The effect of instream rehabilitation structures on macroinvertebrates in lowland rivers


and present address: Simon Harrison, Department of Zoology, Ecology and Plant Sciences, University College Cork, Lee Maltings, Prospect Row, Cork, Ireland (tel. +353 21 4904195; e-mail:


  • 1Many lowland rivers in Western Europe have been substantially modified to aid land drainage and support the intensification of agriculture. Although there have been many attempts at rehabilitation, few have been systematically evaluated on ecological criteria.
  • 2Macroinvertebrates were assessed in 13 UK lowland rivers containing instream rehabilitation structures, seven with artificial riffles (intended to mimic natural gravel riffles) and six with flow deflectors (intended to increase flow, depth and substrate heterogeneity within the channel). In each river, invertebrates were compared between stretches of river with and without rehabilitation structures.
  • 3Rehabilitated and reference stretches were subdivided into different benthic and macrophyte habitats. Three macroinvertebrate samples were taken once in July/August 1999 from each habitat across all schemes and rivers. Current velocity, depth and substratum particle size were recorded at the same time from each habitat.
  • 4Artificial riffle benthos had faster current, a coarser substratum and was shallower than reference benthos. Depth and substratum particle size differed little between flow deflector and reference benthos, although velocity downstream of the deflector tip was greater, and velocity in the lee of the deflector lower, than reference benthos. At a habitat scale, the benthos of artificial riffles, but not flow deflectors, had higher abundance, taxon richness and diversity than reference benthos. The impact of artificial riffles was most marked for benthic rheophilic taxa.
  • 5In all rivers, macroinvertebrate diversity was highest in marginal macrophytes and abundance highest in instream macrophytes. Although invertebrate communities were distinct between artificial riffle (but not flow deflector) and reference benthos, these differences were negligible in comparison to those between benthic and macrophyte habitats.
  • 6Neither artificial riffles nor flow deflectors had any significant impact on the taxon richness of the benthos or of the rehabilitated stretch of the river as a whole. Invertebrate diversity of rehabilitated stretches related closely to that of reference stretches, indicating that larger scale factors constrained any impact of rehabilitation.
  • 7Synthesis and applications. Local rehabilitation structures appeared to have minor biological effects in lowland rivers. We suggest that post-project appraisal should be more rigorously applied to rehabilitation schemes, measuring success against more clearly defined goals. We also advocate a greater emphasis on large-scale riparian, floodplain and catchment rehabilitation, rather than small-scale channel rehabilitation. Such a change in approach needs more effective cooperation and collaboration between all catchment users.


Low-gradient rivers flowing through the agricultural and urban landscapes of north-west Europe have long been subjected to intensive management (Purseglove 1988; Moss 1998; Rackham 2000). Probably more than 95% of lowland river channels in south-east England and Denmark have been modified to enhance land drainage, river navigation and flood prevention (Iversen et al. 1993; Brookes 1995). As a result, many have highly simplified and uniform channels, unnaturally steep banks and little dynamic connectivity with their flood plains.

River modification accelerated in the twentieth century, largely associated with the intensification of agriculture, when many rivers were straightened, deepened and widened to facilitate catchment drainage and to prevent local flooding (McCarthy 1985; Brookes 1988). Instream gravel deposits and most instream woody debris were often dredged from such rivers, further reducing their physical heterogeneity (Swales 1989; Brookes 1988). The characteristic longitudinal and lateral sediment deposition pattern of actively meandering channels was then replaced by a more uniform and diffuse deposition of finer material in constrained channels. The physical complexity of natural marginal and riparian habitats was also usually greatly simplified. Water quality changed to reflect a greater input of nutrients and organic material from more-intensively managed catchments (Sweeting 1996; Riis & Sand-Jensen 2001).

The physical changes associated with river engineering are widely reported to reduce diversity, abundance and biomass among fish, invertebrate and macrophytes, particularly those species associated with coarse substrata, shallow water, high velocity and complex riparian or marginal habitats (Swales & O’Hara 1980; McCarthy 1985; Boon 1988; Swales 1989; Hey 1996). Such effects are increasingly seen as unacceptable in Western Europe and North America and attempts to rehabilitate heavily modified rivers are now being made, where they do not conflict with flood or drainage imperatives (Swales 1989; Boon 1988; Gardiner & Cole 1992; Brookes 1996).

River rehabilitation can be either passive, allowing natural hydraulic forces slowly to re-shape rivers, or active, applying specific measures to modify channel form and structure more rapidly (Gordon, McMahon & Finlayson 1992; Hey 1996). Most recent river rehabilitation projects in the UK have involved small-scale active methods, largely adopted from rehabilitation schemes developed for small salmonid streams in North America since the 1920s (Tarzwell 1932; Swales 1989; Brookes 1996). A variety of instream structures have been developed for these physically dynamic, high-gradient, gravel streams to mitigate the effects of channel engineering. Two of the commonest measures include artificial gravel riffles, intended to mimic natural dynamic riffles, and large boulders and lateral flow deflectors placed on the stream bed to increase channel sinuosity and flow heterogeneity (Tarzwell 1932; Wesche 1985; Brookes, Knight & Shields 1996). Such measures have now been widely adopted for lower-gradient, larger streams and rivers in the UK to enhance populations of lithophilous fish and other biota (Swales & O’Hara 1980; Swales 1982; Swales 1983; McCarthy 1985; Spillett, Armstrong & Magrath 1985; Swales 1989; Hey 1990; Brookes 1992; Brookes et al. 1996).

