Riparian arthropod responses to flow regulation and river channelization


  • Achim Paetzold,

    1. Department of Aquatic Ecology, Swiss Federal Institute of Aquatic Science and Technology, Eawag, 8600 Duebendorf, Switzerland;
    2. Catchment Science Centre, University of Sheffield, Sheffield S3 7HQ, UK;
    Search for more papers by this author
    • Correspondence author: A. Paetzold, Catchment Science Centre, University of Sheffield, North Campus, Broad Lane, Sheffield S3 7HQ, UK. E-mail:

  • Chihiro Yoshimura,

    1. Department of Aquatic Ecology, Swiss Federal Institute of Aquatic Science and Technology, Eawag, 8600 Duebendorf, Switzerland;
    2. University of Yamanashi, Kofu 400-8511, Japan; and
    Search for more papers by this author
  • Klement Tockner

    1. Department of Aquatic Ecology, Swiss Federal Institute of Aquatic Science and Technology, Eawag, 8600 Duebendorf, Switzerland;
    2. Leibniz-Institute of Freshwater Ecology and Inland Fisheries (IGB),12587 Berlin, Germany
    Search for more papers by this author


  • 1Alterations to river flow and morphology widely impact riverine habitats. Little is known about the consequences of such alterations on riparian arthropods, although they contribute substantially to riverine biodiversity and play a critical role in linking aquatic and terrestrial food webs.
  • 2We investigated the interactive effects of flow regulation (hydropeaking) and river channelization on gravel-bar habitat characteristics and riparian arthropods in seven Alpine rivers. Digital elevation models were developed to simulate inundation dynamics of each gravel bar.
  • 3Channelization significantly increased inundation frequency, and hydropeaking increased substrate embeddedness. The total abundance of riparian arthropods was significantly reduced by hydropeaking, whereas arthropod species richness was reduced by both hydropeaking and channelization. Sites that were affected by both hydrological and morphological modifications together were almost devoid of arthropods.
  • 4The sensitivity of riparian arthropods to alterations in flow and morphology differed among taxa. Spider abundance and richness were significantly reduced by channelization only. Ground beetles showed no significant response. Rove beetle abundance and richness were negatively affected by hydropeaking whereas channelization had a negative effect only in rivers with hydropeaking.
  • 5Abundance and richness of all taxa combined, and of spiders independently, were negatively correlated with inundation frequency and substrate embeddedness. Rove beetle abundance and richness were negatively correlated with embeddedness. Spider and rove beetle richness were also correlated with gravel bar area.
  • 6Synthesis and applications. Our results indicate that the richness and abundance of riparian arthropods were predominantly affected by the availability of exposed gravel above the average high-water level and substrate embeddedness. Restoration of riverbank morphology and mitigation of hydropeaking are likely to benefit riparian arthropods. Riparian arthropods, particularly spiders and rove beetles, appear to be sensitive indicators of the ecological effects of hydromorphological alterations in rivers.


Natural rivers and their fringing riparian zones are pivotal centres for biodiversity (Naiman & Décamps 1997; Sabo et al. 2005) and are among the most threatened ecosystems world-wide (Malmqvist & Rundle 2002; Tockner & Stanford 2002). Riparian zones, in particular, have been severely modified by river engineering and alterations of the natural flow regime (Nilsson & Berggren 2000; Nilsson et al. 2005). Elaborate attempts are underway to restore rivers and to balance restoration efforts with continued use of dams for hydropower production (Robinson & Uehlinger 2003; Giller 2005; Palmer et al. 2005). To optimize such efforts, we need in particular to understand the interactive effects of flow regulation and channel modification on riverine ecosystems (Jansson et al. 2005; Revenga et al. 2005). However, most of our current understanding of the ecological consequences of river regulation is based on single impact studies and selected groups of organisms, particularly fish and riparian vegetation (Lillehammer & Saltveit 1984; Jansson et al. 2000; Downes, Entwisle & Reich 2003; Leyer 2005). Little is known about the consequences of river regulations for riparian arthropods but they are assumed to be particularly sensitive to hydrological and morphological river modifications (Reich 1991; Ellis, Crawford & Molles 2001; Manderbach & Hering 2001; Sadler, Bell & Fowles 2004). Riparian arthropods also represent a functionally important component of riverine systems because they have a critical role in linking aquatic and terrestrial food webs (Baxter, Fausch & Saunders 2005; Paetzold, Schubert & Tockner 2005).

