Effects of re-braiding measures on hydromorphology, floodplain vegetation, ground beetles and benthic invertebrates in mountain rivers


*Correspondence author. E-mail: sonja.jaehnig@uni-due.de


  • 1Medium-sized and large mountain rivers are among the most degraded river types in Europe and numerous river restoration projects are currently carried out to achieve ‘good ecological status’. Surprisingly little is known about the effects of river restoration measures on aquatic and terrestrial organisms. We investigated the effects of restoration on hydromorphology, floodplain vegetation, ground beetles and benthic invertebrates of Central European mountain rivers by comparing seven restored, multiple-channel sections with seven nearby non-restored, straight sections.
  • 2Floodplain mesohabitats and aquatic microhabitats were recorded along 20 transects per river section (200 m long). Samples of floodplain vegetation (444 samples), ground beetles (153) and benthic invertebrates (134) were taken per habitat type and section. Two hydromorphological metrics and 13 biotic metrics were calculated.
  • 3The number of floodplain mesohabitats was significantly higher in restored sections, but there was no significant effect on the number of aquatic microhabitats. Floodplain vegetation reacted most strongly to restoration, with more vegetation assemblages and higher number of species in restored sections. The number of ground beetle species also increased, but there was no effect on number of species or diversity of benthic invertebrates.
  • 4Habitat composition and assemblages were compared by cluster analysis. When using mesohabitat data, restored vs. non-restored sections clustered to separate groups, while the use of aquatic microhabitat data produced mixed groups. Floodplain vegetation data clustered in restored and non-restored sections. For benthic invertebrates, the restored and non-restored sections of each individual river were always clustered together. Ground beetle assemblages responded more strongly to restoration than benthic invertebrates but less than floodplain vegetation.
  • 5 Synthesis and applications. River restoration measures which re-created multiple-channel patterns differ in their effect on floodplain vegetation, ground beetles and benthic invertebrates. The strong increase in the number of floodplain vegetation species is due to the creation of additional habitats, while riparian ground beetles react mainly to the increased availability of gravel bars. The lack of response of benthic invertebrates to restoration measures is due to the comparatively small changes in aquatic microhabitat composition. Our results indicate that floodplain habitats react more strongly to re-braiding as a restoration measure compared to in-stream habitats and that floodplain communities might be best suited to judge the immediate effects of restoration.


The objectives of physical river restoration projects are manyfold: they may target the restoration of near-natural hydromorphology and biota, aim at conserving or enhancing biodiversity patterns or at the restoration of ecosystem functions, or aim to restore services such as flood retention (Nienhuis & Leuven 2001; Bond & Lake 2003; Jansson, Nilsson & Malmqvist 2007). In Europe, the main drivers for river restoration are legal requirements: the EU Habitats Directive requires the conservation of selected floodplain species and communities, while the EU Water Framework Directive (WFD) aims at a ‘good ecological status’ for all rivers by 2015, which is defined through the biota (aquatic vegetation, benthic invertebrates, and fish fauna). Yet, about 70% of the water bodies in Europe are at risk of failing the WFD objectives or data are insufficient to assess their ecological condition (EU Commission 2007).

In Germany, hydromorphological degradation is the most widespread reason for the poor ecological status of rivers, affecting 67% of river length, while organic pollution, which was the predominant stressor type in European rivers for most of the 20th century, only affects 34% (BMU 2006). Consequently, river restoration projects in Germany have been directed towards the improvement of the hydromorphology. In three Federal States in West-Germany (total area: 75 000 km2), more than 1390 restoration projects have been carried out in the last two decades (C. K. Feld, D. Hering, S. Jähnig, A. Lorenz, P. Rolauffs, J. Kail, H. P. Henter & U. Koenzen, unpublished). Most of them were single measures, e.g. targeting the passability of individual weirs, and few projects included the hydromorphological improvement of several kilometres of river length.

