The effects of weirs on structural stream habitat and biological communities

Authors


Corresponding author. E-mail: geist@wzw.tum.de

Summary

1. Most of the world’s rivers are affected by dams and weirs. Information on the quantitative and qualitative effects of weirs across biological communities is crucial for successful management and restoration of stream ecosystems. Yet, there is a lack of comprehensive studies that have analysed the serial discontinuity in direct proximity of weirs including diverse taxonomic groups from algae to fish.

2. This study compared the abiotic stream habitat characteristics upstream and downstream of weirs as well as their effects on the community structure of periphyton, aquatic macrophytes, macroinvertebrates and fish at five different study rivers.

3. Physicochemical habitat characteristics discriminated strongly between upstream and downstream sides of weirs in terms of water depth, current speed, substratum composition and the transition between free-flowing water and interstitial zone. Accordingly, abundance, diversity, community structure and functional ecological traits of all major taxonomic groups were indicative of serial discontinuity, but the discriminative power of individual taxonomic groups strongly differed between rivers.

4. The simultaneous inclusion of abiotic habitat variables, taxonomic diversity and biological traits in multivariate non-metric multidimensional scaling was most comprehensive and powerful for the quantification of weir effects. In some cases, the intrastream discrimination induced by weirs exceeded the variation between geographically distant rivers of different geological origin and drainage systems. Community effects were generally detectable on high levels of taxonomic resolution such as family or order level.

5.Synthesis and applications. River sections in spatial proximity to weirs are affected seriously and should be included in the ecological assessments of the European Water Framework Directive. Multivariate models that include several taxonomic groups and physicochemical habitat variables provide a universally applicable tool for the ecological assessment of impacts on serial discontinuity and other stressors on stream ecosystem health.

Introduction

The introduction of weirs into rivers is considered a major threat to aquatic biodiversity (Bunn & Arthington 2002). Alterations of hydraulic components can change the availability of habitat space, habitat quality and the structure of aquatic communities (Brunke, Hoffman & Pusch 2001; Almeida et al. 2009). The ‘serial discontinuity concept’ (Ward & Stanford 1983) describes the effects of physical barriers such as weirs and dams on biotic and abiotic components of lotic systems in a hypothetical framework, but experimental studies on the ecological effects of weirs have mostly focused on single rivers and single taxonomic groups. Habit, Belk & Parra (2007) and Santos et al. (2006) could detect changes in the fish community at upstream sides of hydropower plants, Zhou et al. (2008) showed effects of a small dam on riverine zooplankton composition, and Bredenhand & Samways (2009) recorded a serious decline in macroinvertebrate diversity below a dam in a small river. For a comprehensive assessment of the weir-induced serial discontinuity, it is essential to compare upstream and downstream sides of weirs in their abiotic and biotic habitat characteristics including all major taxonomic groups. This is important as there is recent evidence that cross-taxon congruence in diversity and community composition of aquatic organisms is typically low (Heino 2010). Consequently, comprehensive studies that assess the effects of human impacts on stream ecosystem health (i.e. on aquatic habitat quality and multiple biotic assemblages) are urgently needed.

The main objective of the study presented here was to analyse how abiotic stream habitat characteristics and biotic community effects in the taxonomic groups of periphyton, macrophytes, macroinvertebrates and fishes differ between upstream and downstream sides of weirs, located within carbonate and silicate streams in the three major drainage systems Elbe, Main/Rhine and Danube. Specifically, we hypothesize that upstream and downstream sides of weirs within one river differ in abiotic habitat characteristics, biodiversity and community composition and test whether different taxonomic groups (of different trophic levels) differ in their response to weirs. Furthermore, we hypothesize that multivariate methods that include abiotic (water depth, current speed, substratum composition, water chemistry) and biotic effects on different taxonomic levels (community composition, abundance, biomass, functional groups) are more suitable for the quantification of weir effects compared with the univariate consideration of single taxonomic groups.

Materials and methods

Study Area

The study was carried out between May and July 2009 at five different study rivers distributed throughout major geological units in Bavaria, Germany (Fig. 1). All rivers are located in an area of similar climatic conditions and have similar flow regimes (Table 1), which are governed by snow melt–induced peak flows during spring. All rivers were altered by weirs for hydroelectric power production, which form barriers for fish migration. In this study, the term weir refers to a style of dam which is routinely overtopped by water. The sections above dams (referred to as upstream sides, U) reveal strongly altered velocity distributions, while downstream sides (D) more resemble the natural flow. None of the study stream sections is affected by hydropeaking regimes. In each river, U and D sides were compared using a standardized sampling design comprising 15 replicates in each side (Fig. 2). The length of the sampled river sections on each side was adjusted to the 15-fold river width at respective weir sides, resulting in study sections of 150–420 m (Fig. 2). For safety reasons, the area in direct proximity of the weirs (twofold stream width distance) was excluded. This study was designed to evaluate the effects of serial discontinuity in direct proximity of weirs, because the underlying effectors would be disguised with increasing distance (Ward & Stanford 1983). We are aware that the effects of weirs can exceed those observed in the study area.

