Dispersal as a limiting factor in the colonization of restored mountain streams by plants and macroinvertebrates

Authors

  • Robert J. Brederveld,

    1. Ecology & Biodiversity Group, Department of Biology, Utrecht University, Padualaan 8, 3584 CH Utrecht, the Netherlands
    Search for more papers by this author
    • Present addresses: Witteveen+Bos Consulting Engineers, Ecology group, PO Box 233, 7400 AE Deventer, The Netherlands.

  • Sonja C. Jähnig,

    1. Department of Applied Zoology/Hydrobiology, University of Duisburg-Essen, D-45117 Essen, Germany
    Search for more papers by this author
    • Department of Limnology and Conservation, Senckenberg Research Institute and Natural History Museum, Clamecystrasse 12, 63571 Gelnhausen, Germany; Biodiversity and Climate Research Centre (BiK-F), Senckenberganlage 25, Frankfurt am Main D-60325, Germany.

  • Armin W. Lorenz,

    1. Department of Applied Zoology/Hydrobiology, University of Duisburg-Essen, D-45117 Essen, Germany
    Search for more papers by this author
  • Stefan Brunzel,

    1. Department of Animal Ecology, Faculty of Biology, Philipps-Universität Marburg, Karl von Frisch-Str. 8, 35032 Marburg, Germany
    Search for more papers by this author
  • Merel B. Soons

    Corresponding author
    1. Ecology & Biodiversity Group, Department of Biology, Utrecht University, Padualaan 8, 3584 CH Utrecht, the Netherlands
    Search for more papers by this author

Correspondence author. E-mail: m.b.soons@uu.nl

Summary

1. Over the past centuries, European streams have been heavily influenced by humans through pollution and regulation. As a result, the quality and diversity of freshwater riparian habitats have declined strongly, and the diversity of riparian flora and fauna has decreased. Recent restoration measures have resulted in stream habitat improvements, but biodiversity improvements have failed to follow in fragmented streams. It has been suggested that dispersal limitation could play an important role in the lack of biodiversity improvement in restored streams, but to date, there is no conclusive evidence for this assumption.

2. In this study, we investigated whether colonization of restored streams by plants and macroinvertebrates is limited by dispersal. We hypothesized that colonization success increases with increasing availability of (nearby) source populations and with increasing ability of species to disperse over long distances. We related species composition in seven restored stream sections to species’ abundances in the surroundings and to species’ dispersal abilities.

3. For both plants and macroinvertebrates, colonization success is strongly related to the abundance of species in the local and regional species pools.

4. For plants, dispersal strategy has an additional influence on colonization success: short-lived plants with high production of small, well-dispersed seeds colonized best within the 3- to 5-year period after restoration.

5. The existence of dispersal strategy constraints could not be confirmed in macroinvertebrates, possibly because these are limited by a lack of connectivity on larger spatial scales. On the landscape scale, beneficial effects of increased plant diversity might further improve habitat suitability for macroinvertebrates.

6.Synthesis and applications. Dispersal appears to be a limiting factor for successful (re)colonization of restored streams in fragmented landscapes. In plants, this is attributed to limitations in seed dispersal abilities and likely to a lack of nearby source populations as well. In macroinvertebrates, lack of nearby source populations may also be a limiting factor. Hence, we suggest restoring landscape connectivity at larger spatial scales and optimizing the availability of near-natural ‘source’ areas in the vicinity of restoration projects, at least for plants, to improve the success of biodiversity restoration in fragmented habitats.

Introduction

Human impacts on landscapes have often resulted in degraded landscapes with monotonous land use and low habitat diversity and, consequently, low biodiversity and biological functioning. Especially in freshwater ecosystems, anthropogenic influences have strongly reduced ecological quality (Giller 2005; Søndergaard & Jeppesen 2007). Ecological degradation is problematic because freshwater systems are of great importance, especially with increasing demands for freshwater and for good-quality water (e.g. EU Water Framework Directive; EC 2000). In streams, sewage loading and agricultural run-off have negatively influenced water quality, while modification and regulation have resulted in loss of natural flows and canalization has reduced the number and suitability of habitats; all of these processes have greatly reduced stream biodiversity (Søndergaard & Jeppesen 2007).

