothesis 1) and how much by trait-based processes (Hypotheses 2 and 3).

Methods

Investigations were carried out in 10 mined sites in Saxony-Anhalt, Germany (see Supplementary Material Fig. S1). The data were sampled in areas that developed spontaneously. Most of the sites are characterized by nutrient-poor but generally hospitable substrates with a wide amplitude of site parameters (e.g. pH value: 3–7, coal content: 0·25–65%; moisture: dry to wet). Characterization of predominant site conditions and vegetation types of the selected study sites is shown in Table 1.

Table 1.  Characteristics of the 10 study sites with affiliation to the northern and southern mining regions
Mined siteAffiliation to mining regionApproximate age (years)Approximate size (km2)Predominant site conditionsPredominant vegetation types
Golpa-Nord BachaueNorth2–141Plain with ridges and hollows, artificial runnel, sand-dominated, low pH values, wet and dry conditionsDifferent pioneer communities, shrubs, reed
Goitzsche Holzweißig-WestNorth4–301·8Plain with gravel- and lignite-rich sand with low to very low pH values, small parts characterized by loamy sand with low lignite contentPsammophytic grassland with patches of Calamagrostis epigejos (L.) Roth dominance stands, shrubs
Goitzsche Halde 1035North3–400·6Tip with predominately clayey and loamy substrate, small patches with gravel-rich and lignite-rich sandBirch pioneer forest, small patches of reed and psammophytic grassland
Muldenstein nature reserve BurgkemnitzNorth380·8Wetland area with open water, sandy -clayey loam alternating with loamy sand and sandy loam, moderate to low pH valuesAquatic vegetation, swamps, reed, pioneer forests, Calamagrostis epigejos dominance stands, small patches of psammophytic grassland
Golpa IIINorth6–502·5Final void filled with water, surrounded by slopes with ridges and hollows, loamy sand with low to moderate pH valuesPioneer forests, different grassland types, swamps, reed, aquatic vegetation
Kayna-Süd southern partSouth3–341·8Final void filled with water, surrounded by slopes with ridges and hollows, carbonate-free gravel-rich sand, clay or silt and carbonate-rich loam, in small parts high lignite content and low pHPioneer forests, different grassland types, Calamagrostis epigejos dominance stands, reed
Roßbach southern partSouth1–232·3Final void, outward slopes with loess, inner parts with lignite- and gravel-rich tertiary sand and siltPioneer forests, different grassland types, Calamagrostis epigejos dominance stands, reed
Mücheln InnenkippeSouth1–412·6Large slope system, lignite-rich, carbonate-free loam and silt (eastern part), carbonate-, gravel- and lignite rich loam alternating with carbonate-free, gravel- and lignite-rich sand (western part)Pioneer vegetation (eastern part), shrubs, different grassland types, reed (western part)
Nordfeld Jaucha nature reserveSouth26–551Plain with heterogeneous substrate consisting mostly of boulder clay, loess, quaternary sand and gravel, in parts mixed with small amounts of tertiary sand and ligniteOld pioneer forests developing to deciduous woodland, small lakes, reed, swamps, different grassland types, large orchid populations [e.g. Epipactis palustris (L.) Crantz, Dactylorhiza incarnata (L.) Soó]
Domsen north-eastern partSouth27–411Quite similar to Jaucha with a higher amount of loess and a higher share of tertiary substrate and lignite in some partsPioneer forests, different grassland types, swamps, reed

We follow Eriksson (1993) and Zobel (1997) in defining the species pool as a set of species which are potentially capable of co-existing in a certain community. This concept, thus, implies that, for an ecologically relevant estimation of species pools, both the spatial distribution (geographical filter) and habitat tolerances (environmental filter) of species must be known (Zobel, van der Maarel & Dupré 1998; Ozinga et al. 2005).