Figure 1.

 Location and map of the study area. The three major drainage systems (Elbe, Main/Rhine, Danube) are shown in different shades of grey. Study rivers are marked with black dots. For study river details, see Table 1.

Table 1.   Characterization of the five study streams: catchment characteristics, geology, discharge, weir construction details, water chemistry (mean values from field sampling dates)
 Günz (G)Leitzach (L)Moosach (M)Sächsische Saale (S)Wiesent (W)
Catchment area (km2)526112175523432
DrainageDanubeDanubeDanubeElbeMain
GeologyMolasseLimestone alpsMoraineBasementChalkstone
Mean annual discharge (m3s−1)8·354·652·645·417·48
Year of construction1945189917th. c.a 19051924
Heigth of weir (m)4·04·21·31·51·8
Average river width (m)2414201420
pH value7·88·07·77·27·9
Specific conductance (μS cm−1)556470762292635
Dissolved oxygen (mg L−1)9·310·99·68·310·0
Temperature (°C)18·59·613·517·814·3
Redox potential (mV)580470520490460
Figure 2.

 Schematic diagram of the sampling design with: I–III = tracks (for the fish sampling), a–e = labelling for sampling points of each track, + = sampling points (for the sampling of physicochemical habitat variables, periphyton, macroinvertebrates and macrophytes), x = average width of the river measured 50 m upstream (U) and downstream (D) of the weir, 15x = length of the sampling area, 2x = area excluded from the assessment for safety reasons.

Physicochemical Habitat Characteristics

Because substratum characteristics exert significant control on the quality of streambed habitat and benthic community structure (e.g. Geist & Auerswald 2007), the composition of the stream substratum was investigated at 15 points in each U and D site (Fig. 2). Substratum was sampled with a steel corer of 8 cm diameter (riverside corer; Eijkelkamp Agrisearch Equipment, Giesbeek, the Netherlands). Grain sizes were fractioned with a wet-sieving tower (Fritsch, Idar-Oberstein, Germany) of decreasing mesh sizes (63, 20, 6·3, 2·0 and 0·85 mm). The fractions retained on each sieve were dried at 100 °C and weighed to the nearest gram. The percentage of each grain fraction was determined, and the geometric mean particle diameter (dg) was calculated according to Sinowski & Auerswald (1999). For a hydraulic characterization, water depth, current speed below surface and 15 cm above ground were measured at each sampling point using a HFA flow-measuring instrument (Höntzsch Instrumente, Waiblingen, Germany). Dissolved oxygen, temperature, specific conductance, redox potential and pH were measured in the hyporheic zone in 10 cm depth and in the free-flowing water. Water extractions from the hyporheic zone and redox potential measurements were carried out as described by Geist & Auerswald (2007). Dissolved oxygen, temperature, specific conductance and pH were measured using handheld Multi-340i equipment (WTW, Weilheim, Germany).

Periphyton

As most of primary production in medium-sized streams is related to the algal biomass associated with epilithal biofilms (Müllner & Schagerl 2003), periphyton can play an important role for the assessment of the functionality of stream ecosystems. At each sampling point, periphyton was scraped off a 1- to 4-cm2 total surface area from all available substratum types (stones and dead wood) using a kitchen knife and a flexible plastic tablet to determine surface area. The sampled periphyton mass was dissolved in 200 mL of water and preserved with 20 mL of acidified Lugol’s iodine solution (80% Lugol’s iodine solution, 10% glacial acetic acid, 10% methanol). The Utermöhl technique (Utermöhl 1931; DIN EN 15204 2006) was applied before cell numbers were counted using an inverted microscope. Periphyton samples were left to settle for at least 24 h, and the sample volume for sedimentation was adjusted to 1–50 mL depending on particle concentration in the sample. Algae were determined according to the studies of Ettl et al. (1978–1999) and Cox (1996).

Macrophytes

At each sampling point, all aquatic macrophytes and macroalgae were collected from a surface area of c. 20 m2 using a garden rake according to the methodology described in the study of Deppe & Lathrop (1993). Sampling was continued until no additional species was found (typically c. 15 min). Species were determined according to the study of Weyer & Schmidt (2007). Macroalgae were determined to genus level according to John, Whitton & Brook (2002). The dominance of macrophyte species was calculated as percentage of sampling points at which the particular species was present.

Macroinvertebrates

Macroinvertebrate samples were collected with a Surber sampler (Surber 1930) at 15 sampling points for each U and D side. The substratum inside the sampling area of 0·096 m2 at each sampling point was vigorously disturbed for 2 min to a depth of 10 cm using a metal fork. Macrozoobenthos was then collected in the net (mesh size 0·25 mm) and preserved in 30% ethanol. Macroinvertebrates were classified according to the study of Nagel et al. (1989) using a binocular microscope. Classification was performed on species, family (Chironomids, some Trichoptera and Ephemeroptera) or order level (few Diptera).