Many restoration projects have been carried out in attempts to restore the ecological quality of streams. Commonly used restoration measures include addition of large (parts of) trees to the stream channel (e.g. Kail et al. 2007), removal of bank fixations (e.g. Rohde et al. 2005) and excavation of the floodplain and/or excision of the riverbed (e.g. Gurnell et al. 2006). The hydromorphological effects of these restoration measures have improved habitat number and quality but have failed to positively influence biodiversity in highly fragmented streams (Larson, Booth & Morley 2001; Rohde et al. 2005; Spänhoff & Arle 2007; Jähnig et al. 2009a), although positive biodiversity effects are observed in near-natural streams (Helfield et al. 2007; Kail et al. 2007). It is important for stream restoration to reintroduce habitats, but in heavily fragmented streams, this has not always been sufficient to restore and conserve biodiversity (Suding, Gross & Houseman 2004; Palmer et al. 2005). Therefore, it is necessary to identify the bottleneck for riparian and aquatic biodiversity restoration. As habitat availability and quality are not likely to substantially limit recolonization following restoration projects (Rohde et al. 2005; Kail et al. 2007; Jähnig et al. 2009a), it has been suggested that dispersal limitation plays an important role (Rohde et al. 2005). Fragmentation of (restored) stream habitats hampers dispersal between patches (Jansson, Nilsson & Renöfält 2000), so species composition may be highly dependent on the local species pool (Renöfält, Nilsson & Jansson 2005; Lake, Bond & Reich 2007) and on dispersal of organisms (Nilsson et al. 2002).

As dispersal limitation is thought to play a critical role in the lack of success of biodiversity restoration measures in streams, it is important to know how riparian and aquatic organisms disperse. We focus here on two groups of organisms that are generally used for the assessment of restoration success and as indicators for stream quality: plants and macroinvertebrates (EC 2000). Dispersal by water is the most prominent form of plant dispersal in stream ecosystems (e.g. Nilsson et al. 2002; Gurnell et al. 2007, 2008). For macroinvertebrates, passive dispersal by water also occurs on a large scale, but there is also active compensatory upstream movement (e.g. Elliott 1971; Turner & Williams 2000). Dispersal by wind is also relevant, by dispersing plant seeds and invertebrates in lateral directions and to sites not connected to the source population by surface waters (e.g. Soons 2006; Vanschoenwinkel et al. 2008). Dispersal by animals is often used to explain upstream colonization (e.g. Pollux et al. 2007). Plant seeds and invertebrates can be dispersed by waterfowl (e.g. Figuerola & Green 2002; Soons et al. 2008), although in streams, this probably plays a smaller role than in more important waterfowl habitats, such as ponds and lakes. Different dispersal vectors thus have the potential to disperse plants and macroinvertebrates in streams, but fragmentation of habitats in riparian landscapes disrupts each of these dispersal strategies, reducing colonization ability and species survival (Soons & Ozinga 2005; Ozinga et al. 2009).

Thus far, the theory that dispersal limitation is a critical factor in the recolonization of restored streams has mainly been inferred from the lack of colonization after successful habitat restoration. Although the theory appears plausible, data supporting dispersal limitation as a critical factor for recolonization are missing. Therefore, we investigated the role of dispersal as a limiting factor in the colonization of restored stream sections by plants and macroinvertebrates. We hypothesized that the colonization of restored stream sections (i) increases with availability of nearby source populations and (ii) increases with species’ dispersal ability. We studied this hypothesis in streams in which hydromorphology and habitat heterogeneity have been successfully restored and compared the colonization success of plants and macroinvertebrates with different dispersal abilities and different abundances in the surrounding areas.

Materials and methods

Study system

We investigated colonization success in seven restored mountain stream sections. The restored sections had a multiple-channel pattern, as would naturally occur in such streams, while nearby unrestored sections had a man-made single-channel pattern. The restored sections were located in the rivers Lahn, Orke, Eder, Nims and Bröl in the lower mountains of the Rhenish Slate Mountains in Germany between 49°56′N–51°2′N and 6°29′E–8°51′E. Active restoration measures were taken for three sections from 2000 to 2002, while the other four sections were passively restored by halting floodplain use from 1995 to 2000 onwards. In the passively restored sections, morphological and habitat changes began following a lag phase of several years after halting floodplain use (Jähnig, Lorenz & Hering 2008).