To assess the interactive effects of flow regulation and river channelization on riparian arthropods, we conducted a large-scale comparative field study in seven Alpine braided, or formerly braided, rivers. Natural braided rivers are characterized by extensive areas of exposed gravel that are inhabited by a specialized riparian arthropod fauna (Tockner et al. 2006). Today, however, most braided rivers, which were once widespread in temperate mountain-valley areas, are channelized and impacted by flow regulation (Tockner et al. 2006). As a consequence, much of the gravel bar-associated arthropod fauna has conservation status (Eyre, Luff & Phillips 2001; Manderbach & Hering 2001). In the UK, for instance, almost 20% of beetles on exposed riverine sediments are listed as endangered and vulnerable (Sadler, Bell & Fowles 2004).

Hydropeaking, i.e. diel water-level fluctuations caused by hydroelectric power generation at peak demand (Moog 1993), represents the major type of hydrological alteration in mountainous regions (Petts 1984). In Switzerland, for example, 30% of all hydrologically surveyed rivers are impacted by hydropeaking (BUWAL 2003). Because complete restoration of the natural flow regime is often limited in rivers that are exploited for hydropower production, it is important to understand whether restoration of certain elements of the river morphology, such as river widening, can provide an alternative means of mitigating the potential negative effects of hydropeaking. We inferred the potential effects of morphological restorations on riparian arthropods by comparing channelized sites with sites with remaining natural morphology.

Our main hypothesis was that hydropeaking and channelization reduces species richness and density of riparian arthropods. This was based on the assumptions that (i) hydropeaking increases embeddedness of the gravel (i.e. the degree to which interstitial spaces among gravel and coarser sediments is clogged by fine sediments) because the diel flow fluctuations can remobilize fine sediments that can become deposited along channel margins (Sear 1995); and (ii) channelization reduces the gravel-bar area and increases the frequency of gravel-bar inundation. We expected that an increase in inundation frequency results in a decrease in riparian arthropod species richness because natural braided rivers represent harsh environments that are already located at the decreasing limb of the humped-shaped harshness–diversity curve (Tockner et al. 2006). Negative effects of embeddedness on riparian arthropod abundance and richness were expected because many gravel-bar inhabiting arthropods appear to be dependent on the availability of open interstitial habitats (Schatz, Steinberger & Kopf 2003; Tockner et al. 2006). More specifically, we hypothesized that spiders are particularly sensitive to increased flooding frequency because, unlike most riparian beetles, they cannot escape floods by active flight, whereas riparian beetles are more sensitive to increases in embeddedness (Andersen 1985; Schatz, Steinberger & Kopf 2003). Based on our results, we discuss the implications for river restoration and assessment.


study sites

We investigated 12 gravel bars (sites) along seven mid-sized (5–7th stream order) Alpine rivers in Switzerland and Italy (Table 1). All study sites were characterized by a braided, or formerly braided (channelized), channel style (BAFU 1992). Sites were selected in order to study the ecological effects of flow regulation (hydropeaking) and morphological alteration (channelization), separately and combined, using a full factorial design (Table 1). The channelized sites were generally straightened and narrowed river sections, often with steep ripraps on the upper banks and narrow gravel bars in the active river corridor (Fig. 1). The rivers Tagliamento, Sense and Thur exhibited an essentially natural flow regime driven by snow melt and heavy rain events (Uehlinger 2000; Arscott et al. 2002). The rivers Alpenrhein, Vorderrhein, Hinterrhein and Upper Rhône were strongly affected by hydropower operation. Their natural flow regime was controlled by rainfall, snow and glacier melt. Hydroelectric power operation caused major diel flow variations and reduced seasonal and interannual flow extremes as a result of increased winter and decreased summer discharge (Loizeau & Dominik 2000).

Table 1.  Characteristics of the 12 study sites; Fl+, natural flow regime; Hp, flow regime altered by hydropeaking; M+, natural morphology; M–, channelized sites
SiteArea (m2)Catchment area (km2)Mean annual discharge (m3 s−1)Flow regimeMorphologyGravel-bar width (m)Area affected by hydropeaking (%)*
  • *

    Relative proportion of the gravel-bar area that is inundated daily as a result of hydropeaking.

