River rehabilitation and fish populations: assessing the benefit of instream structures

Authors


James Pretty (tel. + 44 20 78823718; e-mail j.l.pretty@qmul.ac.uk).

Summary

  • 1River rehabilitation schemes are now widespread in the UK and elsewhere, but there have been few systematic assessments of their ecological effect, particularly on target organisms such as fish. Fish populations were therefore assessed in 13 lowland rivers using point abundance measures and depletion electrofishing. Each river was sampled in two reaches, respectively containing a small-scale rehabilitation scheme (artificial riffles or flow deflectors) and an unrehabilitated control reach. Detailed geomorphological surveys were undertaken for the two study reaches in each river to assess the physical and hydraulic effect of rehabilitation.
  • 2There were large qualitative and quantitative differences among rivers and some had relatively impoverished fish faunas. Overall, total fish abundance, species richness, diversity and equitability were not significantly different between rehabilitated and control reaches. This was true for both the sampling methods used. Bullhead Cottus gobio and stone loach Barbatula barbatula tended to be more abundant in rehabilitated reaches, but this was significant only for artificial riffles. There was a significant between-year difference in fish abundance.
  • 3In general, rehabilitation schemes increased depth and flow heterogeneity, and fish species richness and diversity appeared to respond positively to increased flow velocity in restored reaches. However, there were few significant relationships between the fish fauna and physical variables, indicating that increasing physical (habitat) heterogeneity does not necessarily lead to higher biological diversity. We therefore caution against the use of physical responses to rehabilitation as a surrogate or reliable predictor of ecological response.
  • 4The weak response of fishes to rehabilitation may have been because the schemes were inappropriate in design and scale for low-gradient rivers. Furthermore, fish assemblages may have lacked the potential for recovery because of poor water quality and/or because the schemes were isolated within longer sections of degraded river. More extensive and directed biological monitoring is essential to improve understanding and enable future improvements in the design of schemes and the selection of sites with greater potential for successful rehabilitation.
  • 5Synthesis and applications. From this substantial sample of lowland rivers, there is little evidence of any general benefit to fish of small-scale instream structures in river rehabilitation. From present ecological knowledge it may be that resources would be better devoted to promoting the development of lateral and off-channel habitats within the river corridor. Physical restoration will be most effective when used alongside other strategies to augment fish populations such as water quality management.

Introduction

As with elsewhere in north-western Europe, the British landscape has changed dramatically over the past century, largely through the intensification of agriculture. An important feature of this change has been the modification of lowland rivers for the purposes of flood control, land drainage and navigation (Purseglove 1988). The most intensive period of river modification occurred between 1930 and 1980, when over 8500 km of river were heavily channelized (Brookes, Gregory & Dawson 1983). In addition, most rivers are managed by the routine clearance of in-channel and bankside obstructions that impede the passage of water (Brookes et al. 1983; Swales 1988).

It is widely acknowledged that intensive farming reduces habitat diversity and quality to the detriment of some terrestrial wildlife (see Vickery et al. 2001; Benton et al. 2002; Robinson & Sutherland 2002). Similarly, aquatic communities may be susceptible to changes in habitat diversity and quality. Channelization and river clearance produce structurally simplified and hydraulically efficient river channels, which permit the rapid clearance of water from the floodplain (Brookes 1985; Smith, Harper & Barham 1990; Wilcock & Essery 1991; Hodgson & O’Hara 1994). Such dramatic change to the physical structure of a river has an inevitable effect upon its ecological functioning (McCarthy 1985).

Spatial and temporal heterogeneity across a range of scales are fundamental characteristics of aquatic systems (Frissel et al. 1986; Townsend & Hildrew 1994; Palmer & Poff 1997; Crook et al. 2001; Ward et al. 2001). Fish exploit a wide range of habitats within a river system, and many species exhibit distinct preferences. Different guilds of fishes require a variety of spawning substrata, whilst the survival of fish larvae and juveniles depends crucially upon the availability of nursery habitats (Mann 1996; Copp 1997a; Cowx & Welcomme 1998). Furthermore, fish community structure and diversity, and resilience to disturbance, may be related to habitat complexity (Gorman & Karr 1978; Schiemer et al. 1991; Pearsons, Li & Lamberti 1992). Therefore, the loss of structural complexity and degradation of spawning, nursery and refuge habitats through river modification can have implications for the fish fauna.