All sections were surveyed in 2004–2005 for hydromorphological features and for plant and macroinvertebrate species distributions. This survey took place 3–5 years after the start of the active restoration measures or, in the case of passive restoration, 3–5 years after the start of the occurrence of natural changes in stream hydromorphology. The results of these surveys show that, in comparison with upstream unrestored sections, restored sections had significantly higher numbers of channel elements and aquatic and riparian mesohabitats but that the Shannon diversity indices of aquatic and riparian organisms showed no significant increase (Jähnig, Lorenz & Hering 2009b). Biodiversity thus appears to be unrestricted by habitat but might be limited by dispersal, making these sites ideally suited for our investigation.

To determine whether colonization of restored stream sections by plants and macroinvertebrates is limited by dispersal, we investigated whether colonization success is related to the abundance of species in the surrounding area and/or to the dispersal ability of the species. Two methods were used: (i) analyses with a high level of detail, based on species abundances, and (ii) analyses based on species’ presence/absence data. The first analyses could be performed only for plants, the latter for both plants and macroinvertebrates. In both methods, we related the species that colonized restored stream sections to the expected species composition in the restored sections based on a reference condition. The method of using a stream reference image to describe the macroinvertebrate reference condition is often applied in stream restoration ecology (the ‘reference condition approach’, e.g. Reynoldson et al. 1997; Bailey, Norris & Reynoldson 2004). We also applied this method to the stream and riparian flora, for a comparable approach.

Colonization success 1 – species abundance data

We first quantified colonization success by using vegetation survey data of the restored sections. These data were collected for each section along ten equidistant transects perpendicular to the main stream, laid out across a section of the main stream approximately 200 m in length (Jähnig et al. 2009a; Fig. S1, Supporting Information). The transects were 2 m wide and covered the entire restored area perpendicular to the stream in length, including the riverbed and the riparian zones (Fig. S1). Along each transect, all vegetation types present were identified (following Oberdorfer 1992), and the length of each vegetation type along the transect was measured. It was not practically feasible to make a complete vegetation survey at every single location along every transect within which a vegetation type was present; therefore, for each vegetation type present in a section, a representative number of vegetation surveys were taken. The number of surveys increased with the total cover of that specific vegetation type in the section. This approach resulted in an average of one vegetation survey per 7 m transect length and a set of vegetation surveys (varying from 1 to 24 surveys, with an average of 10 m2 per survey) to represent each vegetation type present in a section (Jähnig et al. 2009a). In each survey, the presence of all higher plant species was recorded. From these data, we calculated the abundance of each species per vegetation type per section as the percentage of vegetation surveys (of that vegetation type) in which the species was present. Given that there were multiple vegetation types per stream section, we then calculated the abundance of each species per section (hereafter called ‘actual abundance’) by averaging its abundance across all vegetation types present in the section weighted by the total cover of each vegetation type. To calculate colonization success, we then related these actual abundances of species to their expected abundances in the restored sections based on reference conditions of the vegetation types found. As reference condition, we used the detailed quantitative descriptions of well-developed vegetation types in the standard work by Oberdorfer (1992), which gives species abundances (as the percentage of vegetation surveys of that vegetation type in which the species was present, for the region that contains our study sites) for all vegetation types identified in the restored sections (hereafter called ‘expected abundance’). Expected abundances were also averaged across vegetation types (weighted by total cover of each vegetation type) to obtain one abundance value for each species per stream section. Colonization success was then calculated for each species in each restored stream section, as follows:

image(eqn 1)

A small value (0·1) was added to each abundance to prevent zero values in expected abundances, which would result in missing values in the analyses.