Tagliamento 110 6292580 90Fl+M+ 60
Tagliamento 240 5872580 90Fl+M+170
Sense 16 118 132  9Fl+M+ 45
Sense 21 300 408  9Fl+M− 15
Thur 13 5931678 47Fl+M− 10
Thur 23 6541665 47Fl+M− 30
Alpenrhein 122 9903969156HpM+11025–30
Vorderrhein24 3891235 32HpM+ 9015–35
Hinterrhein13 4871695 42HpM+12015–50
Alpenrhein 229 7624018156HpM− 4025–55
Rhône 13973368104HpM−  940–60
Rhône 21 4913841130HpM− 1445–60
Figure 1.

Typical gravel bar at the channelized sites (Alpenrhein 1); section is straightened and lined with riprap (see opposite bank); gravel bars develop in the active river corridor. The white line indicates the extent of the gravel bar and the dotted line the area sampled for riparian arthropods.

The length of each site was 5–10 times the channel width, corresponding to a pool–riffle sequence (Ferguson 1981). The lateral extent of each site was defined by the extent of exposed gravel and the absence of dense vegetation (Figs 1 and 2). All sites were dominated by exposed gravel and cobble with small patches of sparse vegetation of similar type, i.e. patches of grass and individual bushes of Salix spp. The average altitude was 483 ± 60 m a.s.l. Altitude did not differ significantly among the flow regime/morphology treatment groups (anova F3,8 = 0·47, P= 0·71). Sites along the Tagliamento River were located in an 800-m wide island-braided section with a complex channel network (Tockner et al. 2003) and included a gravel bar with a steep eroding (Tagliamento 1) and a wide shallow bank (Tagliamento 2). Sense 1 included steep and shallow banks. Sense 2 and both sites on the Thur River were shallow gravel bars. Alpenrhein 1 and Alpenrhein 2 included steep and shallow banks in a morphologically semi-natural and a channelized river section (Fig. 1), respectively, and both were affected by daily water level fluctuations of up to 1·5 m. The sites Hinterrhein (Fig. 2) and Vorderrhein contained shallow and steep banks with diel water level fluctuations of c. 0·9 m. Rhône 1 and Rhône 2 were isolated shallow gravel bars with a diel water-level fluctuation of 0·8 m and 1·4 m, respectively.

Figure 2.

Gravel bar in a river reach with a natural morphology (Hinterrhein, impacted by hydropeaking) at low water level (left) and high water level (right) with the respective inundation model based on a digital elevation model. The white line indicates the extent of the gravel bar.

inundation dynamics and substrate mapping

All gravel bars were mapped during low water using a differential global positioning system (GPS) (Trimble Pro XR/XRS; Trimble Navigations Ltd, Sunnyvale, CA, USA) operated in carrier phase mode with a local base station <1 km apart. Data were collected in a 5 × 5-m grid. In addition, important breaks in the slope were surveyed (Brasington, Rumbsy & McVey 2000). We obtained average accuracies (±95% confidence interval) from continuous position measurements of a fixed base station using Pathfinder Pro software (Trimble Navigations Ltd) of ±2 cm for plain and ±5 cm for elevation measurements, similar to accuracies reported in other studies (Brasington, Rumbsy & McVey 2000).

For each gravel bar, a digital elevation model (DEM) was derived from triangular irregular networks based on three-dimensional point data (Fig. 2). DEM were corrected for the slope of water level by subtraction of a DEM derived from the respective water-level measurements. The slope of the water was derived from several point measures of water level along the river's edge. The water level was measured repeatedly at each site. Inundation dynamics of each gravel bar were simulated using a linear regression between measured water levels at the site and at the nearest permanent gauging station (distance 3–35 km). No major tributaries entered the stretches between the sites and their respective gauging station. Sites were revisited at high water to evaluate inundation models (Fig. 2). We included the year before sampling (2000–01) in the analysis because inundation dynamics of the previous year can affect recruitment of ground-dwelling arthropods (Manderbach & Plachter 1997).