Fishes

Fish sampling was conducted using a boat-based electrofishing generator (EL 65 II; Grassel, Schoenau, Germany). Each D and U side was subdivided into three separate tracks (I–III in Fig. 2), which were sampled from downstream to upstream direction by the same electrofishing crew. A single anode was used, and stunned fish were collected with a dipnet. Fish of each track were kept in separate plastic tanks with oxygen supply. The total length of all specimens was measured to the nearest 0·5 cm. Fish of 10 cm or more were individually weighed to the nearest gram. For smaller specimens, a representative number of at least 15 fish were weighed to determine the condition factor and to determine the total biomass as described in the study of Pander & Geist (2010). After sampling all three tracks, the fish were released.

Univariate Data Analysis

In order to assess the exchange between the free-flowing water and interstitial zone, the difference in dissolved oxygen concentration, temperature, pH and redox potential was calculated per sampling point. The catch per unit effort (CPUE) of fish (abundance per 100 m3, fish biomass in g 100 m−3), macroinvertebrate abundance, number of periphyton cells per cm2 and species richness of each taxonomic group per sampling point (for all groups except fishes)/track (fishes) and arithmetic means for each U and D side for each river were calculated. Normality of data was tested with the Shapiro–Wilk test, and the homogeneity of variances was tested with the Levene’s test. Because data were not normally distributed, Mann–Whitney U tests were performed for comparisons between pooled U and D sides over all rivers. For multiple comparisons between sides and rivers, Kruskal–Wallis anova and – in case of significance – Mann–Whitney U post hoc tests were carried out. Bonferroni correction was applied for multiple testings. All statistical analyses were performed in the open-source software r (R Development Core Team 2008).

Shannon index (H), maximum diversity (Hmax) and evenness (E) were computed for all taxonomic groups at each U and D side using the r-package vegan (Oksanen et al. 2009). The saprobic index (SI) (DIN 38410-1, 2004), the Index of Fish Regions (FRI, Dußling et al. 2005) and the following ecological traits were determined for pooled data of all sampling points/tracks per side: fishes were assigned to categories of flow current preference, feeding type and structural requirements (Jungwirth et al. 2003). Macroinvertebrates were assigned to functional feeding groups (FFGs) (Merrit & Cummins 1996), locomotion types (Moog 1995) and flow current preference (Schmedtje & Colling 1996). The FFGs ‘filtering collectors’ and ‘gathering collectors’ were grouped as ‘collectors’, and the locomotion types ‘swimming/diving’, ‘borrowing/boring’ and ‘sprawling/walking’ were grouped as ‘active moving’. The percentage of individuals from each functional trait was calculated per study side and additionally compared over pooled U and D sides from all study rivers. Additionally, the percentage of Ephemeroptera, Plecoptera and Trichoptera (EPT%) was calculated for each side. Characteristic indicator species for U and D sides were determined using one-way simper analysis in Primer v. 6 (Plymouth Marine Laboratory, Plymouth, UK). Untransformed species count data of all taxonomic groups, pooled for each U and D side, were used as input data, with Bray–Curtis similarity for the resemblance matrix and a cut-off value for low contributions of 90%.

Multivariate Data Analysis

In order to detect differences in the response of different taxonomic groups, non-metric multidimensional scaling (NMDS) was performed using taxa abundance data of each of the four groups as input variables for the function metaMDS of the r-package vegan (Oksanen et al. 2009). For a comprehensive assessment, NMDS was performed with the input matrix containing physicochemical habitat characteristics and functional traits of each taxonomic group. The resemblance matrix was calculated in Primer v.6 based on Euclidian distances of the sampling sides for the variables FRI, saprobic index, FFGs, EPT%, percentage of active moving taxa, percentage of rheophilic macroinvertebrates, species richness (for all taxa), eveness (for all taxa), CPUE of fish and macroinvertebrate abundance, cell number of periphyton, abundance of macrophytes, fish biomass, water depth and current speed below surface. For homogenizing different measurement scales before calculating the distance matrix, raw data were normalized using the pretreatment normalization function in Primer v.6. For validation of this NMDS method, regular NMDS and detrended correspondence analysis (DCA) based on commonly used taxa abundance data of all taxonomic groups were performed using functions metaMDS and decorana of the r-package vegan (Oksanen et al. 2009). In order to test the discrimination of U and D sides at different levels of taxonomic resolution, NMDS and DCA was compared for all taxa on the species, family and order level. Environmental fitting on all NMDS plots was performed with 1000 permutations. Only environmental variables with significant ( 0·05) correlation with the NMDS were considered as ordination plot vectors. In addition, β-diversity was calculated as species turnover (βt) between U and D in each river for fishes, macroinvertebrates, aquatic macrophytes, periphyton and all taxa with the function betadiver (r-package vegan) using index g (Koleff, Gaston & Lennon 2003).