One of the main problems with river restoration projects is the lack of information on their effects: a comprehensive overview of projects in the USA revealed that only 10% of more than 37 000 projects have been monitored (Bernhardt et al. 2005). Although precise data are lacking, similar figures are expected for Central Europe (Bratrich 2004). In the German Federal State of North Rhine-Westphalia only 6·4% of 426 recent river restoration projects have been monitored by investigating at least a single organism group before and after restoration (MUNLV 2005). Comparative investigations of restoration effects on different organism groups are lacking completely.

To fill this gap in our knowledge, we investigated the effects of restoration measures by comparing restored and non-restored sections of rivers. We focused on projects that re-created natural channel forms and channel dynamics (as a prerequisite of enhanced habitat diversity and near-natural ecosystem functions) in medium-sized Central European mountain rivers. Depending on size, slope, discharge and sediment, many Central European mountain rivers are characterized by multiple-channel patterns under near-natural conditions (LUA NRW 2001). Multiple-channel rivers offer a particularly diverse habitat matrix and correspondingly diverse aquatic and riparian biota (Tockner et al. 2006). Although historically abundant, these rivers have almost completely vanished from Central Europe outside the Alps; starting in the 18th century, hydrological engineering straightened the rivers, leading to a dominance of single-channel rivers (Gurnell & Petts 2002). Today, medium-sized and large mountain rivers are among the most degraded river types in Germany (Böhmer et al. 2004). The need for hydromorphological restoration is therefore particularly high. Only few (approximately 15) short sections have recently been restored back to multiple-channel systems (Jähnig, Lorenz & Hering 2008).

The aim of this study was to compare restored multiple-channel river sections with non-restored single-channel sections in terms of channel and floodplain morphology, and to determine restoration effects on aquatic and riparian organism groups, which were selected to cover a lateral gradient from the river to its floodplain. Benthic invertebrates are the most commonly used organism group in river biomonitoring (Hering et al. 2006), while ground beetles (Carabidae) indicate the intensity of aquatic–terrestrial interactions (Hering & Plachter 1997; Paetzold, Yoshimura & Tockner 2008). Floodplain (riparian) vegetation, as a floodplain organism group, is an important component of the EU Habitats Directive.

This study is based on the following hypotheses: (i) restoration increases habitat diversity and availability; (ii) increased habitat diversity enhances species richness and diversity of the organism groups investigated, that is, the restored sections should display a higher number of floodplain and river habitats, as well as higher species richness and diversity measures of the organism groups, compared to the non-restored sections. Different communities can be distinguished in restored and non-restored sections.

Materials and methods

We investigated seven restored multiple-channel sections and neighbouring non-restored single-channel sections (referred to as ‘paired sites’; Table 1). The paired sites were selected based on an intensive screening of recent restoration measures in Germany; we selected sites that were unpolluted and where multiple-channel patterns developed following restoration. Information on hydromorphology and biota prior to restoration was not available. All paired sites were located in the central German highlands. They comprised the upper parts of the rivers Lahn (three paired sites), Bröl and Nims in the Rhine catchment, and Eder and Orke in the Weser catchment. The catchment size at sampling sites ranged from 180 to 650 km2. Catchment geology is mostly comprised of acid rock (schist), with the exception of the Nims (with 60% carbonate rock). Land use in the study catchments consists of forest (60%), pasture and agriculture (30%) and about 10% urban areas (Corine land cover data Germany 2000). According to the German saprobic system, water quality of all streams is classified as ‘good’, indicating that the streams are not polluted by organic substances. However, most medium-sized rivers in Central Europe were severely polluted in the past, which probably still affects the current assemblages of aquatic organisms. Although several smaller weirs were present upstream of the restored sites, coarse sediment was transported over the weirs at high water levels. There was no water abstraction for hydropower generation at any of the sites.

Table 1.  Characteristics of sites, catchments and restoration measures. Data are valid for both non-restored and restored sections. All non-restored sections are located approximately 200 m upstream of restored sections, except Nims (3·4 km upstream) and Bröl (2·7 km downstream). The terms ‘Lahn-W’, ‘Lahn-LH’ and ‘Lahn-C’ name different paired sites of the river Lahn. Section area, hydromorphological status and mean channel width compare values of non-restored vs. restored section
Site nameLahn-WLahn-LHLahn-COrkeEderNimsBröl
  • m a.s.l., metres above sea level. *approximately 200 m thalweg (following the lowest points of the valley) between embankments.