Colonization success 2 – species’ presence/absence data

In our second method, we used species presence/absence data per section. For this, we used survey data from the restored sections in combination with survey data from nearby upstream unrestored ‘control’ sections. This approach could be used for both plant and macroinvertebrate data. In this analysis, we grouped all species into the following four groups (relating to their colonization success) and compared these groups among each other: (1) species present in the restored section and the unrestored ‘control’ section, (2) species present only in the unrestored ‘control’ section and not in the restored section, (3) species present only in the restored section and not in the unrestored ‘control’ section and (4) species expected to be present in the restored section but not present. The species in groups 1 and 3 represent species that colonized successfully. The difference is that species in group 1 were also present directly upstream and could have colonized from there, whereas the species in group 3 had to colonize from elsewhere. The species in group 4 represent species that failed to colonize. The species in group 2 do not represent any level of colonization success; however, they help form a clearer picture of the group of species present upstream and are therefore included in the overview.

For plants, presence/absence of each species in each section was derived from the vegetation surveys described earlier, which recorded species presence. Following exactly the same design, these vegetation surveys were repeated in the upstream control sections. Expected presence/absence of plant species was determined from the vegetation types present in each section and from the presence/absence of species in these, following the vegetation descriptions in Oberdorfer (1992). For macroinvertebrates, the presence/absence of each species in each section was derived from macroinvertebrate surveys. Macroinvertebrate survey data were obtained for each section (restored and unrestored controls) using a shovel sampler (500-μm mesh size, 0·0625 m2 sampling area), according to the multihabitat sampling protocol (Hering et al. 2003), taking a shovel sample on each of the different stream substrates present in a section. Although abundances could be calculated from these detailed surveys (Jähnig & Lorenz 2008), species presence/absence data for each section were most reliable for the comparison and therefore these were used. The expected presence of macroinvertebrate species was determined from the species composition of the stream reference condition (or ‘Leitbild’); for the streams in our study, we used regional stream type number 9 ‘Silikatische, fein- bis grobmaterialreiche Mittelgebirgsflüsse’ or ‘mid-sized fine to coarse substrate-dominated siliceous highland rivers’ (Haase et al. 2004). Before data analysis, all species lists were adjusted to the same taxonomic identification level.

Occurrence of species in the surroundings

As a potential explaining factor for colonization success, we quantified the availability of sources of possible colonizers in the surroundings of the restored sections. We did this for two distances from the restored sections: (i) nearby: the abundance of species in the nearby upstream unrestored section, at a maximum distance of a few 100 m (hereafter ‘local species pool’) and (ii) regionally: the abundance of species in the 25 × 25 km area surrounding the restored section (hereafter ‘regional species pool’).

For plants, the abundance in the local species pool was derived from the species surveys in the unrestored control sections, in the same way that the actual species abundances for the restored sections were calculated. The abundance in the regional species pool was scored using the distribution of plant species following flora atlases of the area at a 5 × 5 km grid cell resolution (Becker, Frede & Lehmann 1996; Haeupler, Jagel & Schumacher 2003). Because of missing species distributions for two sites herein, abundances in the regional species pool could not be constructed for the sites Lahn Cölbe and Nims, and these were excluded from further analyses of plant data. For the other sites, abundance in the regional species pool was calculated for each species as the fraction of grid cells in the 25 × 25 km area in which the species was recorded as present.

For aquatic macroinvertebrates, we also used the species abundances at the unrestored control sections to represent the abundances in the local species pools. The abundance of species in the regional species pool was derived from a large data base of macroinvertebrate survey data (J. Böhmer, unpublished data 2008), using recordings from each 25 × 25 km area surrounding a restored section. Here, abundance was calculated for each species as the fraction of surveys in the area in which the species was recorded as present.