Percentage cover of substrate types (silt and sand, gravel, pebbles, cobbles and boulders) and classes of substrate embeddedness (<5%, 5–25%, 25–50%, 50–75% and >75%) were visually estimated at each GPS point (5 × 5-m grid). Point estimates classified by their dominant substrate type (cover >50%) were used to extrapolate substrate surface cover for each site. Point estimates of embeddedness of >50% were classified as embedded to calculate percentage cover of embedded substrate for each gravel bar. All sites were mapped by the same two people. ArcInfo and ArcView GIS (ESRI) were used for spatial analysis of the data.

sampling for terrestrial arthropods

In April–May, June–July and September 2001, ground-dwelling arthropods were quantitatively sampled within quadrats (0·25 m2) randomly positioned within 0–2 m of the water's edge (n = 24 per site and season). Quadrat sampling provides the most reliable estimate of density and richness of ground-dwelling arthropods (Andersen 1995). To justify the focus on the 0–2-m shoreline strip, we took an additional eight samples (per stratum and season) at 2–5 m and 5–30 m (when present) at each site. Analyses demonstrated that average seasonal abundance and species richness of all riparian arthropods were significantly higher along the shoreline than at more distant habitats (anova F3,103 = 24·44, P < 0·001, F3,103 = 8·91, P < 0·001, respectively). Seventy-five per cent of the average abundance of ground beetles and rove beetles and 48% of spiders occurred within the 0–2-m shoreline strip; 74% of all taxa occurred in the 0–2-m strip (see Appendices S1 and S2 in the supplementary material). Arthropod abundance within the narrow shoreline strip also provides the best indicator of potential trophic linkages between aquatic and terrestrial systems as trophic interactions occur predominantly close to the shoreline (Paetzold, Schubert & Tockner 2005).

For arthropod collection, loose stones, gravel and debris were carefully removed to a sediment depth of 20 cm and water was poured on the sampling plots to drive hidden organisms to the surface. Arthropods were stored in 70% ethanol, counted, and the dominant taxonomic groups, including spiders (Araneida), ground beetles (Carabidae) and rove beetles (Staphylinidae), were identified to species. Ants were excluded from the analyses as they are not strictly associated with riparian habitats (Hammond 1998) and their clumped distribution (aggregation around their nests) complicates representative abundance estimates in a randomized sampling approach.

data analysis

To compare responses of environmental factors and riparian arthropods to river regulation, sites were grouped by flow regime (natural flow regime vs. hydropeaking) and morphology (braided vs. channelized). Three sites for each combination of altered morphology and flow regime were investigated (Table 1). Because of the limited availability of comparable sites in different rivers of similar size, we had to select some sites within the same river for individual combinations of morphology and flow regime. We expected a limited rate of exchange of riparian arthropods between sites within rivers as they were >10 km apart. The fact that river affiliation as a random factor explained almost no variation in the response variables (species richness and abundance) suggested that local site conditions were more important in determining species composition than some spatial dependency among sites within a river (Jansson et al. 2000). Furthermore, within treatment groups (flow–morphology combinations) mean species similarities between sites within the same river were not significantly different from similarities between rivers (Jaccard's index = 0·26 and 0·22, respectively; t = 0·46, d.f. = 3, P= 0·68). However, to account for some potential spatial dependency between sites, we applied a conservative approach. In treatment groups in which two sites were located in the same river, we aggregated the data for each response variable into a single value (mean of the two sites).

Differences in environmental variables (gravel-bar area, inundation frequency and duration, embeddedness, relative proportions of gravel-cobble and sand) in response to flow alteration and channelization were tested using two-way factorial anovas. We focused on variables describing inundation dynamics and sediment composition because they are assumed to be primary factors in controlling riparian arthropod abundance and diversity (Uetz et al. 1979; Hammond 1998). Vegetation, as another potentially important variable (Ballinger, MacNally & Lake 2005), was not considered because sites had very low vegetation cover and the vegetation coverage was of similar quantity and type.

Arthropod abundance at each site was expressed as the average of all samples and number of species as the sum of all taxa found at each site over the entire study period (72 quadrat samples per site). We applied individual-based rarefaction of species number to estimate species richness, standardized to the site of lowest arthropod density (Gotelli & Colwell 2001), using ecosim Version 7 (Gotelli & Entsminger 2006). We excluded species with restricted geographical distribution (two rove beetles at the Tagliamento River) to account for regional variability. Because some sites were almost devoid of riparian arthropods, we excluded single counts of species within a site in the rarefaction to avoid the situation that all sites had to be rarefied to only 1–2 individuals. Unless indicated otherwise, rarefied values are reported for species richness. We applied the same anova design as for the environmental variables to test for effects of hydropeaking and channelization on riparian arthropod abundance and species richness. anovas were performed for the entire riparian arthropod assemblage as well as for spiders, ground beetles and rove beetles separately.