  • LAWA (2002): The result of the standardized hydromorphological assessment is given as a number between one and seven: 1, unimpaired; 2, slightly changed; 3, moderately changed; 4, distinctly changed; 5, obviously changed; 6, strongly changed; 7, completely changed.

  • from Jähnig et al. (2008).

  • §¶

    Proportion and potential proportion of restored (multiple-channel) sections are based on actual / potential section length compared to stream length from source to restored site.

  • **

    derived from the position along the river and the designated length of multiple-channel reaches (documented in Sommerhäuser & Pottgiesser 2005).

Latitude (N)50°55′37″50°55′29″50°51′47″51°9′8″51°1′38″49°56′48″50°49′36″
Longitude (E)8°29′20″8°29′59″8°47′25″8°50′37″8°34′21″6°29′3″7°22′58″
Catchment size (km2)278288650289480222181
Section area (ha)*0·29–0·620·31–1·140·48–1·240·43–0·920·73–0·740·35–0·550·45–1·04
Altitude (m a.s.l.)300300190300300240104
Mean discharge (m3 s−1)5·15·28·36·310·52·83·4
Local channel slope (m km−1)0·210·400·200·450·200·480·60
Restoration year and measure2001 (active), excavation of floodplain material to the river bottom level, one vegetated island was spared; initiation of one secondary channel2002 (active), excavation of floodplain material to the river bottom level, one vegetated island was spared; initiation of one secondary channel2000 (active), initiation of two secondary channels, removal of bank fixation to initiate bank erosion1998 (passive), naturally initiated by tolerated gravel bar development2000 (passive), initiated by a fallen tree and subsequent tolerated bank erosion and gravel bar development1996 (passive), naturally initiated by tolerated gravel bar development1995 (passive), naturally initiated gravel bar development by fallen trees in the absence of bank fixations
Hydromorphological status6–36–33–25–35–35–35–2
Mean channel width (m)18·5–56·015·8–67·325·5–57·822·9–49·234·6–39·317·9–31·419·0–54·2
Distance to source (km)25 2650 31744430
Length of restored sections (m)2004001000200250200600
Proportion of restored (multiple-channel) sections (%)§ 0·81·520·60·60·32
Potential length of multiple- channel sections (m)6000700036 00015 50040 00017 00011 800
Potential proportion of multiple- channel sections (%)**24  277250543939
Catchment land use (%)
 Urban4·4 5·3 5·4 1·23·0 1·8 5·1
 Agriculture1·7 3·516·218·52·412·1 0·8

Three of the restored sections resulted from active restoration measures aiming at the improvement of hydromorphological conditions and biota, while four had developed naturally because maintenance was abandoned several years ago (‘passive restoration’). Following removal or decay of bank fixations, floods generated side-arms and additional habitats. Restored river sections occupy more of the valley floor, but probably less than the former extent due to land use restrictions. The study sites had at least one secondary channel besides the main channel, which was separated by an island or mid-channel bar; standing water bodies were usually present. Non-restored sections were characterized by a singular main channel, which was usually fixed at the banks. In a recent nationwide survey of river hydromorphology (LAWA 2002), the restored sections were assessed as ‘slightly’ or ‘moderately changed’; none was rated as ‘unimpaired’. The non-restored sections were assessed as ‘obviously’ or ‘strongly changed’ (Table 1). The sections were chosen to belong to the same stream type (thus having similar restoration aims), have similar catchment geology and land use, with at least 2 years of development time following restoration and a length of at least 200 m. In relation to the overall stream length from the sources to the restored sites the multiple-channel sections comprised between 0·6 and 2·0% of the river length (Jähnig, Lorenz & Hering 2009).