Dispersal ability

To determine the dispersal ability of the species involved, data bases on several dispersal traits were used. We selected species traits to quantify the different stages of dispersal. For the plant species, we used the following traits: seed production and life span to estimate propagule source strength; floating capacity, wind dispersal ability and seed longevity to describe spatial and temporal dispersal ability of these propagules; and habitat requirements, to correspond to germinating and growing conditions of the dispersed propagules. This selection of traits resulted in the following available variables: seed mass (mg), seed number per individual (count), plant life span [ordinal scale: annual (1), strict monocarpic bi-annual (2), perennial (3)], seed terminal velocity (m s−1) and seed buoyancy (percentage of floating seeds after 1 week), all of which were taken from the LEDA plant traitbase (Knevel et al. 2003; Kleyer et al. 2008); seed bank type (ordinal scale: transient [1], short-term persistent [2], long-term persistent [3], from Thompson, Bakker & Bekker 1997) and Ellenberg indicator values for moisture and nutrient requirements (ordinal scale of 1–12 and 1–9, respectively, from Ellenberg et al. 2004).

For aquatic macroinvertebrates, life span, aquatic locomotion ability and aerial dispersal ability were used as dispersal traits in the analysis. Life span was converted to a binomial score [<1 year (0), >1 year (1)]; aquatic locomotion type traits were weighted [swimming/skating (2), swimming/diving (2), burrowing/walking (1), sprawling/walking (1), (semi)sessile (0)] and summed to produce a locomotion score. Life span and locomotion type were taken from http://www.freshwaterecology.info (Euro-limpacs Consortium 2008). Aerial dispersal ability was scored using aerial stage information (no/yes) and flight ability (bad/good) on an ordinal scale [no flight ability (0), poor flight ability (1), good flight ability (2); A.W. Lorenz, unpublished data 2008).

Data analysis

Before analysis, we checked for correlations among the various explanatory variables (species abundances in the surroundings and species dispersal traits). Then, we related our first measure of colonization success (a continuous variable, eqn 1) to the potential explanatory variables using multiple regressions. After that, we compared the values of the potential explanatory variables between the discrete species groups identified as a second measure of colonization success using anova. Both methods provide complementary information, on the importance of individual variables in explaining differences in colonizing success and the comparison of those variables in a quantitative manner.

The multiple regressions were used to build models with the local and regional species pool data and species’ dispersal traits as explanatory variables in a multiple stepwise regression. As the dependent variable, we used the colonization success indicator (eqn 1). Regression residuals were checked for normality, and collinearity levels in all regressions were acceptable (all variance inflation factors <1·41). The anovas were carried out nonparametrically because of non-normality and heterogeneity within groups. We tested for differences in abundance in the local and regional species pools and in dispersal abilities between species in the different groups using Kruskal–Wallis tests in combination with Bonferroni-corrected pairwise Mann–Whitney U tests. Differences between groups for macroinvertebrate data were calculated nonparametrically for continuous variables and as a Pearson’s chi-square for ordinal data. Because of a low number of observations in groups 2 and 3 (0 and 11 observations, respectively), only differences between groups 1 and 4 could be calculated. Data were transformed for statistical analysis when necessary. All statistics were calculated using spss 15.0 (SPSS 2006).

Results

In total, 1970 plant and 1178 macroinvertebrate colonization events (a species colonizing one of the restored sections) were analysed.

Correlations among explaining variables

For the plant species, abundance in the local species pool and abundance in the regional species pool were significantly positively correlated (Spearman r = 0·272, < 0·001, N = 1970). This indicates that the species most abundant in the unrestored control section correspond to the species most abundant in the 25 × 25 km area surrounding the sections, and vice versa. Plant dispersal traits also showed correlations among each other (Table S1, Supporting Information), indicating the existence of different life-history strategies: seed mass is negatively correlated with seed number, indicating that species with light seeds produce more seeds, which are also well dispersed by wind (positively correlated with terminal velocity), float well (negatively correlated with buoyancy) and are persistent in the soil seed bank (negatively correlated with seed bank type). In particular, species with persistent seeds are also short-living species (negative correlation between seed bank type and life span).

For macroinvertebrates, species abundance in the local species pool and species abundance in the regional species pool were also significantly positively correlated (Spearman r = 0·609, < 0·001, N = 1178). This result indicates that the species that are most abundant in the upstream control section match the species that are most abundant in the area. Macroinvertebrate dispersal traits also showed significant positive correlations (Table S2, Supporting Information) between life span and locomotion type and also between locomotion type and aerial dispersal ability, suggesting that macroinvertebrates that are more mobile in the water also have higher aerial dispersal ability.