So far, few lowland river restoration projects have incorporated quantitative analysis of the impacts on fish and invertebrates, allowing little assessment of their true ecological impact or value (Friberg et al. 1994; Brookes 1996; Friberg et al. 1998). Where there have been ecological impact assessments, they have usually been conducted within single river systems and at small spatial scales, offering little insight into their general applicability. In view of the growing interest and expenditure on river rehabilitation, there is therefore an urgent need for a general, quantitative assessment of such practices, both in terms of species richness and ecosystem processes (Muotka & Laasonen 2002).

Here, we sought to quantify the biological impact of two common types of instream rehabilitation measure, artificial riffles and flow deflectors, installed over the last 10–15 years in heavily engineered, low-gradient rivers in central and eastern England. We adopted a paired sampling design, replicated across different rivers, comparing a rehabilitated stretch with a nearby, unrehabilitated reference stretch. This appears to be the first truly replicated assessment of such rehabilitation measures for any British river and we asked whether there were any general changes associated with artificial riffles and flow deflectors across a sample of schemes. We deal only with data on macroinvertebrates, having previously described impacts on fish (Pretty et al. 2003).


site selection

River sites similar in depth, width, gradient and substratum were chosen for the study (Table 1). Smaller, low order (< 4 m wide) streams and limestone/chalk rivers were excluded. Each site consisted of a rehabilitated stretch and a non-rehabilitated reference stretch within 100–500 m up- or downstream (the position up- or downstream depending on the location of the nearest suitable reference stretch to the rehabilitated stretch). The two stretches were sufficiently far apart to appear hydraulically independent (i.e., the rehabilitation structures had no hydraulic impact on the reference stretch) but close enough to have similar riparian structure and water quality. Reference stretches were chosen to represent, as nearly as possible, the channel form that existed in the rehabilitated stretches before installation of the rehabilitation measure. Seven sites with artificial riffle schemes and six sites with deflector schemes were selected. Natural riffles were present at various distances upstream in all rivers.

Table 1.  Summary of locations and physical properties of rehabilitation schemes (u = upstream; d = downstream)
RiverMeasure (number per scheme)LocationMean width (m)Mean depth (m)Length (m)Material used for rehabilitation measureDominant substratum of reference stretchLocation of reference stretch
River IvelRiffle (3)N 52:02:12W 0:16:43 80.7 150Large pebbles (32–64 mm)Fine Gravel200 m (u)
Barlings EauRiffle (6)N 53:16:03W 0:22:43 60.7 250Large pebbles (32–64 mm)Medium Gravel100 m (u)
River LymnRiffle (10)N 53:09:21E 0:08:01 50.5 500Small pebbles (16–32 mm)Fine Gravel100 m (u)
River ThameRiffle (3)N 51:47:58W 0:56:26 81.3 500Large pebbles (32–64 mm)Sand250 m (u)
River WithamRiffle (7)N 52:58:32W 0:38:25 90.7 550Small cobbles (64–128 mm) Fine Gravel250 m (d)
River LarkRiffle (8)N 52:18:32E 0:37:25100.91000Large pebbles (32–64 mm)Silt/Sand250 m (d)
Great EauRiffle (4)N 53:19:12E 0:37:25101.6 350Small cobbles (64–128 mm)Silt/Sand150 m (d)
Little Ouse, upstreamDeflector (20)N 52:23:25E 0:52:11 40.4 350BouldersMedium Gravel400 m (d)
Little Ouse, downstreamDeflector (4)N 52:27:27E 0:52:11141.2 100Wooden stakesFine Gravel150 m (d)
River WithamDeflector (3)N 52:57:56W 0:37:33100.5 350Small cobbles (64–128 mm)Fine Gravel100 m (d)
River EvenlodeDeflector (15)N 51:49:50E 0:52:11110.65 500Large stonesMedium Gravel350 m (d)
Great Ouse (New Cut)Deflector (6)N 52:08:05W 0:25:41100.6 400BouldersMedium Gravel100 m (d)
River HizDeflector (12)N 52:00:17W 0:16:31 50.8 400BouldersSilt/Sand100 m (u)

Each rehabilitation scheme contained a single type of structure, either artificial riffles or flow deflectors, each of which was broadly similar in design across rivers. Riffle schemes consisted essentially of piles of coarse gravel or cobbles placed on the stream bed, between 5 m and 20 m in length, depending on the width of the river. The exact design and number of individual riffles per site differed between rivers (Table 1). Flow deflector schemes consisted of either single (from one bank) or paired (from both banks) flow deflectors, generally constructed from large cobbles (128–256 mm diameter) or, in one case, wooden stakes driven vertically into the stream bed. All structures had been installed for a minimum of 3 years, such that any lack of colonization of biota should not have been a limiting factor in comparisons between them.