Much concern has been expressed over changes in fish assemblages as a result of river channelization in Britain (Mann 1988; Swales 1988). These impacts have occurred alongside alteration in water quality associated with intensified agriculture (Mason 1996). Numerous studies have reported a significant decline in fish abundance as a result of channelization in lowland rivers in Britain (Swales 1982, 1988; Spillet, Armstrong & Magrath 1985; Cowx, Wheatley & Mosely 1986; Punchard, Perrow & Jowitt 2000). Similarly, studies from warmwater rivers in North America (Chapman & Knudsen 1980; Edwards et al. 1984; Portt, Balon & Noakes 1986) and rivers in Northern Europe (Jungwirth, Moog & Muhar 1993; Muotka & Laasonen 2002) report similar findings. Over recent decades, efforts have been made to rehabilitate rivers by reducing pollution (Moss 1988; Axford 1994) and diversifying river habitat (Swales 1989; Hey 1992). In many rivers, natural patterns of sediment transport, erosion and deposition enable rivers to re-form morphological features such as riffles and pools following channel modification (Brookes 1985, 1992; Hey 1992). In many low gradient rivers in Britain, however, natural formation of such features is slow because stream power is insufficient to transport bed sediment. Because natural recovery from channel modification may be limited, active intervention may be required to rehabilitate such river habitats.

River rehabilitation projects are now widespread in Britain and many techniques are employed to enhance the physical and hydraulic environment for fish and other organisms (Holmes 1998). In general, these techniques involve restoring natural river features that have been lost through channelization. Common techniques include the narrowing and re-meandering of channelized reaches, re-profiling banks that are very steep, and creating specific features such as riffles and backwaters (Cowx & Welcomme 1998; RRC 1999). The scale of rehabilitation projects is varied. Large-scale projects (a few to several kilometres of actively rehabilitated river) often employ many different rehabilitation techniques and may cost a large amount of money (e.g. Biggs et al. 1998; Kronvang et al. 1998). Smaller projects, rehabilitating a few to several hundred metres, may use one or two techniques and cost considerably less.

Despite the increasing prevalence of rehabilitation schemes, any effects they have on river biota are poorly understood. For British rivers, published studies following the biological response to rehabilitation have concentrated on invertebrates and plants (Ebrahimnezhad & Harper 1997; Biggs et al. 1998; Harper, Ebrahimnezhad & Cot 1998) although benefits to fish are often inferred. Furthermore, while detailed follow-up studies are most likely on the larger scale projects, the smaller (but more widespread) projects are often overlooked. In terms of fish, there is a wealth of information on the rehabilitation of salmonid dominated rivers in Europe and North America (Brittain et al. 1993; Jungwirth, Muhar & Schmutz 1995; Linløkken 1997; Thorn et al. 1997; Kelly & Bracken 1998; Muotka & Laasonen 2002) and warmwater rivers in North America (Shields, Cooper & Knight 1993; Shields, Knight & Cooper 1998). However, despite using similar rehabilitation techniques, these studies provide limited insight for British lowland rivers because they differ greatly in their physical characteristics and fish assemblages.

This study describes the effect of two common instream rehabilitation techniques (artificial riffles and flow deflectors) on the fish assemblages in a series of lowland rivers in central and eastern Britain. Our aim was to assess the benefit of small-scale rehabilitation schemes for fish assemblages, in terms of their abundance, species richness and diversity. This study represents one of the first truly replicated assessments of such rehabilitation schemes for any British rivers.