Colonization success 1 – species abundance data

For plants, the linear multiple regressions carried out for each site separately and for all sites together (Table 1) show that plant life span and abundance in the local species pool are significant explanatory variables in all regressions. Abundance in the local species pool was positively related to colonization success, whereas life span predominantly was negatively correlated, indicating that colonization mainly occurs by species already present upstream and by species with short-living individuals. Furthermore, seed mass was negatively related to success of colonization and positively to seed bank type, indicating that species with smaller seeds and long-term persistence in the seed bank are more successful in colonization of restored sites. The regional species pool was also positively related to colonization success in almost all sites.

Table 1.   Regressions on colonization success for the individual streams and all streams joined show life span and abundance in the local species pool (LSP abundance) as significant parameters in every stepwise regression (in bold). Additional parameters included in the models are abundance in the regional species pool (RSP abundance), seed mass, Ellenberg moisture value, seed bank type and Ellenberg nutrient value
SiteRegression statisticsVariables includedStandardized regression coefficientP
R2Pdf
B (1995)0·2630·000176RSP abundance B0·2450·000
Life span−0·2800·000
LSP abundance B0·2530·000
Seed mass−0·1530·021
O (1998)0·2310·000187Life span−0·2670·000
LSP abundance O0·2540·000
Seed mass−0·2170·001
RSP abundance O0·1570·018
E (2000)0·2230·000181LSP abundance E0·2540·000
RSP abundance E0·2140·002
Life span−0·2300·001
Ellenberg moisture0·2010·003
W (2001)0·2000·000238Life span−0·2750·000
LSP abundance W0·2430·000
Seed bank type0·1740·006
L (2002)0·2520·000260Seed bank type0·2290·000
LSP abundance L0·2720·000
Life span−0·1700·004
RSP abundance L0·1320·018
Seed mass−0·1280·032
All sites0·2130·0001062LSP abundance0·2240·000
Life span−0·2110·000
Seed mass−0·1680·000
RSP abundance0·1750·000
Seed bank type0·0920·004
Ellenberg nutrient0·0730·012

Colonization success 2 – species’ presence/absence data

For the plant species, the four groups reveal differences in most of the variables (Fig. 1); there are significant differences between the groups in abundance in the regional species pool (< 0·001), seed mass (< 0·001), life span (< 0·001), seed bank type (< 0·001), Ellenberg moisture (< 0·001) and Ellenberg nutrient (< 0·001) values. Seed number was shown as marginally significant (= 0·044) but failed to show significant contrasts.

Figure 1.

 Means ± standard errors for different plant species’ parameters show significant differences between groups for most parameters. Group numbers represent (1) species present in both the restored and the nearby unrestored stream sections, (2) species present only in the unrestored sections, (3) species present only in the restored sections and (4) species expected to be present in the restored section but failing to colonize. Significant differences are indicated by different letters.

The regional species pool data show that the species found in both unrestored and restored sections (group 1) also have the highest regional abundances, whereas the expected species that failed to colonize (group 4) have the lowest regional abundances. This means that the species that colonized least well are the species with the lowest abundance in the regional species pool. Seed mass of the species present only in the restored sections (group 3) is lowest, whereas the expected species that failed to colonize (group 4) have the highest seed mass, implying that colonization occurs most by species with low seed mass and least by species with high seed mass. A similar pattern can be seen for life span: species best able to colonize restored sites have the shortest life spans. Seed number and seed bank type show inverse patterns, indicating that the species best able to colonize the restored sections have higher seed numbers and seed bank persistency. The seed dispersal traits terminal velocity and buoyancy show no relation to colonization success. Furthermore, Ellenberg indicator values for moisture requirements tend to be lower in the species present only in the restored section and in the expected species group (groups 3 and 4) than in the groups of species already present in the unrestored sections (groups 1 and 2). Ellenberg indicator values for nutrient requirements are highest in the species found in both sections (group 1) and significantly lower in the other groups. These results on the Ellenberg values indicate that there are no differences in moisture and nutrient requirements between the species that colonized and the species that failed to colonize.