Results

Physicochemical Habitat Characteristics

Physicochemical habitat characteristics discriminated strongly between upstream (U) and downstream (D) sides of the weirs (Table 2). Water depth was significantly higher (mean depth U = 1·55 m, D = 0·83 m, < 0·01), and current speed below surface and above ground was significantly lower (mean v a U = 0·15 ms−1, D = 0·24 ms−1, mean v b U = 0·28 ms−1, D = 0·36 ms−1, < 0·01, respectively) in U than in D side. Substratum composition differed significantly between U and D sides as measured by geometric mean particle diameter (dg), percentage of fines and the fraction >63 mm. Mean particle size dg in D was nearly twice the value of U (< 0·05). The percentage of fines in D was 9% lower than in U (mean D = 28%; mean U = 37%, = 0·029), and the fraction >63 mm was 7% higher in D compared to U (mean D = 10%, mean U = 3%, < 0·05). The differences in substratum composition were also reflected in the water chemical gradients between free-flowing water and interstitial zone. For instance, differences in the concentrations of dissolved oxygen between free-flowing water and the interstitial zone were 20% higher in U than in D (< 0·05). Similarly, gradients in temperature (0·3 °C higher in U than in D, < 0·05) and in pH (0·1 higher in D than in U, < 0·05) were observed. Differences in redox potential and specific conductance were least discriminative between U and D owing to high standard deviations (Table 2).

Table 2.   Physicochemical habitat characteristics
RiverSideDepth (m)SDv a (m s−1)SDv b (m s−1)SDdg (mm)SDΔO2 (mg L−1)SDΔT (°C)SDΔpHSDΔ Eh (mV)SDΔ sc (μS cm−1)SD
  1. D, downstream; U, upstream; v a, current speed 15 cm above ground; v b, current speed 10 cm below water surface; dg, geometric mean particle diameter; Eh, redox potential; sc, specific conductance; Δ, difference between the free-flowing water and interstitial zone; bold numbers show significant differences between respective U and D sides.

GU2·220·270·110·030·190·062574·61·8−1·00·70·20·295361622
D1·440·230·290·090·420·082474·51·4−0·70·40·30·214793675
LU0·930·210·430·200·730·2845165·62·7−0·80·30·20·25441−3951
D0·310·140·400·470·450·522465·82·0−0·80·40·30·16746−3371
MU1·420·170·110·050·220·05317·12·4−2·71·10·30·212772−1378
D1·040·330·200·140·320·15826·12·4−2·31·30·60·38253−2772
SU1·570·350·020·010·050·02924·41·9−0·80·50·20·219524−172297
D0·670·250·130·050·200·093084·11·7−0·10·30·20·116055−289317
WU1·580·440·070·080·200·111034·82·1−1·10·50·40·313532−3969
D0·690·450·190·140·400·2150192·42·1−1·10·60·40·36647−1625
PooledU1·550·510·150·170·280·27695·32·3−1·31·00·20·211765−49153
D0·830·480·240·240·360·2711124·52·3−1·01·00·30·210572−65188
% significant differences 100 80 100 60 20 20 20  20    20 

Periphyton

A total number of 129 periphyton taxa were identified. Species richness was significantly lower in the river S than in all other study rivers (< 0·01), and cell numbers per cm2 differed significantly between the five study rivers (< 0·01, Fig. 3). Over all study rivers, a consistent trend towards higher species richness and cell numbers in D sides could be observed (Fig. 3), with two additional periphyton species in D compared to U (mean U = 10; mean D = 12, < 0·05), and the number of cells per cm2 being 40% lower in U than in D (mean U = 611 406; mean D = 993 605, < 0·01). Significant differences in cell numbers per cm2 between the U and the D side of single study rivers were found in the rivers G and S, with 17-fold higher cell counts in the D than in the U side of river G (= 0·05) and twofold higher cell counts in the D than in the U side of river S (< 0·01) (Fig. 3). Beta diversity between U and D was very similar between study rivers, ranging from 0·22 to 0·33 (Table 4). Characteristic periphyton taxa according to simper analysis were Chlorophyceae and Cyanophyceae for U sides and Diatoms from the genera Navicula and Gomphonema for D sides.

Figure 3.

 Characterization of periphyton and macroinvertebrate abundance in U and D sides (15 replicates each) of the five study streams: G, L, M, S and W refer to the different study streams, as described in Table 1. Box: 25% quantile, median, 75% quantile; whisker: minimum, maximum values. Square brackets between boxes show significant differences in single comparisons within one study river. Significant differences between study rivers are given as text. Significance levels are indicated as follows: 0·01 <  0·05*, 0·001 <  0·01**,  0·001***.