All species within 17 km of the study areas (see Supplementary Material Fig. S2) were integrated into a geographical species pool using the data bases of the floristic mapping of the states of Saxony-Anhalt and Saxony (data base Higher Plants of Saxony-Anhalt, Landesamt für Umweltschutz Saxony-Anhalt Halle, working status 1998; data base Flora of Saxony, Sächsisches Landesamt für Umwelt und Geologie Dresden, working status 1999). The total area (mined site and surroundings) was 3609 km2. Floristic mapping in both federal states began in 1949 and provided an inventory of all higher plant species based on grid cells with a mesh size of 5·5 km. Therefore, we know which species occurred in which grid cells during the last five decades. Mapping of the grid cells that contain the mined sites was carried out exclusively outside the mined sites because mining sites were forbidden zones with limited access before the political change in Eastern Germany in 1989. Each mined site is located in a different grid cell. The ongoing floristic mapping in the surrounding area ensured that species which are not present in the close vicinity most probably reached the mined sites by LDD. Starting from the geographical species pool, we created a subset using data compiled from extensive floristic mapping of approximately 80 former and active mining sites in Saxony-Anhalt (1994–2002: own data base). We used these data to create a habitat species pool containing only species which proved to be capable of growing under mining site conditions in general. Additionally, for each study site, environmental filters were installed to ensure the exclusion of species unable to grow under the current site conditions of the respective mined site (e.g. wetland species on dry, sandy sites, ancient woodland species on raw soil). Consequently, as a subset of the habitat species pool for each study site, a regional species pool was compiled containing information about the abundance of each species in the 5·5 × 5·5 km grid cells of the 0–17 km surroundings. Finally, we built an actual species pool containing all species already present in each of the 10 mined sites. There were 675 species in the analysis including 74 Red List species (Frank et al. 1992) with 6750 species × site records. Red List categories followed the IUCN (1994). The share of neophytic species was about 8% in the total species number.

Functional traits (see Table 2) and Ellenberg indicator values (Ellenberg et al. 1992) were derived from different data bases, such as LEDA (Knevel et al. 2005; http://www.leda-traitbase.org), Biolflor (Frank, Klotz & Westhus 1990) and M. von Lampe (unpublished data). Each species from the habitat species pool was assigned their available Ellenberg indicator values for light, nitrogen availability and moisture. Therefore, the species from the actual species pool can be used to characterize the special site conditions in the study sites. Evidence for the accuracy of these indicator values has been provided by several studies reporting close correlation between indicator values and corresponding measurements of environmental variables (e.g. Thompson et al. 1993).

Table 2.  Overview of functional traits, compiled species pools and other parameters used
ParametersDescription
GSPGeographical species pool, species that are present in the whole region. The abundance of these species is expressed by the number of occupied 30·25 km2 grid cells.
HSPHabitat species pool, species that are present in the whole region and have proved to be able to grow under mining site conditions. The abundance of these species is expressed by the number of occupied 30·25 km2 grid cells (subset of GSP).
RSP 0_17Regional species pool, species from the habitat species pool that occurred in 0–17 km distance of each mined site. The abundance of these species is expressed by the number of occupied 30·25 km2 grid cells (subset of HSP).
ASPActual species pool, species that are already present in the examined mined sites
DistanceNearest distance to the respective mined site, ranging from class 1 (c. 0–3 km) to class 10 (c. 59–66 km); see Supplementary Material Fig. S1
North_southAffiliation to the northern or southern mining region
Light availabilityEllenberg indicator for light availability
MoistureEllenberg indicator for moisture
Nitrogen availabilityEllenberg indicator for nitrogen availability
Seed weight classRanging from 1 (light) to 8 (heavy) depending on the seed weight
Seed longevityPersistence in the soil seed bank (classified with the seed longevity index ranging from 0 = low to 1 = high)
Terminal velocityTerminal velocity of seeds (m s−1)
Dispersal potential windCapacity for long-distance dispersal by wind (0 = low, 1 = high)
Dispersal potential waterCapacity for long-distance dispersal by water (0 = low, 1 = high)
Dispersal potential furCapacity for long-distance dispersal by fur of animals (0 = low, 1 = high)
Dispersal potential dungCapacity for long-distance dispersal by dung of mammals (0 = low, 1 = high)
Dispersal potential birdsCapacity for long-distance dispersal by bird droppings (0 = low, 1 = high)

According to their affiliation in the northern Bitterfelder/Gräfenhainicher coalfield or the southern Geiseltal/Profener coalfield, the study sites can be assigned to a northern or a southern mining region. First, we examined if special features of these two mining regions affected the abundance of species in the respective mined sites. Therefore, land cover based on a classification of satellite images (Landsat7 ETM+, 14·08·2000) was used to characterize both regions (U. Nocker and C. Gläßer, unpublished data). For all species in each of the 10 investigated mined sites (actual species pool), occurrences in the surrounding grid cells of the floristic mappings were located and all species originating from reclamation measures (planting or seeding) were omitted.

Secondly, an analysis of the species pools was carried out with binary logistic regression with forward selection using spss 12·0. The criterion for inclusion of variables in the forward selection was a probability of 0·05. In the full model, all parameters shown in Table 2 were included. The Wald statistic was used as a measure of the relative effect size of the variable in the full model. Nagelkerke's R2 gives the cumulative proportion of explained variance after entrance of the variable in the model. The regression coefficient indicates a positive or negative effect of the independent variable.