invertebrate sampling

Macroinvertebrates were sampled from rehabilitated and reference stretches in summer 1999, when discharges were generally low and stable. For sampling, each stretch (rehabilitated and reference) was divided into recognizable habitats that potentially differed from each other in current velocity, depth, substratum and other habitat features (Fig. 1). For riffle schemes, the habitats were: R1 referred to the (non-riffle) benthos immediately upstream of the artificial riffle; R2 and R3 were patches of riffle benthos at the ‘head’ and ‘tail’ of the riffle, respectively; and R4 was the non-riffle benthos immediately downstream of the artificial riffle. Habitats R6 and R7 were any patches of instream macrophyte on and off the riffle, respectively. Habitat R5 consisted of patches of marginal emergent macrophytes that grew along the banks anywhere over the whole stretch. Not all habitats were available at all sites (e.g. patches of instream macrophytes were not invariably found on artificial riffles). For deflector schemes, the habitats were: D1, the benthos upstream of the flow deflector; D2, the benthos immediately downstream of the tip of the flow deflector (the area of maximum velocity associated with the deflector); D3, the benthos in the downstream lee of the deflector (the area of minimum velocity associated with the deflector); D4, the benthos immediately downstream of habitat D2; and D5 and D7, any patches of marginal and instream macrophyte, respectively. In the reference stretches for both riffle and deflector schemes, three habitats were sampled; C1, the mid-channel benthos; C2, marginal macrophytes; and C3, instream macrophytes. In the results section, we use the terms ‘impacted’ and ‘non-impacted’ to distinguish between habitats directly affected by the rehabilitation measures (i.e., R2 & R3 and D2 & D3 for riffles and deflectors, respectively) from unaffected habitats.

Figure 1.

Schematic diagram showing the location of sampling habitats in stretches with rehabilitation structures and reference stretches.

Three, 15-second kick- (benthos) or sweep- (macrophytes) sample units were taken once, during July–August 1999, from each habitat for each scheme. Samples were all taken from habitats around a single structure installed in the stretch, except for smaller rivers, where samples were taken from two structures. Kick-sampling was carried out using a long-handled pond net, with 0·5 mm mesh, wearing a dry suit in deeper rivers. Samples were preserved in the field and invertebrates sorted and identified in the laboratory, normally to species. Some taxa were identified to order (e.g., Hydracarina and Oligochaeta), family (including many dipteran families), subfamily (e.g., Chironomini and Orthocladiinae) or genus (e.g., Simulium, Baetis and Hydroptila).

physical measurements

Depth and current velocity (at 60% of the depth) were measured at five points, taken haphazardly, in each habitat. The dominant substratum particle size for each benthic habitat was quantified visually into the following six classes: class 1, silt/sand (< 0·5 mm); class 2, sand (0·5–1 mm); class 3, coarse sand (1–2 mm); class 4, fine gravel (2–10 mm); class 5, medium gravel (10–20 mm); and class 6, coarse gravel (20–50 mm).

data analysis

Mean taxon richness, Shannon diversity (H′), total invertebrate numerical abundance (termed simply ‘abundance’ hereafter) and the abundances of individual taxa per 15-second kick/sweep sample were calculated from the three replicates taken from each habitat. For comparisons between the habitats for each type of measure (riffles and deflectors), a two-way block anova was performed on log-transformed data, with habitat as main factor and river as blocking factor and individual rivers as replicates (n = 7 for riffle schemes; n = 6 for deflector schemes).

To compare the physical and biological attributes of the benthos (i.e. excluding samples from marginal and instream macrophytes) for both riffle and deflector schemes, a block anova was used to test for differences among benthic habitats only in mean depth, current velocity and substratum particle size and also in mean invertebrate abundance, taxon richness and Shannon diversity. The effect of rehabilitation structures on the main rheophilic macroinvertebrate families encountered in the investigation (Baetidae, Simuliidae, Ephemerellidae, Elmidae, Hydropsychidae and Gammaridae) was tested by a block anova on mean abundance in the directly impacted benthos and non-impacted benthos (i.e., riffle vs. nonriffle benthos and deflector vs. nondeflector benthos, respectively).

For invertebrate community analysis, all invertebrates were grouped into families (or order, for Hydracarina and Oligochaeta). A correspondence analysis (CA) or principal components analysis (PCA) was performed on the pooled invertebrate samples from each habitat, depending on total gradient length of the ordination of the particular data set (CA for greater than 2·5 SD, PCA for less than 2·5 SD) using canoco, version 4 (ter Braak 1987). The effect of different rivers in the ordination was taken into account by using partial ordination analysis (analogous to the blocking factor in the anovas).