We undertook the following steps: (i) a hydromorphological survey on habitat composition at two spatial scales, i.e. mesohabitats of river and floodplain, and microhabitats at the (aquatic) river bottom; (ii) habitat-specific sampling of floodplain vegetation and ground beetles in the mesohabitats and of benthic invertebrates in the aquatic microhabitats, respectively; (iii) generation of overall taxa lists per section by calculating average communities (considering respective taxa densities) per habitat and combining the habitat-specific lists according to the relative proportion of habitats; and (iv) comparison of non-restored and restored sections based on metrics and cluster analysis.

Field investigations were carried out at every site, either in May 2004 or in June 2005; non-restored and restored sections of a river were sampled on consecutive days.


The standard experimental unit was a 200-m stretch of river and adjacent floodplain. The width of mesohabitats was measured along 20 equidistant transects running between the left and right edge of the embankments (Fig. 1). This area was coherent to the area flooded at mean maximum annual discharge (referred to as ‘floodplain’ henceforth). Partly based on the River Habitat Survey protocol (Raven et al. 1997), the mesohabitats included six aquatic (main channel, secondary channel, connected and disconnected side arm, permanent and temporary standing water body), three transient (bank, mid-channel bar and side bar) and three terrestrial elements (embankment, vegetated islands and floodplain area) (Jähnig et al. 2008). At the Lahn-W site, investigations were limited to 16 transects per section due to limited access to the riverbed by overgrown, steep embankments and water depth deeper than 150 cm.

Figure 1.

Sampling scheme: example from site Lahn-C. Hydromorphology was assessed by measurement of mesohabitat width between left and right edge of embankments; microhabitat composition was determined in aquatic parts. Ground beetles were sampled by pitfall traps or hand-sampling along three transects including all terrestrial and transient mesohabitats recorded. Floodplain vegetation was mapped on every second (hydromorphological) transect. Benthic invertebrates were sampled in main and secondary channels and backwaters where present, on all occurring substrates.

In the aquatic (i.e. submerged) sections of the transects, microhabitats were examined at 20 equidistant points (each point 25 × 25 cm). Thus, a total of 400 points were recorded per river section, yielding the overall number of different microhabitats per section and substrate composition (%). Microhabitats were distinct substrate types as described in the multi-habitat sampling protocol (Hering et al. 2003): blocks (> 200 mm), cobbles (> 60–200 mm), coarse gravel (> 20–60 mm), fine gravel (> 2–20 mm), sand (> 0·006–2 mm), loam (< 0·006 mm), submerged macrophytes, living parts of terrestrial plants, coarse particulate organic matter, organic mud. If several substrates were present at a point, this complex substrate composition was reflected by weighting the substrates according to their (co-) occurrence at the measuring point. Substrates were weighted threefold (if they covered more than 60% of the relevant area), twofold (if they covered 40–50% or when steep banks, large tree roots or living parts of terrestrial plants added other substrates to those covering the bottom), while other substrates were unweighted.

floodplain vegetation

Vegetation was mapped between the embankments in June 2005 along 10 transects of 2 m width (every second transect of the hydromorphological survey). Vegetation assemblages were classified to vegetation types according to Oberdorfer (1983, 1992) and Ellenberg (1996) to the order level. Where phytocoenological units did not apply, we used ‘aquatic vegetation’ and ‘gravel bars’ as units. Every classified vegetation assemblage (from here on ‘vegetation unit’) of each vegetation type along the transects was sampled. This resulted in one to 24 samples taken for each vegetation type present at a section (on average one sample per 7-m transect length; see Supporting Information, Table S1), depending on the number of vegetation units. Since the study aimed to compare total species richness and total numbers of habitat types and sizes of non-restored and restored sections, each vegetation unit was sampled completely: every species which was identifiable in the sampling period at the end of June was recorded. For each sample (on average 10 m2), all species occurring were recorded and their coverage was estimated using the following coverage classes: 1%, 5%, 10%, 15%, 20% and continuing in 10% steps up to 100%.