Differences in seed weight and terminal velocity between immigrated and non-immigrated species (without phanerophytes and nanophanerophytes) of the regional species pools in the surroundings of 0–3 km of the mined sites were tested using either T-test (Levene test > 0·05) or U-test (Levene test ≤ 0·05) depending on the normal distribution of the data (analyses done using spss 12·0). Significance levels were indicated in the following way: [*] 0·05 ≥P > 0·01; [**] 0·01 ≥ P > 0·001; [***]P ≤ 0·001. The number of cases in the analysis depended on the availability of data for seed weight and terminal velocity as well as the occurrence of species in the respective regional species pools.

Following Westoby, Leishman & Lord (1995), who stated that each species present in a vegetation type represents an independent item of evidence for colonization by possessing a particular set of attributes, we decided not to apply a phylogenetic correction to the data set. Additionally, we did not have an imbalance in our data, because most of the genera in the analysis contain between one and six species. Only Carex spp. (24 species) and Hieracium spp. (17 species) showed distinctively higher species numbers. But even for those genera, the trait values were different (e.g. terminal velocity 0·35–0·92 in Hieracium).

Results

comparison of mining regions and ecological potential of mined sites

The classification of satellite images of the area surrounding the mined sites showed that more than 70% of the area in the northern region consists of woodland and grassland compared to the southern region, which is dominated by arable land (Table 3). This is reflected by higher species numbers in the grid cells of the floristic mapping. In the northern region, 89% of the species already present at the mined sites occur within 3 km. In the southern regions, on the other hand, only 65% of the species at the mined sites occur within 3 km. A comparison of species numbers between mined sites and surrounding grid cells showed that, in both regions, 55–60% of the species from the regional species pool had already immigrated into the mined sites.

Table 3.  Potential seed sources in the surroundings of the mined sites in comparison to species already present
 Northern mining region (n = 5)Southern mining region (n = 5)
Percentage of woodland and grassland areas in 0–17 km distance to the mined sites 72·226·6
Percentage of arable land in 0–17 km distance to the mined sites23·567·9
Average number of all recorded species in 0–3 km surroundings of the mined sites (= grid cell from floristic mapping)642 (SD ± 70)264 (SD ± 35)
Average number of species from the regional species pool of the mined sites in the 0–3 km surroundings of the mined sites455 (SD ± 60)259 (SD ± 46)
No. of species from the regional species pool of the 0–3 km surroundings already present in the mined sites248 (SD ± 52)154 (SD ± 16)
Occurrences of species already present in the mined sites in the grid cells of the floristic mappings (%)
< 3 km88·9 (SD ± 2·7)64·7 (SD ± 5·4)
3–10 km8·8 (SD ± 2·3)29·8 (SD ± 3·5)
10–17 km1·0 (SD ± 0·6)3·9 (SD ± 2·1)
> 17 km1·3 (SD ± 0·7)1·6 (SD ± 0·6)

An analysis of all recorded species in the mined sites showed that, in the northern region, 11% of these species occurred farther than 3 km away, whereas the proportion increased to almost 36% in the southern region (Table 3).

analyses of species pools

In the full model, eight variables were significant in determining the occurrence of species in mined sites (Table 4). The variable with the greatest effect in the full model was abundance in the grid cells of the regional species pool (RSP 0_17). Significant dispersal traits were capacity for long-distance dispersal by wind and birds as well as terminal velocity. Significant Ellenberg indicator values were light and nitrogen availability. The analysis also showed that, despite the differences in species numbers and landscape pattern, affiliation to the northern or southern mining region was not significantly important for the occurrence of species in the actual species pools of the mined sites.

Table 4.  Results of multiple logistic regression (full model) with occurrence in the actual species pools of the mined sites as dependent variable (RSP, regional species pool; HSP, habitat species pool; P, significance level)
VariableRegression coefficientWaldd.f.PNagelkerke R2
RSP 0–17 km0·09062·4861≤ 0·00050·235
Terminal velocity−0·22429·2681≤ 0·00050·276
Light availability0·16628·1061≤ 0·00050·284
HSP0·01431·6601≤ 0·00050·289
Dispersal potential wind−0·61420·6671≤ 0·00050·295
Nitrogen availability−0·08216·3421≤ 0·00050·299
Distance−0·267 8·42610·00370·302
Dispersal potential birds−0·514 4·32310·03760·303
Constant−1·66716·7521≤ 0·0005 

mined sites as refuges for rare species

In total, 74 rare species from the Red List of Saxony-Anhalt were present in the 10 study sites (see Supplementary Material Table S1). In Table 5, the number of establishing events in all 10 mined sites is provided for all these species subdivided into four distance classes. Eighty-five per cent of the Red List species could only be found in one mined site. On the other hand, only one species was able to establish in all sites. Occurrences more than 17 km away from the mined sites were recorded for six species (e.g. Dactylorhiza fuchsii (Druce) Soó, Hieracium zizianum Tausch, Platanthera chlorantha (Custer) Rchb., Utricularia vulgaris L.).