The success of artificial riffles and flow deflectors in increasing the biodiversity of rivers at a stretch-wide scale was evaluated by calculating: (i) the numbers of macroinvertebrate taxa found only in the new habitats created by rehabilitation structures (i.e., habitats R2, R3 and D2, D3) and in no other habitat in the river; (ii) the difference in benthic taxon richness between rehabilitated and reference stretches; and (iii) the difference in total taxon richness for all habitats between rehabilitated and reference stretches.

To examine the impact of structures on stretch-scale benthic diversity, a paired t-test on log-transformed data was used to test the difference between the taxon richness of all samples from impacted habitats (R2 and R3 for artificial riffles, D2 and D3 for flow deflectors) and the taxon richness of all samples from non-impacted habitats (R1 and C1 for artificial riffles and D1 and C1 for flow deflectors) with individual rivers as replicates. Although habitats R1 and D1 were not formally in the reference stretches, there was no significant difference for either measure in any biological or physical factor between them and the benthic reference habitat C1 and they thus represent the background, non-impacted benthos. Their inclusion ensures equal sample effort, which could have had a large impact on taxon richness.

To examine the impact on overall invertebrate diversity of a stretch (including benthic and macrophyte habitats), a paired t-test on log-transformed data was used to compare the taxon richness of samples from all four habitats in the rehabilitated stretch (habitats R2, R3, R5 and R7 for artificial riffles and D2, D3, D5 and D7 for flow deflectors) to the taxon richness of all four habitats in the reference stretch (habitats R1, C1, C2 and C3 for artificial riffles and D1, C1, C2, C3 for flow deflectors). As above, although habitats R1 and D1 were not formally in the reference stretches, their benthos was indistinguishable from that in the reference stretches (Fig. 2) and their inclusion ensures equal sample effort.

Figure 2.

Mean (± 1 SE) invertebrate abundance, taxon richness and Shannon diversity (H′) per 15-second kick/sweep sample for each habitat: Riffle, deflector and reference stretches (n = 7 for riffles; n = 6 for deflectors). See Fig. 1 for key to habitats.


comparisons among all habitats

There were highly significant differences in abundance, taxon richness and Shannon diversity among the habitats for both artificial riffle and flow deflector schemes, largely due to differences between benthic and macrophyte habitats (Fig. 2; Table 2). For both types of scheme, total invertebrate abundance was greatest in instream macrophytes and lowest in marginal macrophytes and the non-riffle/deflector benthos (Fig. 2; Table 2). For riffle schemes, invertebrate abundance of habitat R7 was significantly greater than R1, R5 and C2; abundance of habitat R2 was significantly greater than R5 (post hoc Tukey tests). For deflector schemes, abundance of habitat D7 was significantly greater than D3, D5 and C2 and abundance of habitat C3 greater than D3 and D5 (post hoc Tukey tests).

Table 2.  Summary of results of a two-way block anova performed on each invertebrate community parameter with habitat zone as main factor and river as blocking factor
Type of structureFactord.f.AbundanceTaxon richnessDiversity (H′)
F ratioPF ratioPF ratioP
Artificial rifflesRiver6,475.666< 0.001 8.055< 0.001 2.447  0.038
Habitat zone9,474.012  0.001 6.361< 0.001 6.372< 0.001
Flow deflectorsRiver5,354.376  0.00311.362< 0.001 6.217< 0.001
Habitat zone8,354.67  0.001 9.339< 0.00110.874< 0.001

In contrast, taxon richness was highest in marginal macrophytes and lowest in instream macrophytes and the non-riffle/deflector benthos (Fig. 2; Table 2). For riffle schemes, habitat R5 had significantly greater taxon richness than R1, R4, R7, C1 and C3 while habitat C2 had significantly greater taxon richness than R1, R7, C3 (post hoc Tukey tests). The taxon richness of habitat R2 was also greater than R7 and C3 while the taxon richness of habitat R3 was greater than C3 (post hoc Tukey tests).

The pattern of Shannon diversity between habitats was similar to that of taxon richness, with diversity highest in the marginal macrophytes and lowest in instream macrophytes (Fig. 2, Table 2). For riffle schemes, habitats R5 and C2 had higher diversity than R1, R7, C1 and C3 (post hoc Tukey tests). Habitats R2 and R3 also had significantly higher diversity than R7 and C3 (post hoc Tukey tests). For deflector schemes, the instream macrophytes D7 and C3 had significantly lower Shannon diversity than all other habitats (post hoc Tukey tests).

comparisons among benthic habitats only

There were significant overall differences in depth, current velocity and substratum particle size between the various benthic habitats of riffle schemes (Fig. 3; Table 3). Habitats R2 and R3 had significantly shallower and faster-flowing water and a coarser substratum than R1, R4 and C1 (post hoc Tukey tests). These physical differences between benthic habitats were associated with significant differences in invertebrate abundance, taxon richness and Shannon diversity (Fig. 2; Table 3). Habitat R2 had a significantly greater abundance and taxon richness than R1, although other differences between individual benthic habitats were non-significant (post hoc Tukey tests).

Figure 3.