ground beetles

Ground beetles were sampled in vegetated habitats, open gravel and sand bars. In all vegetated habitats, pitfall traps (250 mL volume) were exposed for one week in June 2005; on open gravel and sand bars, an area of 0·82 m2 was hand-sampled. All terrestrial and transient mesohabitats mapped in the hydromorphological surveys were investigated; in non-restored sections, samples were taken only on embankments and twice on side bars (Orke and Eder), while in restored sections, samples were taken on embankments, side bars, mid-channel bars, vegetated islands and floodplains. The number of samples ranged from six to 17 (Supporting Information, Table S2), and was dependent on the total area covered by the mesohabitat (on average one sample per 20-m of transect length). All individuals captured were identified to species level.

benthic invertebrates

Altogether, 134 macroinvertebrate samples were taken in late June and July using a shovel sampler (500 µm mesh size, 0·0625 m2 sampling area). Each available substrate (mapped in the hydromorphological survey) was sampled independently of its frequency and relative cover so that important but area-limited substrates were assessed equally (Supporting Information; Table S3). The sampling scheme considered main channel, secondary channels and backwaters, if present. Samples from Lahn-C and Orke were taken in 2004 for detailed investigation of substrate-specific assemblages between non-restored and restored sections (Jähnig & Lorenz 2008). As the data collected in 2004 did not show variability of substrate-specific assemblages, the sampling design was simplified for 2005 and each substrate type was sampled once in the non-restored section and once in the restored section. No substrate occurred solely in the non-restored sections. Substrates occurring only in the restored section were sampled twice to ensure comparable sampling effort for individual substrates. Samples were preserved in 70% ethanol and sorted in the laboratory following the RIVPACS-sorting scheme (Murray-Bligh et al. 1997). The organisms were identified to species level, except Oligochaeta which were recorded as such or identified to family level, and Chironomidae, identified mostly to tribe level.

data processing and analysis

Morphological data included the total length of mesohabitats (m) and relative microhabitat composition (%). The number of meso- and microhabitats is inherent in these data. Quantitative taxa lists of each organism group and their abundance were generated for the individual sections from the habitat-specific samples. First, an average taxa list for each habitat was generated by averaging the species’ abundances for all samples in the respective habitat. Secondly, the abundances in the habitat-specific taxa lists were multiplied with the proportion of the respective habitats in the particular sections (habitat-weighting). Thirdly, for each section, the habitat-weighted taxa lists resulting from the second step were summed. The relative habitat proportion available for the individual organism groups was taken from the length of vegetation units along transects (floodplain vegetation), from the length of mesohabitats along transects (ground beetles), and from the proportion of microhabitats in the aquatic zone (benthic invertebrates), respectively. In all cases, habitat area was significantly larger in the restored sections. However, we did not correct the taxa number for area, since the increase in habitat area was an effect of restoration.

Data analysis comprised (i) comparison of the seven restored and the seven non-restored sections using a set of 15 metrics describing the (absolute) composition of habitats and organism groups, and (ii) a quantitative comparison of habitat and taxa data from all 14 sections. Richness and diversity metrics used for all organism groups were: number of taxa, number of genera, number of families (only for floodplain vegetation and benthic invertebrates), and Shannon–Wiener Diversity. We calculated additional richness metrics with the floodplain vegetation and the ground beetle lists: the number of floodplain vegetation types and the proportion of riparian ground beetle species (according to Koch 1993). Hydromorphological metrics applied were the number of mesohabitats and the number of aquatic microhabitats. For all metrics, the group of restored sections and the group of non-restored sections were tested for differences with the Mann–Whitney U-test.

The following data were subject to cluster analyses: mesohabitat widths standardized to mean width of the respective section (%), microhabitat composition (%), and taxa abundance. For cluster analysis, Soerensen (Bray–Curtis) Index and flexible beta linkage method (flexible beta set as −0·25) were applied (PCOrd5, McCune & Mefford 1999).


comparison of non-restored and restored sections

Seven out of 15 metrics were significantly different between non-restored and restored channel sections (Fig. 2). The number of mesohabitats on average doubled in the restored channel sections, indicating large structural changes in the floodplains. The mean length of mesohabitats increased by a factor of 3·0 (terrestrial parts), 1·4 (aquatic parts) and 5·9 (transient parts) (Jähnig et al. 2008). The number of aquatic microhabitats was not significantly higher, although quantitative microhabitat composition differed between non-restored and restored sections: the share of the dominant substrate was on average reduced from 75 to 62% and larger amounts of smaller mineral grain sizes and organic substrates were observed in the restored sections (Jähnig et al. 2008).