Methods

study sites

Typical small/medium sized rivers of lowland Britain were chosen for study. During initial site selection, an effort was made to choose structurally similar rivers and similar rehabilitation structures. All sites were low gradient, channelized and predominantly surrounded by agricultural land with little riparian buffer zone, although some variation in site characteristics and riffle/deflector design was inevitable (Table 1). Physico-chemical data at or close to our study sites (monitored routinely by the UK Environment Agency) indicated that water quality was typical of lowland rivers (Table 2). While many sites were enriched by nutrients, water quality was considered sufficient to support many non-salmonid species, based on water quality criteria outlined by the Environment Agency. Organic pollution indices based on the British ‘Biological Monitoring Working Party’ BMWP scores and ‘average score per taxon’ ASPT (see Mason 1996), calculated from invertebrate samples taken at each study site, confirmed that some sites were enriched (Table 2). As a general rule, scores of over 100 indicate good water quality while scores around 70 or less indicate a loss of taxa sensitive to organic pollution.

Table 1.  Physical characteristics of the 13 study sites and description of the rehabilitation scheme. ‘Number of measures’ refers to the number of individual riffles or deflectors making up the scheme
SiteLocationSchemeWidth (m)Depth (m)Length of scheme (m)Numberof measuresMaterial usedChannel substratum
LatitudeLongitude
R. IvelN 52°02′12″W 0°16′43″Riffle 7–10    0·5–1 150 3Coarse gravelSand/gravel
Barlings EauN 53°16′03″W 0°22′43″Riffle5–7    0·5–1·5 250 6Coarse gravelSand/gravel
R. LymnN 53°09′21″E 0°08′1″Riffle 5< 0·5 50010Medium gravelSand/fine gravel
R. ThameN 51°47′58″W 0°56′26″Riffle 7–10    0·5–2 500 3Coarse gravelSilt/clay
R. WithamN 52°58′32″W 0°38′25″Riffle10    0·1–1·5 550 7CobblesSand/gravel
R. LarkN 52°18′32″E 0°37′25″Riffle10–12    0·5–21000 8Coarse gravelSilt/sand/gravel
Great EauN 53°19′12″E 0°37′25″Riffle10    0·5–2 350 4CobblesSilt
R. WithamN 52°57′56″W 0°37′33″Deflector10    0·5–1·5 350 3CobblesSand/fine gravel
R. EvenlodeN 51°49′50″E 0°52′11″Deflector10–12    1–1·5 50015Large stonesSand/gravel
R. HizN 52°00′17″W 0°16′31″Deflector 6    0·5–1·5 40012BouldersSilt/sand
Great Ouse (New Cut)N 52°08′05″W 0°25′41″Deflector10    0·5–1 400 6BouldersSand/fine gravel
Little Ouse (Knettishall)N 52°23′25″E 0°52′11″Deflector 4–6< 0·5 35020BouldersSand/fine gravel
Little Ouse (St Helens)N 52°27′27″E 0°52′11″Deflector12–15    0·5–2 100 4Wooden stakesSand/fine gravel
Table 2.  Water chemistry from monitoring stations near the study sites. Mean values (and 95% C.I.) are based on monthly Environment Agency samples for 1999. Biotic indices (BMWP and ASPT) were calculated from an invertebrate survey carried out at each study site during summer 1999 (S.S.C. Harrison et al. unpublished data)
SiteSchemeTemperature (°C)pHDO (mg l−1)BOD (mg l−1)Ammonia (mg l−1)TON (mg l−1)P sol. react. (mg l−1)BMWPASPT
R. IvelRiffle13·5    7·9      8·0    1·3    0·1410·9    2·5108    4·0
  (2·75)(0·13)  (1·2)(0·20)(0·05)  (0·87)(0·48)        (9·0)(0·2)
Barlings EauRiffle11·4    8·211·2    1·3    0·1512·8    1·0       90    3·5
  (3·52)(0·10)  (1·23)(0·20)(0·14)  (2·40)(0·68)        (9·3)(0·1)
R. LymnRiffle11·2    8·312·6    1·44    0·05      9·8    0·1      87    4·5
  (2·82)(0·07)  (1·04)(0·31)(0·02)  (0·92)(0·02) (19·6)(0·2)
R. ThameRiffle11·7    8·0      8·3    1·5    0·09      6·6       95    3·9
  (3·36)(0·14)  (1·27)(0·46)(0·06)  (1·58)   (11·9)(0·1)
R. WithamRiffle13·3    8·414·3    1·1    0·0311·8    0·3      70    4·3
  (3·12)(0·14)  (2·00)(0·09)(0·01)  (1·65)(0·08)        (7·4)(0·2)
R. LarkRiffle13·2    8·110·7    1·6    0·20      7·9    0·2      67    4·1
  (3·78)(0·05)  (0·60)(0·32)(0·10)  (1·22)(0·04)          (7·4)(0·2)
Great EauRiffle10·7    8·112·2    1·3    0·0610·6 102    3·8
  (2·67)(0·09)  (1·42)(0·17)(0·03)  (0·84)          (9·4)(0·1)
R. WithamDeflector13·3    8·414·3    1·1    0·0311·8    0·3 56    3·9
  (3·12)(0·11)  (2·00)(0·09)(0·01)  (1·65)(0·08)        (8·8)(0·3)
R. EvenlodeDeflector11·3    8·210·7    1·6    0·05      9·0 124    4·3
  (2·79)(0·07)  (0·88)(0·30)(0·03)  (0·97)   (17·0)(0·2)
R. HizDeflector12·9    8·010·0    1·3    0·14      9·6    1·4      73    3·5
  (2·34)(0·08)  (1·36)(0·21)(0·09)  (0·56)(0·33)        (8·7)(0·2)
Great Ouse (New Cut)Deflector13·1    8·311·1    2·6    0·05      7·6    0·5      84    4·0
  (3·66)(0·18)  (1·37)(1·32)(0·14)  (1·85)(0·12)        (8·5)(0·2)
Little Ouse (Knettishall)Deflector11·2    7·9      9·3    1·1    0·21      7·6    0·05      80    4·2
 2·61(0·04)  (1·67)(0·15)(0·14)  (0·90)(0·01)        (8·2)(0·2)
Little Ouse (St Helens)Deflector12·4    8·1      9·3    1·5    0·17      7·5    0·4      80    4·3
 2·88(0·13)  (0·99)(0·19)(0·08)  (1·50)(0·05)      10·7)(0·2)