Table 5.  Frequency of establishment and distance to nearest occurrence in the grid cells of the floristic mapping for Red List species at mined sites
 No. of mined sites with successful establishment of Red List species
12345678910
Nearest occurrence in the surrounding area
< 3 km1859 610 3  8
3–10 km208511 4 8  2
10–17 km 921 3 1 3   
> 17 km 133       
Total number of established Red List species48965302001

A comparison of immigrant and non-immigrant species present 0–3 km from the respective mined site showed that both the difference in seed weight and terminal velocity were highly significant for all species as well as for non-Red List species (Table 6). For the Red List species, only terminal velocity showed significant differences between immigrant and non-immigrant species. In addition to the regression analysis, this supports the hypothesis that species with adaptations to wind dispersal, and both low seed weight and low terminal velocity, have a higher probability of reaching suitable sites. In general, rare Red List species showed lower seed weights and terminal velocities compared to more common species in the same landscape unit. The higher mobility of Red List species, as indicated by lower seed weight and lower terminal velocity, increases their LDD potential.

Table 6.  Differences in species traits important for LDD between immigrant and non-immigrant species that are present in the 0–3 km surrounding area of the 10 mined sites with exclusion of woody species (phanerophytes, nanophanerophytes). SE, standard error; P, level of significance between immigrant vs. non-immigrant species
 All speciesOnly non-Red ListOnly Red List
ImmigrantNon-immigrantImmigrantNon-immigrantImmigrantNon-immigrant
  • *

    0.05 ≥P > 0.01,

  • ***

    P≤ 0.001.

Seed weight (mg)2·092·762·122·880·821·11
 n18281411178713154196
 ± SE0·110·170·110·180·160·12
 P (U-test)*** ***   
Terminal velocity (m s−1)1·992·392·012·421·361·94
 n16561078161010074671
 ± SE0·030·040·030·040·220·16
 P (T-test)*** *** * 

Discussion

Long-distance dispersal (LDD) is rare and largely driven by chance; therefore, it is very difficult to observe. The probability of catching single diaspores travelling over large distances in small seed traps is very low (Bullock & Clarke 2000). Colonization of large-scale open mined areas offers the opportunity to study LDD indirectly. Based on our analyses, we are able to assemble a general model that explains the colonization of large-scale, open and nutrient-deficient sites.

We assessed the amount of variance explained by the abundance of species in the vicinity of the mined sites (Hypothesis 1). Interestingly, the number of occupied grid cells in the regional species pool was more important than actual distance to the seed source. This could be due to large quantities of released seeds. By contrast, the proportion of successfully established rare species with very small seeds and a low terminal velocity was generally high at our study sites. Although these species have a low abundance in the surrounding area, they are able to bridge relatively large distances and accumulate in the mined sites, which act as huge seed traps in the landscape.

In addition to this abundance-driven ‘recruitment lottery’ for rare species, trait-related parameters also proved to be significant (Hypothesis 2) although less important than the abundance in the regional species pool. In particular, low terminal velocity, which is an important determinant of the ability for long-distance dispersal by wind (e.g. Bullock & Clarke 2000; Nathan et al. 2002; Tackenberg et al. 2003; Soons et al. 2004), increased the probability of local occurrence. Many rare species are well-equipped for LDD (e.g. Orchidaceae, Ophioglossaceae, Pyrolaceae). Bradshaw (1983) and Ash, Gemmell & Bradshaw (1994) observed in their studies on industrial waste heaps that species with very small seeds and specialized soil preferences (e.g. orchids) are able to bridge distances up to 40 km to reach suitable sites. In recent studies (e.g. Tackenberg et al. 2003; Nathan et al. 2005), weather conditions characterized by thermal turbulence and updrafts, but low horizontal wind speed, have been identified as important factors for long-distance diaspore transport. Additionally, it has been argued (e.g. Higgins et al. 2003) that non-standard means of dispersal are often responsible for LDD events (e.g. wind-dispersed seeds may occasionally be dispersed by birds).