Figure 2.

Comparison of hydromorphological and biotic metrics between non-restored and restored sections (Median; Box: 25–75%; Whisker: min–max); * = differences are significant at P < 0·05 (U-test).

All metrics calculated for the floodplain vegetation, except Shannon–Wiener Diversity, were significantly different between non-restored and restored sections (P < 0·05; Fig. 2). On average, two additional vegetation types occurred in the restored sections. While in some non-restored sections (Lahn-W and Nims) only two vegetation types were present, at least five vegetation types were found in each restored section. Altogether, 287 plant species were recorded (60–173 species per section; mean = 103 species). The median species richness was 60 in the non-restored and 125 in the restored sections. The median number of genera increased from 62 to 86, and the median number of families from 28 to 35. Ruderal and short-living taxa inhabiting rather dry gravel bars (e.g. Melilotus spp., Atriplex spp., Barbarea spp.) and taxa occurring on moist loam and sandy habitats (e.g. Chenopodium spp., Bidens spp., Rorippa spp.) benefited most from floodplain restorations.

For the ground beetles, species richness and the proportion of riparian species were significantly different between non-restored and restored sections (P < 0·05; Fig. 2). Altogether, 48 ground beetle species were recorded (2–17 species per section; mean = 10 species). In restored sections, median number of species was 12 (5 in non-restored sections), and medium number of genera was 7 (4 in non-restored sections); but differences were not significant. The proportion of riparian species was 75·2% in the restored sections and 29·5% in the non-restored sections (P < 0·01). Similar numbers of species were captured in the pitfall traps (median = 5 for both, non-restored and restored sections). Much higher species numbers were recorded by hand-sampling on the open gravel and sand bars (median = 0 in the non-restored sections; median = 10 in the restored sections). Shannon–Wiener diversity of ground beetles did not significantly differ.

Altogether, 169 benthic invertebrate taxa were recorded (97–111 taxa per section; mean = 93 taxa). Values of metrics were always slightly higher in restored sections, e.g. median taxa number was 91 in the non-restored and 96 in the restored section (Fig. 2). Shannon–Wiener Diversity showed a trend towards an increase in the restored sections (median = 2·9 in non-restored sections; median = 3·2 in restored sections). However, none of the calculated metrics was significantly different between non-restored and restored sections. There was also no difference in ‘ecological quality’ between restored and non-restored sections measured with benthic invertebrates according to the German standard procedure (Meier et al. 2006).

comparison of quantitative habitat data and taxa lists

Cluster analyses of habitat data and taxa lists support the results of the metric-based analyses (Fig. 3). The mesohabitat data analysis reveals two main clusters composed of the non-restored sections and restored sections, respectively. In contrast, the analysis of microhabitat data leads to a mixture of non-restored and restored sections.

Figure 3.

Figure 3.

Comparison of quantitative data of non-restored (n) and restored (r) sections for: mesohabitat widths standardized to mean width of the respective section (%), microhabitat composition (%), and taxa abundances of each organism group (Soerensen/Bray–Curtis similarity, flexible beta = −0·25).

Figure 3.

Figure 3.

Comparison of quantitative data of non-restored (n) and restored (r) sections for: mesohabitat widths standardized to mean width of the respective section (%), microhabitat composition (%), and taxa abundances of each organism group (Soerensen/Bray–Curtis similarity, flexible beta = −0·25).