Multiple regressions were performed to analyse how much of the variation in arthropod species richness (rarefied) and abundance was explained by the environmental variables. A stepwise backwards procedure was used to determine the variables explaining most of the variation. All tolerance values were >0·3 and condition indices of the variables were <30, indicating that independent variables were not highly correlated (Weisberg 1980).

All data except gravel-bar area and rarefied species richness were square-root transformed to standardize variances and improve normality. The area was log10-transformed because species richness was expected to be linearly correlated with area on a logarithmic scale (Williamson 1981). For multiple comparisons, we adjusted significance levels with Bonferroni corrections. Statistical analyses were performed with systat 10·0 (SPSS, Chicago, IL, USA). Unless indicated otherwise, values presented are mean ± standard error of the mean.


effects on abiotic habitat conditions

Hydropeaking and channelization significantly affected gravel bar inundation frequency and embeddedness. We found no significant effect on inundation duration, gravel-bar area and the relative proportion of gravel-cobble. In channelized reaches, the inundation frequency of entire gravel bars was significantly higher than in braided reaches (Fig. 3 and Table 2). This was particularly pronounced at rivers with a natural flow regime. Hydropeaking significantly increased the embeddedness of the gravel bars. Channelization resulted in a reduced embeddedness in river reaches with a natural flow regime (Fig. 3 and Table 2).

Figure 3.

Combined effects of gravel bar morphology (M+, natural morphology; M–, channelized river section) and flow regime (Fl+, natural flow regime; Hp, hydropeaking) on inundation frequency of entire gravel bars and on the relative area of embedded substrate (embeddedness of gravel/pebble >50%) (see Table 2 for statistics). n = 3 gravel bars for each pair of morphology and flow regime. Values presented are means ± SE.

Table 2.  Results from two-way fixed-effects anovas on the effects of morphology (natural vs. channelized) and flow regime (natural vs. hydropeaking) on inundation frequency and relative area of embedded substrates (embeddedness of gravel/pebble >50%) of gravel-bank sites. Each column summarizes results from a single anova. Degrees of freedom for each effect and the corresponding error term are given in parentheses
EffectInundation frequencySubstrate embeddedness
  • *

    P < 0·05;

  • **

    P < 0·01;

  • ***

    P < 0·001.

Morphology (1,5)46·71***32·97**
Flow regime (1,5) 9·85*40·04***
Morphology × flow regime (1,5) 0·9427·77*

arthropod community composition

A total of 1476 individuals from 87 taxa (spiders 24 taxa, ground beetles 27 taxa, rove beetles 36 taxa) were collected in 864 samples from 12 gravel bars (see Appendix S1 in the supplementary material). The spider Pardosa wagleri Hahn (64 ± 11% of total spider abundance, present at eight sites), the rove beetle Paederidus rubrothoracicus Goeze (36 ± 9% of total rove beetle abundance, present at eight sites) and the ground beetles Nebria picicornis Fabricius (19 ± 7% of total ground beetle abundance, present at 10 sites) and Bembidion fasciolatum/ascendens (15 ± 4%, present at eight sites) dominated the arthropod communities (see Appendix S1 in the supplementary material).

effects on riparian arthropods

Hydropeaking significantly reduced riparian arthropod density, and hydropeaking and channelization had significant negative effects on riparian arthropod species richness (Fig. 4 and Table 3). Mean arthropod density and richness were highest at natural sites (18·9 ± 3·4 individuals m−2 and 7·4 ± 0·7 species, respectively) and lowest at channelized sites that were also affected by hydropeaking (3·6 ± 1·2 individuals m−2 and 2·7 ± 0·8 species, respectively). Impacted sites contained primarily a subset of the species that occurred at natural sites (see Appendix S1 in the supplementary material).

Figure 4.

Combined effects of gravel bar morphology (M+, natural morphology; M–, channelized river section) and flow regime (Fl+, natural flow regime; Hp, hydropeaking) on number of species (top panel), rarefied species richness (middle panel) and abundance (bottom panel) of ground-dwelling terrestrial arthropods (mean ± SE; see Table 3 for statistics). n= 3 gravel bars for each pair of morphology and flow regime.