Table 4.   Beta diversity (βt) for each taxonomic group and for all taxa as measure of similarity between U and D side in each study river
Riverβt
PeriphytonMacroinvertebratesMacrophytesFishesAll taxa
G0·330·770·000·470·53
L0·300·360·600·400·43
M0·320·580·290·580·49
S0·220·640·750·600·52
W0·220·440·230·640·36

Macrophytes

Species richness of aquatic macrophytes was generally low and strongly variable between study rivers. Overall, 18 species of macrophytes from 13 families were found. Numbers of species were almost balanced in U and D (total number of species U: 15; D: 16) with a slightly higher mean species richness in D (species richness U: mean = 5; D: mean = 6, Fig. 4). Species richness, Shannon index and evenness were not significantly different between U and D. Macrophyte dominance was higher in U sides of three rivers in comparison with the corresponding D sides (U–G 67%, D–G 53%; U–L 100%, D–L 93%; U–M 100%, D–M 80%), lower in U–S (20%) than in D–S (40%) and equal in U–W and D–W (100%). The Shannon index of macrophytes was higher in D than in U sides by a factor of 1·4 (mean U = 0·87; mean D = 1·18). Beta diversity values also indicated great variability among rivers, with the greatest differences between U and D found in the river S and the lowest difference in the river G (Table 4). Only one characteristic species for D, Fontinalis antipyretica HEDW., and no characteristic species for U was identified by simper analysis.

Figure 4.

 Comparison of species richness and diversity indices of the investigated taxonomic groups in U and D sides: data are pooled for U and D sides in each study river, resulting in five replicates each box except for Shannon index and evenness of macrophytes (four replicates). Box: 25% quantile, median, 75% quantile; whisker: minimum, maximum values.

Macroinvertebrates

A total of 11 921 specimens from 93 species of macroinvertebrates comprising 51 families were identified. The most common taxa were Diptera (23%), Amphipoda (16%), Ephemeroptera (15%), Plecoptera (10%), Coleoptera (5%) and Trichoptera (4%). The most abundant FFGs were collectors (52%) and shredders (36%), whereas predators (2%), scrapers (2%) and all other groups (8%) were less abundant. Macroinvertebrate abundance was significantly lower in the river G than in all other study rivers (< 0·01, Fig. 3). Differences in abundance between the U and D within one study river were most pronounced for the rivers W, S and G (< 0·01, Fig. 3). Species richness differed significantly between study rivers (< 0·05) except W–L and S–M. Significant differences in species richness between U and the D were found in the rivers S and W. Pronounced differences between U and D (U: 64 species, mean = 18·2, and D: 81 species, mean = 25·2, < 0·01) occurred, whereas Shannon index and evenness were less discriminative (except for river S with a 2·3 times higher Shannon index and a 1·6 times higher evenness in D–S than in U–S, Fig. 4). Beta diversity as a measure of similarity indicated strong differences in community composition between U and D (Table 4). EPT% values were up to fourfold higher in D sides, with differences varying strongly between streams (Table 3). Characteristic macroinvertebrate taxa according to simper analysis were Oligochaeta and Chironomidae for U and the rheophilic taxa Leuctra nigra (Plecoptera) and Rhyacophila spp. (Trichoptera) for D.

Table 3.   Ecological traits of macroinvertebrates and fishes
RiverSideMacroinvertebratesFishes
SIShredders (%)Collectors (%)Rheophilic (%)EPT%Active moving (%)FRIHigh structural reguirments (%)Benthivor (%)Rheophilic (%)
  1. Values refer to the percentage ratio of the number of individuals in relation to all specimens.

  2. EPT, Ephemeroptera, Plecoptera and Trichoptera; FRI, Index of Fish Regions; D, downstream; U, upstream.

GU2·2027333756·392500
D2·084721026186·09392929
LU1·5427627245634·06697171
D1·6340527248724·05777777
MU1·873234226426·563144
D1·85917867964·41795757
SU2·111645019186·96400
D1·91365334196·10111111
WU1·9113704337343·96879896
D1·7713766638394·16809695
PooledU1·9623593423325·89433524
D1·9041494731494·91575459

The observed differences in abundance, species richness and diversity were even more pronounced considering the functional traits of these groups. Concerning the flow current preference, differences in the percentage of rheophilic taxa of up to 64% between respective D and U sides were observed, with a trend towards higher abundance of rheophilic taxa in D compared with U in three of the streams (Table 3). Accordingly, the percentage of active moving taxa was lower in U sides than in D sides for all study rivers (U: 32%, D: 49%). A classification according to FFGs indicated greater abundance of collectors in U sides (U: 59%, D: 49%) and of shredders in D sides (U: 23%, D: 41%). The saprobic index was higher in U sides compared with the respective D sides (except for L, Table 3).