A total of 13 rivers was selected for the study, seven with artificial riffles and a further six with flow deflectors. Artificial riffles essentially consisted of piles of coarse mineral substrata placed at intervals along a river reach. Flow deflectors were more variable in their design, but usually consisted of structures protruding from the bank into the river channel (as pairs or in a staggered pattern), which serve to narrow the channel and create areas of faster flow (Hey 1996). All rehabilitation schemes were installed between 1992 and 1997, allowing a reasonable period of time for the fish populations to respond to the altered river conditions.

sampling regime

Two reaches in each river were sampled. One reach contained the rehabilitation scheme (hereafter referred to as manipulated) and the other served as a control reach (100–500 m up or downstream). Control reaches resembled the manipulated reaches except for the rehabilitation measure itself. A control reach had a similar riparian structure and water quality to the manipulated reach, a channel shape typical of the river section as a whole, and was located where there was no hydraulic impact from the rehabilitation scheme.

Fish populations at all sites were assessed using point-abundance electrofishing (Copp & Peňáz 1988; Persat & Copp 1989) by means of a portable (backpack), battery-powered electroshocker (Table 3). Between 20 and 50 haphazardly chosen points were sampled in each manipulated reach, and the same number at the control. The fish caught at each point were identified and counted in the field. Sampling was conducted at all sites in July/August 2000 and repeated in July/August 2001, although five sites could not be visited in 2001 due to access restrictions.

Table 3.  Summary of sampling regime
SiteSchemePoint abundance samplingDepletion sampling (2000)
20002001ManipControl
R. IvelRiffleXXXX
Barlings EauRiffleX   
R. LymnRiffleX   
R. ThameRiffleXXX 
R. WithamRiffleX   
R. LarkRiffleXXXX
Great EauRiffleX   
R. WithamDeflectorX   
R. EvenlodeDeflectorXXX 
R. HizDeflectorXXXX
Great Ouse (New Cut)DeflectorXXXX
Little Ouse (Knettishall)DeflectorXXXX
Little Ouse (St Helens)DeflectorXXXX

Point-abundance electrofishing is commonly used to make quantitative estimates of fish populations. It is effective in catching juveniles and small fish species, but may be less effective at catching larger, more mobile individuals, and is best used in conjunction with other sampling methods (Cowx, Nunn & Harvey 2001). Thus, to complement the point-abundance sampling, a second method (repeated depletion sampling electrofishing) was also used to assess fish populations. Stop nets were used to section off the manipulated and control reaches, which were fished throughout on two or three repeated runs whilst retaining all captured fish in aerated bins until the end of the survey. Depletion sampling was carried out by the Environment Agency in summer 2000. Due to resource constraints, only six rivers were fished in both control and manipulated reaches, and a further two in the manipulated reach only (Table 3).