In our large-scale mined areas, accumulation of species has been taking place continuously over several decades. The high number of species which could be found in only one or two study sites indicates that LDD events are not regular. Therefore, both small-scale restoration sites and short time frames will impede successful colonization. In comparison with short-distance dispersal, colonization via LDD is not predictable at the species level (e.g. Higgins et al. 2003, del Moral, Wood & Titus 2005). Suitable species groups with a high colonization probability can only be identified at the level of ecological species groups (e.g. pioneer species of acid and sandy sites with suitable dispersal traits).

While dispersal by birds is often identified as a crucial factor for LDD (e.g. Wilkinson 1997), dispersal by fur or dung was not of significance probably because the study sites are relatively isolated from suitable habitats (woodlands, grasslands) with insufficient corridors for animal movement (e.g. hedgerows). Other species traits, such as persistence in the soil seed bank or seed weight class, also did not play a significant role in our model. Seed morphology, such as wing-loading or surface structure, influences terminal velocity (e.g. Augspurger & Franson 1987; Andersen 1993), but in many cases seed weight class is correlated to terminal velocity.

Site conditions typical of pioneer stands, such as open space (increasing light availability) and nutrient deficiency (decreasing nitrogen availability), proved to be important factors for the colonization of adapted species (Hypothesis 3) (Ozinga et al. 2004; Rehounková & Prach 2006). In general, nutrient deficiency can be a major factor controlling colonization (see Ash et al. 1994). Additionally, Soons & Ozinga (2005) stated that, in times of increasing nutrient availability, plant nutrient requirements can outrank LDD for regional plant survival. The special features of large-scale mined areas additionally enhance immigration probability, especially for species that are adapted to open, nutrient-deficient habitats.

Irrespective of our results, it is important to keep in mind that extraction of minerals by surface mining destroys whole landscapes with all ecosystems involved. Functional ecosystems, especially when difficult to restore (e.g. Harris & van Diggelen 2006), should not be destroyed by mining activities. All interventions by mining must be compensated by target-oriented restoration measures. Therefore, it is important to examine how damage by mining can be compensated, and which restoration strategy can achieve the best possible result. In the temperate climate of Central Europe, one method of restoring surface-mined land could be the designation of large-scale areas reserved for spontaneous succession (e.g. Prach & Pyšek 2001; Prach 2003; Tischew & Kirmer 2007). In mined sites that are not restored, substrate, hydrologic and geomorphologic heterogeneity resulted in a large number of different niches for plant establishment (e.g. Tischew & Kirmer 2007). In fragmented landscapes, large-scale pioneer habitats offer a colonization opportunity for remnant populations. For species with low dispersal abilities, immigration can be facilitated by appropriate re-introduction measures (e.g. Kirmer & Mahn 2001, Donath et al. 2007).

This study has demonstrated for the first time that the occurrence of species up to a distance of at least 17 km plays an important role in colonization processes following large-scale destruction of ecosystems. Previous studies of primary succession also document that post-glacial colonization as well as the colonization of islands have to take place over large distances and during a time frame of several hundred years (e.g. Fridriksson 1975; Whittaker, Bush & Richards 1989; Jordan 2001). In this study, we have been able to prove that, in large-scale, nutrient-deficient, open sites, accumulation of species can be expected in time frames amenable to planning (several decades). In regions where such habitats have become rare and fragmented, restoration planning of derelict land (e.g. surface mines, quarries, landfills) that supports the creation of such conditions can contribute to the regional survival of rare pioneer species. In addition, the results underscore the importance of conserving remnant populations of rare species in the surrounding landscape. In areas reserved for nature conservation, it should be possible to await re-vegetation via spontaneous succession. This also applies to other potential sites where there is no danger of erosion and no time limit on re-vegetation. In future restoration planning, a new protocol should be established that exploits site potentials while also allowing for long-term spontaneous colonization processes.

Acknowledgements

Investigations were funded by the German Federal Ministry of Education and Research, the state of Saxony-Anhalt, the Lusatian and Central German Lignite Mining Management Company (grant no. 0339647, 0339747, and 0339770) and by the European Science Foundation (grant no. 3B_071-SURE and 05_EDIV_FP040-ASSEMBLE). We would like to thank all co-workers within the mentioned projects for their participation in collecting data. We are very grateful to Robert Freckleton (University of Sheffield), Jan Lepš, Karel Prach (both University of South Bohemia) and one anonymous referee for advice and constructive comments on earlier drafts of this paper and to Keith Edwards (University of South Bohemia) for language corrections.

Ancillary