Fishes

Overall, 27 species from nine families and one lamprey species (Lampetra planeri) were sampled, comprising a total of 2508 specimens and a total biomass of 244 kg. Species richness was higher in D than in U over all study rivers (U: 19 species, mean = 7·4, D: 23 species, mean = 9·6, Fig. 4). The CPUE per number of specimens was significantly higher in D than in U (mean U = 4·9 per 100 m3, mean D = 5·8 per 100 m3, < 0·05). The CPUE biomass was three times higher in D than in U (mean U = 270 g per 100 m3, mean D = 851 g per 100 m3, = 0·01). Fish diversity was higher and more even in D than in U (Shannon D: 2·37, evenness D: 0·74, Shannon U: 2·05, evenness U: 0·68, Fig. 4). Species richness was most discriminative between U–S (8) and D–S (13) and between U–G (10) and D–G (14). In contrast to the other study rivers, there were two more species in U–L (5) than in D–L (3). Fish diversity was higher in D–G, D–M, D–S and D–W than in the corresponding U sides and more even in D–W, D–G and D–L than in the corresponding U sides. Beta diversity values ranging from 0·40 to 0·64 indicated pronounced differences in species composition between U and D (Table 4). In addition to the differences in abundance, species richness, diversity and community composition among D and U, the highly different FRI between U (5·89) and D (4·91) sides (stream-specific difference between U and D of 0·02–2·15) indicated pronounced weir effects on fish community structure and the availability of ecological niches for rheophilic specialists (Table 3). The difference in the FRI mostly results from the higher abundance of rheophilic species such as Salmo trutta L., Thymallus thymallus L., Cottus gobio L., Gobio gobio L., Barbatula barbatula L. and Barbus barbus L. in D (59%) than in U (24%). The most characteristic fish species according to simper analysis were Rutilus rutilus L. for U sides and S. trutta, C. gobio, G. gobio and Squalius cephalus L. for D sides. In addition to flow current preference, the fish community composition of U and D also represented different feeding types and structural requirements, with lower abundance of benthivoric and habitat structure-specialized species in U than in D sides in all rivers except the river W (Table 3).

Multivariate Data Analysis

The consideration of single taxonomic groups instead of comprehensive community response analysis in NMDS revealed strong river-specific patterns (Fig. 5, Table S1, Supporting information). For instance, there was a strong separation between U and D in the river M for fishes and macroinvertebrates, but weak to no separation for periphyton and macrophytes, respectively. In contrast, differences in the river S were greatest for periphyton and macrophytes, but less pronounced for macroinvertebrates and fishes.

Figure 5.

 Non-metric multidimensional scaling performed for different taxonomic groups, based on taxa abundance data and Bray–Curtis similarity. Fishes: non-metric stress = 0·06*10−4; Macroinvertebrates: non-metric stress = 0·03; Aquatic macrophytes: non-metric stress = 0·02; Periphyton: non-metric stress = 0·06. Study rivers are displayed with different pictograms, upstream sides (U) with filled symbols and downstream sides (D) with open symbols. Environmental variables and metrics ( 0·05 based on 1000 permutations) are displayed as vectors and can be distinguished by colour according to their relatedness to physicochemical habitat characteristics (grey), fishes (blue), macroinvertebrates (orange), macrophytes (green), periphyton (brown) and all taxa (red) as well as by capital letters according to their relatedness to feeding type (F), locomotion type (L), reproductive strategy (R), productivity (P), diversity (D) and habitat requirements (H). Codes and coefficients of variance (r2) of the environmental variables are shown in Table 4.

The simultaneous inclusion of abiotic habitat variables and biological traits of all taxonomic groups in NMDS resulted in a more comprehensive and universally applicable assessment (Fig. 6). Both normalized distance matrix–based NMDS (including ecological traits and physicochemical variables of U and D as input variables) as well as classical NMDS and DCA (based on taxa abundance data, DCA not shown) revealed similar discrimination patterns of sides and rivers, indicating a strong linkage of ecological traits, community composition and habitat characteristics. For instance, the strongest separation of U and D was in both NMDS input scenarios found in the river M, and the weakest in the river L. Generally, community effects were detectable not only on the species level but also on higher levels of taxonomic resolution such as family and order level (Fig. 7, Table S1, Supporting Information). Remarkably, differences upstream and downstream of weirs at adjacent sides within the same river were often greater than the differences observed among rivers from different geological units and drainage systems. For instance, differences on all levels of taxonomic resolution were greater between adjacent U and D sides at the river M than between the river M and the rivers G, S and W (Figs 6 and 7), which belong to different drainage systems (G, M: Danube, S: Elbe, W: Main/Rhine) and which are geographically separated by more than 200 km (Fig. 1). On the other hand, the differentiation between the rivers L and W (Figs 6 and 7) as well as between G and S was remarkably low in comparison with the other study rivers (Fig. 6), although these rivers are geographically separated by 300 km (W–L) and 400 km (G–S), belong to different drainage systems (G, L: Danube; W: Main/Rhine, S: Elbe) and belong to different geological units (G: molasses, L: limestone alps; W: chalkstone, S: basement).

Figure 6.

 Non-metric multidimensional scaling of the U and D sampling sides based on Euclidean distances resulting from biological traits and physicochemical habitat characteristics. Study rivers are displayed with different pictograms, upstream sides with filled symbols and downstream sides with open symbols, non-metric stress = 0·03.