The channel form (non-restored or restored) mainly determines the composition of the floodplain vegetation: all restored sections except the Eder form a cluster, which is opposed to a cluster formed by all non-restored sections. In contrast, for benthic invertebrates, each paired site is grouped together, indicating that site is the main descriptor of invertebrate community composition, while channel form is less important. Catchment size at the sites determines the composition of the main cluster branches, with the larger rivers Lahn and Eder forming one group as opposed to the smaller rivers Orke, Nims and Bröl. The clustering result of the ground beetle data is intermediate with four of the non-restored sections forming one cluster, while the ground beetle fauna of the three other non-restored sections (Eder, Lahn-C, Orke) is more similar to the restored sections. Thus, the composition of the ground beetle assemblage reflects both: sampling site and channel form, that is, if the section has been restored or not.


Restoration increases habitat diversity and availability of biota in floodplain mesohabitats, while the effects on aquatic microhabitats and assemblages were less obvious. The analyses partly reject both hypotheses stated in the introduction. The first hypothesis (restoration increases habitat diversity and availability) is supported for the mesohabitat level, but aquatic microhabitat diversity is not increased although many aquatic microhabitats extended their area (Jähnig et al. 2008). Despite restored sections being closer to reference conditions than non-restored sections, near-natural habitat range and extension have not yet been achieved. In the restored sections about 2–5 large logs per 100 m stream section were found (Jähnig et al. 2009), which is much lower then the values typical for near-natural streams in lower mountainous areas (20 logs per 100 m; Kail 2005); Hering et al. (2004) report 10–70 logs per 100 m in different stream sections for the alpine river Isar (Germany). The availability of bare sediment, crucial for habitat turnover, is still lower at the restored sections: at the near-natural alpine river Tagliamento (Italy), 65% of the floodplain is ‘exposed area’ and 5% are vegetated islands (Van der Nat et al. 2003), compared to about 12% exposed and 25% islands at the restored sections in our study. Rivers such as Tagliamento, Isar, or Upper Rhone (France) reflect habitat diversity in near-natural floodplains, which cannot be achieved by restoring 200 m of the channel (Cellot et al. 1994; Hering et al. 2004; Tockner et al. 2006).

The second hypothesis (increased habitat diversity enhances species richness and diversity of organism groups) is also partly supported: the species richness, but not the species diversity, of floodplain vegetation and ground beetles increased following increased habitat diversity. Diversity differences were not significant due to the large range of values observed.

Floodplain vegetation changes observed in this study, such as an increase in pioneer vegetation, were also recorded by Rohde, Kienast & Bürgi (2004), who evaluated spatial patterns of riparian habitat types following river widenings. For ground beetles and other riparian arthropods, the presence of bars and islands with sparse vegetation are important for increased species richness (Boscaini, Franceschini & Maiolini 2000; Paetzold et al. 2008), due to rapid re-colonization ability (Günther & Assmann 2005). Marginal or no effects on benthic invertebrates as observed in this study support the findings of several other investigations (Brooks et al. 2002; Pretty et al. 2003; Lepori et al. 2005). The poor restoration effects on benthic invertebrates in the study sites are possibly due to the scale of the restoration measures (Bond & Lake 2003), and a lack of recovery potential if restored sections are isolated within longer sections of degraded rivers (Pretty et al. 2003).

The observed restoration effects on different organism groups reflect changes in meso- and microhabitat composition. Restoration generated several additional mesohabitats, and all mesohabitats increased in size. As a consequence, the number of vegetation units increased and eventually the number of plant taxa. Partly, the increased taxa richness is an area effect since the transect length between the embankments increased in the restored sections. For ground beetles, the average species number in the vegetated habitats was not different between non-restored and restored sections. On the open gravel bars in the restored sections, however, species richness greatly increased. Enhanced ground beetle species richness in restored sections is thus mainly a function of gravel bar availability. For benthic invertebrates, overall taxa richness was not affected, although most microhabitats increased in size and frequency thus theoretically increasing niches for population establishment. However, additional habitats were not created and high quality habitats (e.g. large woody debris), required by several sensitive taxa, were not generated in sufficient density (Jähnig et al. 2009).