In addition to the fish survey, a detailed physical and hydraulic survey was carried out in the control and manipulated reaches at each site. This was repeated at two flows, representing baseflow and medium flow conditions. Five cross sections were surveyed in manipulated reaches, encompassing the deepest/shallowest and widest/narrowest sections. A similar survey was conducted in the control reaches, with 3–5 cross sections depending on the morphological variability along the reach. At each cross section, velocity profiles were taken at 1-m intervals across the river using a Sensa RC2 electromagnetic current meter. Flow velocity was measured at 2, 4, 8, 15 and 25 cm above the riverbed and then at 15 cm intervals to the surface. From these data it was possible to estimate mean flow velocity and depth within each study reach, as well as a measure of velocity and depth heterogeneity (expressed as a coefficient of variation CV). These physical parameters were compared with the fish data to examine whether fish assemblages responded to physical changes resulting from rehabilitation.

data analysis

The design of the study enabled a replicated, paired comparison of manipulated reaches (riffles or deflectors) and their controls. Data analysis was carried out separately for the riffles and deflectors. A two-way anova was used with ‘reach’ (manipulated vs. control) and ‘year’ as factors. The fish assemblage was analysed in terms of several response variables: total fish abundance, taxon richness, assemblage diversity, equitability, and the abundance of individual species sufficiently numerous for meaningful comparison. In addition, Spearman's rank correlation was used to test whether the age of the schemes had a significant effect on any of the response variables. This was carried out for the 2000 point sampling data only (because all sites were sampled at this time) and the strength of the response was calculated as the difference between control and manipulated reaches.

The depletion sampling data were not separated into riffle and deflector sites because data from both manipulated and control reaches were obtained for only six sites. Data for riffles and deflectors were combined for rivers and the manipulated and control reaches were compared using a paired t-test (with site as the replicate). Before analysis, fish abundance in each reach was standardized as density (100 m−2) to control for variation in the total area fished at each site. Analysis was performed for the same response variables chosen for the point sampling method. Where data did not conform to assumptions of normality and homoscedasticity they were log10 (n + 1) transformed.

Relationships between fish fauna and physical variables were investigated using Spearman's rank correlation. The point abundance data from 2000 (when all sites were sampled) was compared with mean flow velocity and depth and flow heterogeneity (CV), at both baseflow and medium flow conditions. To examine whether the schemes with the largest physical effect corresponded to those with the greatest biological effect, data were adjusted to represent the percentage change in the manipulated reach relative to its control.

Results

Over the duration of the study a total of 1393 point abundance samples were collected, with a total catch of 4652 individual fish belonging to 18 species (Fig. 1). Overall, the minnow Phoxinus phoxinus was numerically dominant, representing 54% of the total number of fish caught by point sampling. There were considerable qualitative and quantitative differences among rivers, and some (notably the Hiz, Witham and Great Eau) had relatively impoverished fish faunas. The wide variation among sites in fish abundance limited the number of comparisons that could be made. For example, although roach Rutilus rutilus (L.) were the second most abundant species caught in the study, they were only abundant in three of the 13 study rivers. Only two species, stone loach Barbatula barbatula and bullhead Cottus gobio were sufficiently abundant among rivers to warrant separate analysis.

Figure 1.

Summary of point abundance fish surveys at each site (control and manipulated reaches combined). Species listed in descending order of numerical dominance (total numbers caught in brackets). Size of circle represent number of individuals for each species at each site. Key to site names: NEW, New Cut; EVE, Evenlode; LOK, Little Ouse at Knettishall; LOS, Little Ouse at St Helens; WIT, Witham; HIZ, Hiz; BAR, Barlings, Eau; LYM, Lymn; IVE, Ivel; LAR, Lark; GTE, Great Eau.