Table 3.  Results from two-way fixed-effects anovas on the effects of morphology (natural vs. channelized) and flow regime (natural vs. hydropeaking) on rarefied species richness and abundance of riparian arthropods. Each column summarizes results from a single anova. Degrees of freedom for each effect and the corresponding error term are given in parentheses
EffectAll taxaSpidersGround beetlesRove beetles
  • *

    P < 0·05;

  • **

    P < 0·01.

Species richness
Morphology (1,5)14·66*7·05*4·71 5·01
Flow regime (1,5) 7·45*0·850·01 9·59*
Morphology × flow regime (1,5) 0·160·030·5710·55*
Morphology (1,5) 5·587·93*1·98 9·02*
Flow regime (1,5) 6·86*5·350·5860·38**
Morphology × flow regime (1,5) 0·190·280·12 6·57*

Arthropod groups differed in their sensitivity to hydropeaking and channelization (Fig. 4 and Table 3). Spider abundance and richness were negatively affected only by channelization. The most abundant spider Pardosa wagleri occurred at all sites with a natural morphology but only at two out of the six channelized sites. Hydropeaking and channelization had no significant effects on ground beetle abundance and richness. Effects of channelization on rove beetle abundance and richness differed between the two flow regimes. Abundance and richness were highest at channelized sites with a natural flow regime and lowest at channelized sites affected by hydropeaking. The most abundant rove beetle Paederidus rubrothoracicus was absent at all channelized sites that were also affected by hydropeaking.

correlations between abiotic habitat conditions and species richness and abundance

Embeddedness and inundation frequency were the best predictors for riparian arthropod abundance and richness (Table 4). Abundance and species richness of all groups except ground beetles were negatively correlated with the percentage of embedded substrate (Table 4). Abundance and richness of spiders and all taxa combined were also negatively correlated with inundation frequency. Species richness of spiders and rove beetles were correlated with gravel bar area. Ground beetle abundance and richness of all taxa were negatively correlated with inundation duration.

Table 4.  Results of backwards-selected multiple regressions of species richness (rarefied) and abundance of riparian arthropods in relation to environmental characteristics of gravel bars. Standardized partial regression coefficients for each environmental variable and F-values from anovas for selected regressions are given
VariablesAll taxaSpidersGround beetlesRove beetles
  • *

    P < 0·05;

  • **

    P < 0·01;

  • ***

    P < 0·001.

Species richness
Selected regressionR2 = 0·95R2 = 0·82R2 = 0·52R2 = 0·56
F = 49·78***F = 12·41**F = 3·63F = 5·65*
Inundation duration−0·28*
Inundation frequency−0·90***−0·87**
Selected regressionR2 = 0·72R2 = 0·65R2 = 0·43R2 = 0·49
F = 15·32***F = 11·26**F = 9·14*F = 11·70*
Inundation duration−0·69*
Inundation frequency−0·79**−0·97**


Our results have demonstrated that both river channelization and hydropeaking had negative effects on riparian arthropods. The combination of both impacts resulted in a highly impoverished riparian arthropod fauna of low abundance.

effects on abiotic habitat conditions

Small floods result in the total inundation of the gravel bars in channelized reaches, as evidenced by the shallow bars that have developed. Furthermore, a reduction in cross-section area of the active river corridor generally associated with channelization tends to result in a faster rise in water level. This explains the higher inundation frequency of entire gravel bars in channelized sites compared with sites with a natural morphology. Hydropeaking did not significantly alter the inundation dynamics of the entire gravel bars because diel water-level fluctuations associated with hydropeaking affected only the shoreline parts of the gravel bars.