In addition, different dispersal abilities of the organism groups might have influenced the effect of the restoration measures. Benthic invertebrates comprise taxa with very different dispersal distances (e.g. flightless hololimnic organisms such as mussels and Crustacea), and effective dispersers such as merolimnic caddisflies (Trichoptera) (Hoffsten 2004). Most riparian ground beetles are r-strategists with high dispersal capacities (Bonn, Hagen & Wohlgemuth-von Reiche 2002) capable of finding and colonizing tiny habitat patches. In contrast, many ground beetle species living in vegetated habitats are flightless, but habitats for these species were present prior to the restoration. Many riparian plants are dispersed by air and by water (Soons 2006). Habitats generated by restoration measures might be colonized more rapidly by floodplain vegetation and by riparian ground beetles, than by most aquatic invertebrates, as we found in this study.

In a similar way, source populations for re-colonization are likely to differ between organism groups. Floodplain plants may benefit from seed banks which are mobilized if sediment is relocated (Hölzel & Otte 2001). Many benthic invertebrates, particularly of medium-sized and large rivers, became extinct in the majority of Central European catchments due to water pollution (Zwick 1992). Source population of sensitive taxa are thus not present in most catchments. Ground beetles were affected to a lesser extent than benthic invertebrates by water pollution, since they are less dependent on water quality. They reflect the linkages between aquatic and terrestrial habitats, as they mainly feed on aquatic organisms emerging along the shoreline or washed ashore (Paetzold et al. 2008). Even in degraded sections, suitable habitats for sensitive species might occasionally occur (e.g. small gravel bars), thus securing relict populations. If new habitats are generated, these populations may spread rapidly. An example is the ground beetle Bembidion ascendens Daniel 1902, which is very rare in the Central European lower mountain ranges (Trautner, Müller-Motzfeld & Bräunicke 1997), but was found in abundance on gravel bars in the restored Eder section.

The results of this study are of relevance to water managers in two ways. First, river restoration effects on different organism groups reveal spatial and temporal differences in the response of organism groups. The design of monitoring programmes should reflect this adequately. Organism groups occurring on the floodplain (vegetation and ground beetles, specifically riparian species) are the most suitable indicators of the effects of restoration measures at a section-scale. Even restoration of comparatively short sections in degraded catchments may significantly enhance species richness and the proportion of sensitive species. Benthic invertebrates, however, might better reflect long-term and catchment-scale restoration success, as they seem to be strongly dependent on catchment-scale processes (e.g. eutrophication or input of fine sediments; Wood & Armitage 1997; Hering et al. 2006). Recovery is less likely to be achieved if catchment-wide measures are not implemented. But benthic invertebrates are also sensitive to instream habitat, and if the habitat does not change, the community structure is unlikely to change. To reflect temporal differences in the response of organism groups to restoration, a temporally stratified sampling design including pre-restoration samples is required instead of the space-for-time substitution approach applied in this study.

For the purpose of the Water Framework Directive, river restoration success is mainly evaluated with reference to aquatic organisms, such as benthic invertebrates. Yet, there is considerable doubt that the aim of a ‘good ecological status’ is attainable by 2015 in many European catchments if measures do not significantly increase in size, frequency and comprehensiveness (Bjerring et al. 2008). Riparian organism groups, such as floodplain vegetation and ground beetles, are not considered by the Water Framework Directive, although according to our results they respond more rapidly to restoration than benthic invertebrates. The time-lags and hysteresis-effects of the benthic invertebrate fauna following restoration do not allow for immediate success of catchment management measures. Riparian organism groups may therefore be used as early-responding additional parameters to evaluate the short-term effects of restoration, while aquatic organisms may be better suited to reflect long-term effects. In addition, positive results of restoration are more rapidly achieved according to the aims of other directives: enhanced retention areas or land use changes might support the aims of the EU Floods Directive, while certain vegetation types are of importance to the Habitats Directive. Awareness of the positive effect of river restoration in the context of the legal framework is crucial to increase public support, a preposition for longer-lasting, larger-scale measures.


This study was supported by the EU-funded Integrated Project Euro-Limpacs (GOCE-CT-2003-505540). Three anonymous reviewers are thanked for their comments and suggestions, which improved the manuscript.