There appeared to be no significant relationship between the age of the rehabilitation scheme and the strength of the fish response (Spearman's rank correlation: abundance n= 13, rs = 0·11, P= 0·585; species richness n= 13, rs = 0·04, P= 0·724; diversity n= 13, rs = 0·39, P= 0·211).

In reaches where deflectors had been installed, total fish abundance and the numbers of stone loach and bullhead tended to be higher than control reaches (Fig. 2). However, the standard error around the mean was large (reflecting large variation among replicate sites), and differences between manipulated and control reaches were not significant (Table 4). Species richness, diversity and equitability were less variable among sites, and there was no significant difference between manipulated and control reaches (Fig. 2; Table 4). There were significant between-year differences, with total fish abundance being higher in 2000 and assemblage equitability being higher in 2001. There was no significant interaction between year and reach (Table 4).

Figure 2.

Deflector sites: mean point abundance, species richness, diversity and equitability for manipulated (open columns) and control reaches (shaded columns). y axes represent the mean of replicate sites ± 1 SE (note different scales on y axes). Six sites were sampled in 2000 and five in 2001. The 2000 data has been plotted for all sites ‘ 2000 all’ and those revisited in 2001 ‘2000 (5 sites)’.

Table 4.  Deflector sites: summary of the two-way anova (on log10n + 1 transformed data) comparing the effect of reach (control vs. manipulated) and sampling date on fish abundance, species richness, diversity and equitability
 Effect of Year d.f. = 1,18Effect of Reach d.f. = 1,18Effect of Interaction d.f. = 1,18
FPFPFP
Stone loach3·560·0760·900·3550·240·633
Bullhead2·480·1331·210·2860·120·738
Total abundance6·600·0191·020·3260·070·791
Species richness0·940·3460·800·3840·120·738
Diversity (H)0·690·4160·200·6580·0010·975
Equitability (J)6·880·0170·050·8190·0010·980

In reaches where riffles had been installed, only three of seven sites were sampled in 2001, and between-year comparisons should be treated with caution. There was no significant difference between manipulated and control reaches in total fish abundance, species richness, diversity or equitability (Fig. 3; Table 5). Bullhead abundance tended to be higher in manipulated reaches and this pattern was significant in both years (Table 5). Stone loach were significantly more abundant in manipulated reaches in 2000 (paired t-test, d.f. = 6, t= 3·40, P= 0·015) but not in 2001 (paired t-test, d.f. = 2, t= 1·47, P= 0·279).

Figure 3.

Riffle sites: mean point abundance, species richness, diversity and equitability for manipulated (open columns) and control reaches (shaded columns). y axes represent the mean of replicate sites ± 1 SE (note different scales on y axes). Seven sites were sampled in 2000 and three in 2001. The 2000 data has been plotted for all sites ‘2000 all’ and those revisited in 2001 ‘2001 (3 sites)’.

Table 5.  Riffle sites: summary of the two-way anova (on log10n + 1 transformed data) comparing the effect of reach (control vs. manipulated) and sampling date on fish abundance, species richness, diversity and equitability
 Effect of Year d.f. = 1,16Effect of Reach d.f. = 1,16Effect of Interaction d.f. = 1,16
FPFPFP
Stone loach2·940·1061·570·2280·640·223
Bullhead2·010·1753·730·0720·1070·748
Total abundance3·460·0810·340·5660·0010·993
Species richness0·050·8241·640·2190·090·764
Diversity (H)1·750·2052·470·1350·050·823
Equitability (J)2·000·1760·010·9450·030·874

For the six sites assessed using depletion sampling, total fish abundance was low, on average < 20 fish 100 m−2 (Fig. 4). The large standard errors for the abundance of individual species reflect the large variation among sites. There were no significant differences between manipulated reaches and their controls (paired t-tests, d.f. = 5, P > 0·1 in all cases) in total fish abundance, species richness, diversity, equitability or the abundance of common species (Fig. 4).