Mean (± 1 SE) depth, velocity and substratum particle size for each benthic habitat. See Fig. 1 for key to habitats.

Table 3.  Summary of results of two-way block anova for benthic zones only, performed on physical factors (depth, substratum particle size and velocity) and biological parameters (abundance, taxon richness and diversity), from artificial riffle and flow deflector schemes, with habitat zone as main factor and river as blocking factor
SchemeFactord.f.DepthVelocitySubstratum particle sizeAbundanceTaxon richnessDiversity (H′)
F ratioPF ratioPF ratioPF ratioPF ratioPF ratioP
Artificial rifflesRiver6,24 5.958   0.001 5.121   0.002 1.835   0.1352.4870.052 7.129< 0.001 2.517   0.049
Habitat zone6,2427.592< 0.00118.751< 0.00112.769< 0.0012.9160.042 2.864   0.045 3.131   0.033
Flow deflectorsRiver5,1824.391< 0.00111.921< 0.001 6.568   0.0016.610.00112.847< 0.00110.56< 0.001
Habitat zone4,18 1.976   0.14116.629< 0.001 5.236   0.0065.6640.004 0.735   0.58 3.463   0.029

For flow deflector schemes, current velocity and substratum particle size, although not depth, were significantly different between the benthic habitats (Fig. 3; Table 3), with current velocity and substratum particle size generally higher in the habitat at the tip of the deflector (D2) and lowest in the lee of the deflector (D3). D3 had significantly lower velocity than other habitats and D2 had significantly higher velocity than the reference benthos, C1 (post hoc Tukey tests). Substratum particle size in D3 was significantly smaller than in D1 and D2 (post hoc Tukey tests). Invertebrate abundance and Shannon diversity, although not taxon richness, were significantly different among benthic habitats (Fig. 2; Table 3). Invertebrate abundance was significantly lower in the habitat D3 than other benthic habitats, but there was no significant difference in Shannon diversity between individual habitats (post hoc Tukey tests).

Of the main rheophilic taxa, Baetidae, Ephemerellidae, Elmidae, Hydropsychidae and Gammaridae were all significantly more abundant on the artificial riffles (mean of habitats R2 and R3), than in non-impacted benthos (mean of R1 and C1) (Fig. 4a; Table 4). Only Simuliidae were significantly more abundant in the deflector benthos (mean of D2 and D3) than non-impacted benthos (mean of D1 and C1) (Fig. 4b; Table 4).

Figure 4.

Mean (± 1 SE) abundance of the six main rheophilic invertebrate taxa in (a) artificial riffle benthos (mean of habitats R2 and R3) and non-impacted benthos (mean of habitats R1 and C1) (note different y-axis scale for Gammaridae), and (b) flow deflector benthos (mean of habitats D2 and D3) and non-impacted benthos (mean of habitats D1 and C1).

Table 4.  Summary of two-way block anovas performed on log-transformed abundances of rheophilic invertebrate taxa with habitat zone (unimpacted benthos vs riffle or deflector benthos) as main factor and river as blocking factor
FamilyFactorDeflector schemesRiffle schemes
d.f.F ratioPd.fF ratioP
Baetidae River1,1121.56   0.0021,13 3.890.061
Habitat5,11 0.021   0.896,13 9.710.021
Simuliidae River1,1125.27   0.0011,13 1.110.451
Habitat5,11 8.19   0.0356,13 0.010.935
EphemerellidaeRiver1,11 56.3< 0.0011,13 9.260.008
Habitat5,11 0.23   0.656,1311.490.015
ElmidaeRiver1,1110.61   0.0111,13 6.650.018
Habitat5,11 0.048   0.8366,1327.30.002
HydropsychidaeRiver1,1123.04   0.0021,13 1.50.317
Habitat5,11 5.83   0.066,1313.730.01
GammaridaeRiver1,11 0.68   0.661,1319.230.001
Habitat5,11 0.26   0.636,1343.520.001

community composition

Partial correspondence analysis of invertebrate communities revealed strong differences between benthic assemblages and instream and marginal macrophytes for both artificial riffle (Fig. 5a) and flow deflector (Fig. 5b) schemes. The first axis of both ordinations largely distinguished between marginal macrophytes and other habitats. The second axis distinguished between instream benthic and instream macrophyte assemblages. There was little separation between impacted and non-impacted benthic communities for either artificial riffle or flow deflector schemes.

Figure 5.

Plot of sample scores from a partial correspondence analysis of invertebrate communities from (a) habitats in the riffle and reference stretch and (b) habitats in the deflector and reference stretch.

When macrophytes were excluded from the analysis, a partial principal components analysis of benthic habitats (R1, R2, R3, R4 and C1) clearly distinguished between the benthos of the artificial riffles (R2 and R3) and the other non-impacted benthos (Fig. 6). The pattern of taxon scores along PCA axis 1 showed a clear gradient of habitat preference, with taxa associated with artificial riffles more typical of high-velocity eroding habitats and those associated with non-impacted benthos more typical of lower-velocity depositing habitat (Fig. 6). There was no similar distinction for benthic habitats (D1, D2, D3, D4 and C1) from flow deflector schemes (Fig. 7).