Figure 7.

 Non-metric multidimensional scaling performed for different levels of taxonomic resolution, based on taxa abundance data and Bray–Curtis similarity: species level = including all specimen that could be identified on species level, non-metric stress: 0·05; Families = including all specimen that could be determined to family level or further, summarized on family level, non-metric stress: 0·06; Orders = including all specimen summarized to order level, non-metric stress: 0·06. Environmental variables and metrics ( 0·05 based on 1000 permutations) are displayed as vectors. For codes of study rivers, sides and environmental variables, see legend Fig. 5 and Table S1 (Supporting information).

The comparison of beta diversity (including all taxa) between U and D of individual rivers and between the rivers showed that the difference between U and D equals more than half of the differences between rivers (beta diversity between rivers: mean = 0·68, beta diversity between sides: mean = 0·35, Table 4). Variables that correlated with the ordination distances between study sides based on taxa distribution mostly refer to habitat preferences and FFGs, as well as to diversity characteristics and physicochemical variables.

Discussion

The pronounced weir effects detected in this study suggest that damming strongly alters community structure, productivity and the diversity of stream ecosystems. These alterations are supposed to originate in an interruption of the natural gradient of physical habitat conditions and the biotic responses from the headwater to the mouth of river systems (Ward & Stanford 1983), as originally described in the river continuum concept by Vannote et al. (1980). To our knowledge, this is the first study that comprehensively assesses the ecological effects of weirs on the serial river discontinuity including physicochemical habitat characteristics as well as community effects on all major taxonomic groups.

The most important finding is the overriding influence of weirs on biological communities compared with other variables including geology or drainage system. This finding was unexpected, because several authors suggest strong relatedness of rivers of the same or similar geological origin (Stendera & Johnson 2006; Kim et al. 2007; Mykrä, Heino & Muotka 2007) or drainage system (Corkum 1989; Richards et al. 1996; Robinson 1998; Schaefer & Kerfoot 2003). Consequently, the different geochemical conditions of the rivers included herein were expected to result in entirely different community structures. Remarkably, small-scale effects of heterogeneity in dg, water depth and current speed introduced into adjacent sites of the same stream by weirs greatly exceeded the large-scale effects of geology and geographical isolation.

Differences Between Taxonomic Groups And Rivers

This study shows that none of the single taxonomic groups (periphyton, macrophytes, macroinvertebrates, fishes) alone is a universally suitable indicator of the overall discrepancy in community structure upstream and downstream of weirs, yet they are widely used as indicators for the ecological status of aquatic ecosystems [e.g. macrophytes for the trophic status (Schneider & Melzer 2003); fishes for the ecological status in context of the European Water Framework Directive (WFD) (Dußling et al. 2005); periphyton for the ecological condition (Stevenson et al. 2008); macroinvertebrates for freshwater monitoring (Menezes, Baird & Soares 2010)]. The low congruence between the responses of different taxonomic groups to weirs is also supported by their individual and distinct responses to environmental gradients in other freshwater ecosystem studies (Heino 2010). For instance, Declerck et al. (2005) showed that different taxonomic groups in shallow lakes react individualistically to environmental gradients, and Heino et al. (2005) revealed similar results for running waters. Based upon similarity values, diversity measures, functional traits and multivariate community composition analyses, none of the four taxonomic groups studied was a more integrative and sensitive indicator of weir effects than others. The signal strength of weir effects on biological communities not only turned out to be dependent on the taxonomic group investigated but also differs strongly between rivers within taxonomic groups. This is mostly due to the stream-specific habitat structure, community composition, diversity and productivity, which have strong influence on the discriminative power of different taxonomic groups (Heino 2010).

Periphyton

Periphyton, which constitutes the basis of aquatic food webs (Vannote et al. 1980; Szabo et al. 2008), strongly depends on physical habitat characteristics (Soininen 2002; Müllner & Schagerl 2003). This is supported by our data where most physicochemical variables revealed significant correlation with periphyton community composition. Along with the different abiotic habitat conditions observed in the study streams, this finding can explain the differing suitability of periphyton as an indicator of weir effects in different rivers. For instance, in the comparatively deep and slow-flowing river Günz, cell counts of periphyton were 17-fold higher in the more shallow and high-current D compared with the U side. In the shallow and fast-flowing Leitzach, periphyton cell counts were on average threefold higher than in the Günz and only differed by a factor of 1·01 between D and U (Table 2).

Macrophytes

In contrast to periphyton, macrophytes only occurred in some streams and only one species discriminated between D and U, which limits their use as a general indicator for weir effects. Only in two of the rivers (W, M), macrophyte diversity and abundance was high enough to detect differences between U and D, while differences in community composition were evident for the river S. However, as aquatic macrophytes can play an important role for habitat structure (Balanson et al. 2005), weir-induced alterations of the macrophyte community could affect the entire ecosystem in streams with high macrophyte dominance. This is evident from the strong correlation of macrophyte diversity measures with community composition of fishes and periphyton and of all taxonomic groups on species level.