For the macroinvertebrate species, there are significant differences between groups in abundance in the regional species pool and in aerial dispersal ability (Fig. 2). Abundance in the regional species pool is higher for species present in both sections (group 1) compared to the species expected but not found (group 4). Also, the low number of observations in groups 2 and 3 shows that if species occur, they are mostly found in both sections. As with plants, this means that the species that do not colonize are the species with the lowest abundance in the regional species pool. Aerial dispersal ability is higher for the expected species that failed to colonize (group 4) in comparison with the species occurring in both sections (group 1). This latter result contrasts with our expectations on dispersal ability limitation. There are no differences in life span or aquatic locomotion among the macroinvertebrate groups.

Figure 2.

 Means ± standard errors for different macroinvertebrate species’ parameters. Group numbers represent (1) species present in both the restored and the nearby unrestored stream sections, (2) species present only in the unrestored sections, (3) species present only in the restored sections and (4) species expected to be present in the restored section but failing to colonize. Significant differences are indicated by different letters. Group numbers 2 and 3 had very low numbers of observations and were therefore omitted.

Discussion

Our results clearly show that abundance in the local species pools and abundance in the regional species pools are important factors in predicting colonization success: species with a higher abundance nearby and/or in the surrounding area are more likely to colonize restored sections. This means that either there is strong dispersal limitation in the form of source limitation, or the species that fail to colonize have much more specific habitat requirements (that are rarely found in human-dominated, fragmented landscapes), or both. In the case of the plant species, the latter assumptions are unlikely because the species that fail to colonize (group 4) have no different habitat optima than species that colonize (group 3), based on their Ellenberg indicator values for moisture and nutrients. However, differences in habitat requirements cannot be ruled out completely based on this information. For the macroinvertebrates, specific habitat requirements cannot be ruled out at all and may contribute significantly to our findings. Hence, our results show that, to succeed in restoring fragmented streams, it is likely to be of great importance to have well-developed vegetation communities in the vicinity of restored areas to function as sources for colonization, and this could be similar for macroinvertebrate communities.

Dispersal ability-related plant traits appear to be crucial factors contributing to colonization success in plants. This appears to be a reflection of plant strategies, pointing towards the existence of suites of related traits for successfully colonizing restored sites. The plant species with high colonization success are short-living species that produce many small, light seeds. Such seeds disperse well via different vectors: in wind owing to low terminal velocity (Soons et al. 2004), in water owing to high buoyancy (Nilsson et al. 2002; Boedeltje et al. 2004), in waterfowl owing to unimpaired gut passage (Soons et al. 2008) and in time owing to high persistence in the seed bank (Thompson, Bakker & Bekker 1997). The fact that this combination of plant traits closely resembles that of r-selected or pioneer species shows how successful such strategies are for colonizing dynamic environments.

Surprisingly, plant species that failed to colonize show no difference in terminal velocity and buoyancy of seeds compared to the other species. We discovered that these results arise from a bias in available data for these plant traits; most data have been collected on good dispersers, i.e. species with low terminal velocity and high buoyancy. The average seed mass for species with terminal velocity data available is significantly lower than that for species lacking terminal velocity data (< 0·001). Regarding buoyancy, the species with buoyancy data available are also not a representative subset of the total species list (< 0·001). Given that almost all data available on these traits are for well-dispersed species, these variables understandably reveal no significant differences between the groups. Thus, an additional important finding of our study is that further data collection on dispersal traits is necessary, so that more representative analyses dealing with large numbers of species can be carried out.