Figure 6.

Plot of species and sample scores from a partial principal components analysis of invertebrate communities from benthic habitats in the riffle and reference stretches.

Figure 7.

Plot of species and sample scores from a partial principal components analysis of invertebrate communities from benthic habitats in the flow deflector and reference stretches.

the impact of rehabilitation measures macroinvertebrate diversity at a stretch scale

The list of taxa associated with each habitat, for each river, is shown in Appendices S1 and S2 (see Supplementary material). The number of novel taxa ‘added’ to each river by the rehabilitation structures (i.e. those found solely in the new habitats created by rehabilitation structures) was very low, ranging from 3 to 6 taxa for artificial riffles and from 1 to 7 taxa for flow deflector schemes (Table 5). Most of the taxa associated with rehabilitation structures were also found (albeit, for some taxa, at low density) in other habitats, particularly among the marginal macrophytes.

Table 5.  Taxa associated with artificial riffles or flow deflector habitats not found in other habitats elsewhere in the study stretch
Artificial rifflesFlow deflectors
Elmis aeneasOulimnius tuberculatus
Gyrinidae larvaePotamonectes depressus
Erpobdella octoculataValvata piscinalis
Glossiphonia complanataHydropsyche pellucidula
Theromyzon tessulatumHydroptila sp.
Glossiphonia complanata
Barlings EauTricladida
Elmis aeneas
Hydroporus marginatusLittle Ouse
Hydroptila sp.Hydropsyche pellucidula
Ephemerella ignitaLeptocerus tineiformis
PsychodidaeTinodes waeneri
Ancylus fluviatilisHydropsyche angustipennis
Hydropsyche instabilisCaenis luctuosa
Hydropsyche pellucidulaEphemerella ignita
Hydroptila sp.
Glossiphonia complanataGreat Ouse
Leuctra hippopusPolycentropus flavomaculatus
Great EauGlossiphonia complanata
Oulimnius tuberculatusHelobdella stagnalis
LimoniidaeTheromyzon tessulatum
Micronecta poweriSialis lutaria
IvelLittle Ouse
Ceraclea dissimilisTheromyzon tessulatum
Hydropsyche angustipennis
Hydropsyche instabilisEvenlode
Hydropsyche pellucidulaStictotarsus duodecimipustulatus
Cloeon dipterumCeraclea dissimilis
Hydropsyche pellucidula
ThameHabrophlebia fusca
Hydropsyche angustipennisPsychodidae
Hydropsyche pellucidula
Ephemerella ignita
Leuctra geniculata
Elmis aeneas
Oulimnius tuberculatus
Planorbis vortex
Erpobdella octoculata
Leuctra hippopus

Although, for most schemes, the impacted benthos had higher taxon richness than the non-impacted benthos (a mean increase of 19·9% for artificial riffles and 10·8% for flow deflectors), differences were not significant for either artificial riffles (two-tailed paired t-test: P = 0·19) or flow deflectors (two-tailed paired t-test: P = 0·23) (Fig. 8a). There was also no significant difference between the overall taxon richness of the rehabilitated stretch and the reference stretch for artificial riffles (two-tailed paired t-test: P = 0·25) or flow deflectors (two-tailed paired t-test: P = 0·89) (Fig. 8b). There was however, a significant positive correlation (P < 0·05) between the taxon richness of impacted and non-impacted benthos (Fig. 8a) and between the overall taxon richness of the rehabilitated and reference stretches (Fig. 8b).

Figure 8.

Plot of total taxon richness (a) of the directly impacted benthos (total taxon richness of all samples from habitats R2 and R3 for riffle schemes and habitats D2 and D3 for deflector schemes) against total taxon richness of the non-impacted benthos (total taxon richness of all samples from habitats R1 and C1 for artificial riffle schemes and D1 and C1 for flow deflector schemes); and (b) total taxon richness of rehabilitated stretch (total taxon richness of all samples from habitats R2, R3, R5 and R7 for artificial riffles and D2, D3, D5 and D7 for flow deflectors) against total taxon richness of non-rehabilitated stretch (total taxon richness of all samples from habitats R1, C1, C2 and C3 for artificial riffles, and D1, C1, C2 and C3 for flow deflectors). On both plots the dotted line shows equal taxon richness on both axes.


The restoration of engineered channels in Europe has a much shorter history than in North America, but there has been a similar focus on instream habitat improvement (Brookes 1992; Iversen et al. 1993). Artificial rehabilitation structures have now been installed in many lowland rivers in the UK with the objective of enhancing (mainly non-salmonid) fish and macroinvertebrate communities (Driver 1997). Although stream restoration measures have long been the subject of evaluation and appraisal in North America (De Jalón 1995; Gore, Crawford & Addison 1998; White 2002), their success in Europe cannot readily be assessed because biological evaluation is only rarely undertaken (Iversen et al. 1993; Brookes et al. 1996; Harper, Ebrahimnezhad & Cot 1998). As a replicated spatial study on the biological impact of rehabilitation structures on macroinvertebrates in a number of UK lowland rivers, our contribution therefore aims to fill an important gap.