Macroinvertebrates

Macroinvertebrate community structure (diversity indices, FFGs, saprobial index, flow current preference, locomotion types) strongly discriminated between U and D sides, which probably results from different flow conditions up- and downstream of weirs. Whereas the effect of the different flow velocities on flow current preference and locomotion types of macroinvertebrates is obvious (i.e. rheophilic and actively moving taxa being most characteristic for high-current D sides), current also affects the availability and ratio of coarse particular organic matter (CPOM) CPOM to fine particular organic matter (FPOM) FPOM, which can explain the differences in FFGs up- and downstream of the weirs. For instance, the higher abundance of the filter-feeding collectors Simulidae and Chironomidae at U sides with lower current may be explained by higher sedimentation rates of FPOM and consequently higher FPOM/CPOM ratios. Accordingly, shredder organisms that are considered highly sensitive to perturbation (Rawer-Jost et al. 2000) were more abundant at D sides. Analogously, these effects on the distribution of FFGs and organic matter seem to be also true for the general texture of the stream substratum, which was much finer in U than in D. Fine-textured substrata typically reduce the availability of voids and consequently the abundance and diversity of benthic organisms in the hyporheic zone (Gayraud & Phillipe 2003; Geist & Auerswald 2007; Rice, Lancaster & Kemp 2010), which can explain the lower abundance and diversity of Plecoptera, Ephemeroptera and Trichoptera at U sides.

Fishes

The observed differences of up to two fish faunal regions (according to the classification by Dußling et al. 2005) between U and D mirror community structures with entirely different ecological requirements and represent habitat conditions typical for rhithron vs. potamon regions within entire stream ecosystems. Weir-associated fish habitat modifications mostly result from changes of water depths, current speed and substratum composition, which, compared with the natural status, are more pronounced in U than in D. In most cases, U sides cannot fully meet the habitat requirements of species with high structural requirements but are tolerated by indifferent species (Kruk 2007). For instance, the rheophilic species S. trutta, C. gobio, G. gobio and S. cephalus occurred at higher densities in D sides, whereas the generalist species R. rutilus occurred in higher densities in U sides. In rivers with occurrence of high numbers of specialised (e.g. rheophilic, lithophilic or benthivoric) fish species (e.g. M), the most pronounced differences in fish community structure between U and D were observed, while differences in rivers with a high number of tolerant species were generally low (e.g. G).

Implications for Management

The continuity of river systems is a hydromorphological quality element in the European WFD, which requires the evaluation of human impacts on water bodies (European Parliament 2000). However, river sections in spatial proximity of weirs are currently excluded from assessments in context of the WFD. As most European rivers today are a mosaic of upstream and downstream sides of weirs that succeed each other in short geographical distances, information on the qualitative and quantitative effects of weirs on these river sections is crucial for representative assessment of their ecological status and for conservation and restoration management. For example, restoration measures that form a variety of shallow overflowed habitats could improve the overall biodiversity in weir-regulated rivers with increased and uniform water depths and reduced current speed (Kemp, Harper & Crosa 1999; Freeman et al. 2001). The normalized NMDS based on physicochemical variables and ecological traits is highly suitable for a comprehensive quantification of weir effects in different rivers on the ecosystem level as well as for the monitoring of restoration measures. Additionally, this method provides the possibility to assign indicator weights to specific target taxa or ecological traits to account for conservation management prioritization.

Because of river-specific differences, the univariate consideration of single abiotic parameters and of single taxonomic groups is not suitable as a generally applicable indicator for the detection of weir effects. Multivariate methods that simultaneously include different taxonomic groups and physicochemical variables produce a more complete and coherent picture of the serial discontinuity and may serve as a comprehensive and universal indicator of ecosystem health.

Community effects and the underlying effectors were generally detectable at high levels of taxonomic resolution such as family and order level. This illustrates that the effects of the interruption of the river continuum caused by weirs are not only restricted to a few sensitive species or taxonomic groups but affect the entire aquatic community structure. Therefore, a classification on the family or even order level may be sufficient for most taxonomic groups. This finding is particularly relevant for the applicability of this methodology in other regions with different community composition. Typically, funding for the ecological monitoring of weir effects and of other impacts on aquatic ecosystems is limited. The results of this study suggest that the inclusion of multiple taxonomic groups at low levels of resolution is advantageous compared with the inclusion of few groups at high levels of taxonomic resolution in ecological monitoring.

Acknowledgements

We thank ‘Fischereiverein Obere Saale Hof’, ‘Fischereiverein Miesbach-Tegernsee’, J. Heinlein and F.J. Schick for their permission to carry out investigations in their fishery rights and for their local support. We acknowledge support to M.M. by the TUM Graduate School and a doctoral scholarship of UniBayern e.V.

Ancillary