Hydropeaking resulted in an increase in the proportion of embedded sediments of the gravel bars, particularly along the ecologically important channel margins. Daily rapid increases in flow associated with hydropeaking can remobilize fine sediments that become deposited in areas of low shear stress, such as channel margins (Sear 1995). Hydropeaking during higher flow conditions (snow melt) may have led to the accumulations of fines in higher elevations of the gravel bars. Additionally, hydropower operations generally reduce the number of smaller floods that can flush out fines from the gravel bars (Loizeau & Dominik 2000). In rivers with a natural flow regime, frequent flow pulses (sensu Tockner, Malard & Ward 2000) resulted in a low embeddedness of the gravel bars, particularly at channelized sites where smaller floods already flushed a large proportion of the shallow gravel bars.

effects on riparian arthropods

The negative effect of channelization on abundance and richness of riparian arthropods, particularly spiders, was probably the result of associated increases in inundation frequency of the gravel bars. This was indicated by the negative correlations of richness and abundance of all taxa combined and of spiders with inundation frequency. Inundation frequency appears to be an important factor in controlling ground-dwelling riparian spiders (Uetz 1976; Bonn 2000) as they depend on flood refugia above the high-water level (Adis & Junk 2002). In forested lowland flood plains, spiders can escape floods by migrating to adjacent upland habitats or climbing trees (Adis & Junk 2002). The gravel bars, however, were treeless and access to terrestrial habitats higher in the flood plain was often limited during higher water levels because most gravel bars became surrounded by water with rising water level (Fig. 2). Furthermore, in contrast to lowland flood plains that are dominated by opportunistic species from surrounding uplands (Adis & Junk 2002; Ballinger, MacNally & Lake 2005), the gravel bars with a natural morphology were dominated by habitat specialists such as Pardosa wagleri and we know very little about their ability to survive floods in adjacent vegetated habitats higher in the flood plain.

Channelization had no significant effect on gravel-bar area, and only rove beetle richness showed a positive correlation with area. Gravel-bar area was also not a good predictor of species richness of riparian spiders and beetles in gravel-bed rivers in the European Alps and UK, respectively (Manderbach & Framenau 2001; Sadler, Bell & Fowles 2004). By confining our sampling to a narrow shoreline strip, we deliberately reduced potential area-related increases in arthropod richness resulting from the addition of xerothermic or ubiquitous taxa that generally occur in dry, sparsely vegetated higher parts of the gravel bars.

Our results suggest that the negative effects of hydropeaking on riparian arthropods largely result from the increase in embeddedness. This reduces structural complexity and the availability of hollows and open interstitial habitats. Hollows provide important daytime shelter for lycosid spiders (Framenau et al. 1996) and structural complexity is considered an important determinant for the density and diversity of ground-dwelling spiders (Uetz 1979). Open, air-filled interstitial spaces appear to be important as flood refugia for riparian ground-dwelling beetles (Andersen 1985; Dietrich 1996). The particularly strong dependence of rove beetles on open interstitial habitats (Schatz, Steinberger & Kopf 2003) can explain their high abundance and species richness at channelized sites in rivers with a natural flow regime where channelization further reduces substrate embeddedness. The reduced abundance and richness of riparian arthropods in rivers impacted by hydropeaking might also be a result of the frequent disturbances associated with the daily lateral movement of the shoreline. However, frequent sampling of riparian arthropods over a diel cycle indicated that riparian arthropods are highly mobile and rapidly follow the moving shoreline in hydropeaking-impacted rivers (A. Paetzold, unpublished data). The combined effects of hydropeaking and channelization created the most hostile habitat conditions for all riparian arthropods, but for rove beetles in particular.

Our findings suggest that the negative impacts of hydropeaking and channelization on riparian arthropods predominantly result from the associated increases in inundation frequency and loss of interstitial habitats through increased substrate embeddedness. Other factors may also play a role. For instance, reductions in the productivity of aquatic insects might have affected the densities of riparian arthropods because aquatic insects can provide an important food source for many gravel bar-inhabiting arthropods (Hering & Plachter 1997; Paetzold, Bernet & Tockner 2006). Detailed stable isotope studies at the Tagliamento River showed that ground beetles fed exclusively on emerging aquatic insects while diets of spiders consisted of 50% aquatic insects (Paetzold, Schubert & Tockner 2005). Even though data on aquatic insect emergence were not available for all of our study sites, it is likely that hydropeaking resulted in a reduction in densities and biomass of aquatic insects (Petts 1984; Moog 1993; Uhlmann 2001). However, the fact that ground beetle abundance was not affected by hydropeaking suggests that the observed responses in riparian arthropod densities did not result primarily from altered aquatic insect productivity.