Headwater streams typically have greater abundance and diversity of benthic macroinvertebrates in natural riffles than in pools and other habitats (Rabeni & Minshall 1977; Logan & Brooker 1983; Allan 1995). This is generally attributed to the effect of current velocity and substratum particle size (Cummins & Lauff 1969; Hynes 1970; Allan 1975; Williams & Mundie 1978; Hildrew, Townsend & Henderson 1980; Erman & Erman 1984). The higher velocities and larger, more stable substratum particles of riffles offer more profitable foraging sites for algal grazers and filter feeders (Williams & Moore 1986; Allan 1995), while larger interstitial pore sizes can increase retention of particulate organic food and act as refugia from adverse flow conditions and/or predators (Gee 1982; Culp, Walde & Davies 1983). Much river rehabilitation in the UK has thus aimed to recreate distinct habitats, such as riffles and pools, in the belief that this will restore the physical heterogeneity of the original system and increase biodiversity within the channel. Such ‘functional’ habitats have been proposed to represent the building blocks of river rehabilitation, and, as such, should become the prime focus of river managers (Harper, Smith & Barham 1992; Harper & Everard 1998; Newson et al. 1998). The riffles and flow deflectors installed in the lowland rivers discussed here can be viewed as an attempt to recreate such functional habitats, and might have been expected to have strong ecological effects.

Our results suggest, however, that the impact on macroinvertebrate assemblages was only modest for artificial riffles and barely detectable for flow deflectors, particularly when viewed at a stretch-wide scale, when taking all habitats, including macrophytes, into account. Taxon richness and abundance was greatest in marginal and instream macrophytes, respectively, rather than in the rehabilitated benthos. Riffles and, particularly, flow deflectors added few invertebrate taxa not found in other habitats (3–6 for artificial riffles and 1–7 for flow deflectors) so that the main effect of artificial riffles was to increase the relative abundance of benthic rheophilic taxa already common elsewhere in the channel. Furthermore, several rehabilitation schemes failed to have any positive impact of stretch-wide diversity. Similarly, the effect of the same structures on fish populations in the same rivers was weak (Pretty et al. 2003). What might explain this limited ecological response to rehabilitation in these rivers?

Within a physical context, the structures may have been limited in several ways. First, they perhaps failed to provide habitat that was sufficiently different from non-rehabilitated habitats elsewhere in the river. Contrary to the expectation of the functional habitat approach – that distinct habitats represent ‘building blocks’ which can effectively be added into a system to create greater physical and biological diversity (Harper et al. 1992) – other factors might override the relationship between invertebrate taxa and particular habitats, particularly given the plasticity of response to habitat by many, although not all, lotic macroinvertebrates (Minshall & Minshall 1977; Jenkins, Wade & Pugh 1984; Barmuta 1989; Palmer & O’Keefe 1991; Bournaud, Tachet, Berly & Cellot 1998; Wright & Symes 1999; Harrison 2000). The apparent changes in physical habitat created by artificial riffles and, particularly, flow deflectors may lie within the normal physical tolerances of taxa living in the unrehabilitated channels.

Secondly, riffles may have only a weak functional role for benthic invertebrates in higher-order lowland rivers. In contrast to hydraulically active upland streams, flow refugia in lowland rivers are likely to be less important to benthic invertebrates, while aggregations of particulate organic matter (POM) will be less limiting (Allan 1995). In addition, both functions may be performed by the abundant instream and marginal macrophytes in lowland systems (Jacobsen & Sand-Jensen 1992).

Thirdly, habitat provided by rehabilitation structures may occur only rarely in lowland rivers. Riffles and pools reflect hydraulic forces acting on the stream bed, with their physical characteristics and longitudinal periodicity resulting from complex interactions between large-scale hydraulic forces and the dominant sediment size (Gordon et al. 1992). A strong active riffle-pool bedform is characteristic of high-gradient rivers of high stream power where there is an adequate coarse sediment supply. They would be less common in low-gradient, finer-sediment rivers of low stream power (Brookes & Sear 1996), possibly occurring at sporadic intervals in local high gradient stretches. The role of artificial coarse gravel riffles in replacing lost ‘natural’ habitat in these latter systems would thus be lessened.

Furthermore, the morphological features of unaltered rivers in their natural state are not static. For example, coarse sediment particles move downstream from one riffle to another during periods of high flow, while finer sediment is constantly deposited and scoured at all points along a river under more common hydraulic events (Gordon et al. 1992). It is likely that this sedimentary dynamism, mediated by hydraulic disturbance, is essential in the maintenance of natural benthic faunas (Hildrew & Giller 1994). Indeed, rheophilic invertebrates are more abundant on natural dynamic riffles that actively recruit new gravel (Rice, Greenwood & Joyce 2001). Artificial riffles in low-gradient systems, with very weak movement of larger particles, are unlikely to act like natural riffles, and will be further compromised by the deposition and infilling of fine sediments. This was certainly characteristic of many of the riffle schemes in our study.