For macroinvertebrates, dispersal parameters were generally without significance, except for aerial dispersal ability, which was higher in the species that failed to colonize. This unexpected finding could have resulted either from a lower abundance of species with high aerial dispersal ability in the regional species pool because of lack of connectivity, or from the existence of highly specific habitat requirements for flying macroinvertebrates. Abundance in the regional species pool is positively and significantly related to aerial dispersal ability [phi (nominal correlation) r = 0·564, = 0·016], suggesting habitat limitation in the restored streams. Macroinvertebrates are evidently capable of (rapid) dispersal over more than a few tens of metres (e.g. Griffith, Barrows & Perry 1998; Petersen et al. 2004), but might be limited by the lack of connectivity on larger spatial scales in such a manner that their dispersal distances are outweighed by the distances between habitat remnants in the fragmented study landscapes. This may imply that even for well-dispersing aquatic macroinvertebrates, habitats are currently isolated (see also Soons et al. 2005; for plants), although further research is needed on dispersal in aquatic macroinvertebrates (Malmqvist 2002; Bohonak & Jenkins 2003). There are also indications that macroinvertebrates benefit from microhabitats created by colonizing plants, because large pieces of wood and living parts of plants provide additional high-quality substrates for macroinvertebrates in restored sections (Jähnig & Lorenz 2008). Therefore, increasing plant diversity in restored sections might interact with macroinvertebrate diversity by enhancing habitat diversity and ameliorating biotic constraints towards suitable conditions for macroinvertebrates, perhaps especially flying species.

In conclusion, dispersal limitation appears to be a determining factor for colonization success in restored streams for plants: their colonization is clearly limited by dispersal ability, favouring short-lived life-history strategies with high dispersal abilities, and probably also by (lack of) propagule sources in the surroundings. These results are in agreement with those of a recent study by Kirmer et al. (2008) on dry ecosystems. For macroinvertebrates, our data suggest that dispersal limitation by lack of availability of propagule sources in the relevant species pools may play a role, but our results are not conclusive. Especially in flying macroinvertebrates, habitat limitation may also (still) play a role 3–5 years after restoration. Macroinvertebrate dispersal ability data did not reveal any limitation in dispersal capacity, which may have been because of the general low resolution of information available on dispersal in macroinvertebrates and other factors such as lack of habitat and habitat connectivity at larger spatial scales. Nevertheless, macroinvertebrates can take advantage of increasing plant colonization, as plants can further increase suitable microhabitats.

The results presented here show that, to improve the success of restoration projects regarding biodiversity, plant dispersal limitation owing to propagule source limitation and limited dispersal ability should be taken into account. Restoration success is often measured as the similarity between biodiversity at the restored site and a reference condition. However, if colonization by plants (and, following that, macroinvertebrates) is limited by dispersal, restoration success can also be limited by the availability of propagule sources and species’ dispersal abilities. These factors should be considered when planning restoration activities. To overcome propagule source limitation, it is crucial to strategically plan restoration areas within the reach of (natural) areas with potential source populations (Soons 2006; Kirmer et al. 2008). After taking this first step, the creation of stepping stones (Jähnig & Lorenz 2008) of restored areas can be a successful strategy to shorten distances and improve connectivity between dispersal sources and restoration areas. This approach is especially important, given that in plants, abundance in the nearby (directly upstream) local species pool is an even better predictor of colonization success than abundance in the regional species pool (in the surrounding 25 × 25 km), indicating that colonization is most successful over very short distances or via the seed bank.

Dispersal as a limiting factor for the colonization of restored stream sections is mainly a problem in landscapes with high habitat fragmentation. A landscape-scale approach to planning restoration projects is of major importance for improving biodiversity and achieving good ecological status of streams in the near future. To surmount the limitation in dispersal ability, species with low colonization potential, e.g. species with high seed mass and therefore high terminal velocity and low buoyancy or species with long life span and short seed bank persistence, could be aided in colonizing restored sites (e.g. Hulme 2005; Donath et al. 2007). Alternatively, autonomous colonization might take a very long time, thereby affecting restoration success as measured by nature and/or landscape managers.

Acknowledgements

This paper was funded under the European Union Sixth-Framework-Programme Integrated Project Euro-limpacs (GOCE-CT-2003-505540) and the research funding programme LOEWE (Landes-Offensive zur Entwicklung Wissenschaftlich-Ökonomischer Exzellenz) of Hesse’s Ministry of Higher Education, Research, and the Arts. M.B.S. acknowledges funding by the Netherlands Organization for Scientific Research (NWO). We thank I.C. Knevel for providing data from LEDA, the environmental agencies of the covered federal states for their data on macroinvertebrate distributions and N. Smits and H.J. During for assistance with vegetation classifications.

Ancillary