The present study focused on factors that are likely to change consistently with hydropeaking and channelization. However, additional factors, such as changes in predation pressure and the wider terrestrial landscape, might have affected the riparian arthropods. Further, our findings are limited to specialist arthropod species that are permanent residents of exposed gravel but non-resident species might also utilize the gravel bars for feeding during the night. However, night-time sampling at two of the gravel bars showed no occurrence of non-resident ground-dwelling arthropods along the shoreline (A. Paetzold, unpublished data).

implications for river management

Riverine gravel bars are threatened habitats and their associated arthropod fauna is of high conservation value (Sadler, Bell & Fowles 2004; Andersen & Hanssen 2005). Our results indicate that the diversity of the gravel bar-inhabiting arthropod fauna can be conserved in its entirety only in reaches with both a natural morphology and a natural flow regime. Full restoration of the natural flow regime, however, conflicts with the continued use of rivers for hydropower production, and therefore alternative river rehabilitation strategies are required that balance the ecological requirements of gravel rivers against the needs of hydropower production (Baron et al. 2002). Our findings suggest that the increase in substrate embeddedness represents the most important impact of hydropeaking for riparian arthropods. Alternative dam operation schemes, including artificial flood releases that flush fines out of the gravel, could therefore provide a potential mitigation strategy (Schmidt et al. 2001; Mürle, Ortlepp & Zahner 2003), although we know little about the efficiency of artificial floods in this regard (but see Robinson, Uehlinger & Monaghan 2003). Reduced embeddedness of the river bottom substrate may also benefit macroinvertebrate and fish production (Osmundson et al. 2002). Consequently, artificial flood management should be carefully applied using an adaptive management approach (Richter et al. 2003; Robinson & Uehlinger 2003). Furthermore, management constraints may limit the possibility of generating desired floods because many existing dams were not designed with appropriate structures to allow flood releases, and flood releases can involve trade-offs with base flow releases (Acreman 2003). Alternatively, hydropeaking might be reduced by a combination of different hard technical and soft operational measures, such as retention reservoirs or slower up- and down-ramping of turbines (Fette et al. 2007).

The positive effects of a natural morphology on the riparian arthropod fauna indicate that morphological river rehabilitation (e.g. channel widening) can benefit riparian arthropods, particularly in rivers that are affected by hydropeaking. The success of morphological rehabilitations, however, is likely to be dependent on the formation of exposed gravel above the average high-water level, as high frequencies of gravel-bar inundations appear to have negative consequences for riparian arthropods. The development of such gravel bars requires a sufficient supply and transport of sediments and enough space for the river to redevelop a more natural river morphology. Sufficient sediment supply can be particularly problematic in dammed rivers where reservoirs can trap large amounts of sediments (Schmidt et al. 2001; Powell 2002; Stanley & Doyle 2003). Furthermore, the long-term success of morphological restorations is dependent on the flow regime and certain levels of disturbance are required to prevent succession of vegetation on the gravel bars (Church 1995; Nilsson & Berggren 2000).

The success of river restorations for riparian arthropods will also depend on the potential for colonization (Andersen & Hanssen 2005; Bates, Sadler & Fowles 2006). Therefore, the spatial configuration of gravel bars along rivers needs to be considered in restoration planning. Upstream gravel bars might be particularly important sources of colonization by flightless arthropods that are primarily dispersed via floating organic matter (Tockner et al. 2006). However, further knowledge on the potential dispersal pathways by riparian arthropods is required to place restoration measurements strategically in a river corridor framework. It is important to use future restorations as ecosystem-level experiments (Jansson et al. 2005) to improve our understanding of the habitat requirements of many riparian arthropods and their ability to colonize new habitats.

We suggest that riparian arthropods should be integrated in future river assessments because they (i) are particularly sensitive to hydrological and morphological alterations; (ii) contribute significantly to overall riverine biodiversity (Hammond 1998); (iii) play an important role in linking aquatic with terrestrial food webs (Paetzold, Schubert & Tockner 2005); and (iv) can provide important prey for other species of conservation interests, such as birds and bats (Hammond 1998). In addition to the assessment of instream biota, riparian arthropods can provide complementary information on the ecological integrity of the riparian zone.


We are grateful to Viviane Uhlman, Helene Baur and Jacqueline Bernet for help in the field. We thank Dr Irene Schatz and Alexander Rief for the identification of rove beetles and spiders and Dr Werner Maggi for confirming ground beetle identifications. The research has been supported by a grant of the Rhône-Thur project (EAWAG).