Lastly, the rivers in which riffles (and riffle-dwelling taxa) occur more naturally are generally characterized by cool, clean, well-oxygenated water. Such conditions would be less usual in higher order, low-gradient systems, particularly those flowing through urban or intensively agricultural catchments. In several of the rivers we investigated, poor water quality resulting from high nutrient inputs is likely to have determined invertebrate community composition more than the lack of suitable physical habitat.

The physical context of small-scale rehabilitation schemes could mitigate against their ecological effectiveness. However, their ecological context, at a landscape scale, might also be important (Wiens 2002). The functional habitat approach may be too simplistic to predict a specific ecological response to small-scale rehabilitation measures (Harper et al. 1995): simply observing that particular habitats have partially distinct faunas does not imply that those faunas would be recreated by providing habitat in a small-scale restoration scheme. Local sets of species do not always interact with each other or their environment in the same way, frustrating site-specific predictions (Lawton 2000). Further, most macroinvertebrates have complex life cycles in which different life stages use different parts of the aquatic and riparian environment. This includes the use of marginal vegetation by newly hatched aquatic larvae, which then disperse into the mid-channel benthos (Cellot, Bournaud & Tachet 1984; Harrison 2000), while aerial adults sometimes use terrestrial bankside vegetation (Harrison et al. 2000; Harrison & Harris 2002). Simple structural features such as artificial riffles and flow deflectors are unlikely to satisfy all these requirements.

An ubiquitous pattern in large-scale ecology is the relationship between local and regional species richness, in which local species richness is constrained by the regional species pool (Lawton 2000). In the context of river schemes, the taxon richness of the rehabilitated reach was more a function of taxon richness in the wider, unimproved river, than of the physical structure itself. Such a result could argue for the primacy of larger scale regional processes in determining the local community in lotic ecosystems (Heino, Muotka & Paavola 2003). Thus, placing patches of gravel in a landscape where suitable species are scarce is unlikely to have a major, local effect. The metapopulation structure and dispersal of aquatic invertebrates is still poorly understood (Wilcock, Hildrew & Nicholls 2001; Berendonk & Bonsall 2002) but such should be considered in river rehabilitation.

The influences of large-scale factors, such as geology and land use on lotic invertebrate communities is well known (e.g., Hildrew & Giller 1994; Sponseller, Benfield & Valett 2001). This suggests the need for more extensive rehabilitation measures at the whole catchment scale, using the natural dynamics of the river to fashion instream and riparian habitats naturally, rather than artificially to create local patches of habitat (Sear 1994; Muhar, Schmutz & Jungwirth 1995; Brookes & Sear 1996; Poole, Frissell & Ralph 1997; Harper et al. 1999). Whole catchment restoration is often impractical in intensively managed lowlands, as the need for flood control and land drainage usually overrides any attempt to ‘restore’ whole river catchments to their natural state. Often, however, features at smaller scales in the catchment hierarchy, such as reach-scale riparian land use are important (Harrison 2000; Sponseller et al. 2001; Harrison & Harris 2002). Marginal and riparian habitats are rich in terrestrial and semi-aquatic species and have an important impact on the structure, functioning and diversity of river ecosystems (Hynes 1975; Sweeney 1993; Harrison 2000; Harrison & Harris 2002).

The future of lowland river rehabilitation will differ from past practices. Ideally, it should involve minimal engineering and should rely on natural fluvial dynamics as far as practicable. Rivers should be given as much lateral space as possible, reinstating floodplains, side-channels and marginal habitats where feasible. This study, in particular, has demonstrated the importance of marginal, emergent macrophytes for invertebrate diversity in managed rivers. Rehabilitation of riparian habitats, by the simple act of fencing them from cattle, can have significant benefits for invertebrate communities in taxon-rich lowland streams (Harrison & Harris 2002).

The study has also highlighted the need for much more clearly defined goals and expectations in river rehabilitation, both in the targets (such as the kinds of species, processes or habitat that is desired) and in the context within which projects are set. A macrophyte-rich, low gradient, downstream reach of a river, for example, would require a very different set of rehabilitation principles compared to a higher-order, high gradient, more erosive system. Even within ‘lowland’ rivers, the physical nature of streams may differ widely, largely as a function of catchment geology. Rivers of similar discharge and gradient flowing through limestone, sandy, clay and sandstone catchments may be expected to have markedly different physical, chemical and biological properties. Equally, rehabilitation in rivers flowing through intensively managed catchments may be constrained by excessive nutrient inputs, precluding the reintroduction of target species that need good water quality. Cost effective rehabilitation should thus be tailored both to the natural physical, chemical and biological conditions of the river and to the anthropogenic disturbance that determines the actual state of the river.

Supplementary material

The following supplementary materials are available online here.

Appendix S1 List of taxa for artificial riffle schemes.

Appendix S2 List of taxa for flow deflector schemes.