Figure 4.

Summary of the depletion sampling data (riffle and deflector sites combined), showing fish abundance, species richness, diversity and equitability for manipulated (open columns) and control reaches (shaded columns). y axes represent the mean of six replicate sites ± 1 SE (note different scales on y axes).

Both artificial riffles and flow deflectors tended to increase flow velocity heterogeneity relative to the control reaches (Fig. 5). Similarly, depth heterogeneity tended to be greater in manipulated reaches, although for a number of sites there was little or no difference between control and manipulated reaches (Fig. 5). There were no significant relationships between fish (abundance, species richness and diversity) and depth or flow heterogeneity at the two flow levels. This was true for both absolute values and when data were expressed as a percentage change between manipulated and control reaches. Similarly, mean velocity showed no relationship to the fish fauna when absolute values were used, but a significant positive relationship existed when data were expressed as a percentage change (Fig. 6). At both flow levels, the relative increase in species richness and diversity was positively correlated to the relative increase in mean velocity (Fig. 6). This suggests that fish may respond positively to an increase in flow velocity in manipulated reaches, but the absence of any other relationships may indicate that physical/biological relationships are not simple, and may be influenced by other factors.

Figure 5.

Comparisons of flow and depth heterogeneity (expressed as CV) in manipulated and control reaches at riffle sites (open circles) and deflector sites (closed circles). Solid lines represent the 1 : 1 ratio between control and manipulated reaches, i.e. points that fall above that line are more heterogeneous in the manipulated reach.

Figure 6.

Relationship between the fish fauna (species richness and diversity) and mean flow velocity. Riffle sites (open circles) and deflector sites (closed circles). Axes are plotted as the percentage difference in the manipulated reach compared to its control. Results from Spearman's rank correlation are given.

Discussion

There was little evidence from this study that adding artificial riffles or flow deflectors substantially improved the conservation value of the fish assemblage, in terms of abundance, species richness, diversity and equitability. Whilst individual sites showed marked increases in abundance in the manipulated reach, this effect was not consistent among sites and the overall effects of rehabilitation were non-significant. Variability among sites showed no predictable pattern related to the age of the schemes, suggesting that among-site differences were not primarily due to different colonization periods. Only two species (stone loach and bullhead) were sufficiently abundant among sites to enable separate analysis. These two species showed no strongly significant response to rehabilitation, although at riffle sites bullhead exhibited a weakly significant increase in abundance and stone loach were significantly more abundant around artificial riffles in 2000. There were a number of possible reasons for the poor overall response of fish to the rehabilitation schemes: (i) the data were unreliable; (ii) the schemes failed to improve the physical environment for fish; (iii) the schemes were inappropriate for low-gradient rivers; and (iv) fish abundance and diversity lacked the potential to increase following rehabilitation because they were limited by poor water quality or the schemes were physically isolated.

are the data reliable?

Access restrictions in 2001 caused by foot and mouth disease, and the fact that only some sites were fished using depletion sampling, led to gaps in our data set. Furthermore, the large qualitative and quantitative differences in the fish fauna among the replicate sites reduced the number of species-specific comparisons that could be made. Despite these problems, a total of 1393 point samples taken from two reaches in 13 rivers over 2 years (plus depletion samples from six rivers) represent a considerable data set. To our knowledge, this is the largest data set aimed specifically at assessing the response to river rehabilitation of fishes in British lowland rivers.

Although not all rivers were fished using both sampling techniques, results from both were consistent in finding no significant effect of rehabilitation. This result suggests that the data obtained from point sampling were representative of the fish assemblage. Furthermore, a comparison of the two methods revealed significant positive relationships for both species richness and abundance of the ‘major’ species (all fish except small sized species like minnow and stickleback: see Appendix for full details). This confirms that, while absolute estimates may differ between methods, point abundance sampling was representative of the fish fauna as estimated by the more thorough depletion sampling technique. Overall, while gaps exist, the data presented appear representative of the fish fauna in each river.

Figure Appendix.

Comparison of point abundance and depletion sampling methods. Each point represents either a control or manipulated reach from the sites where both methods were used in summer 2000 (see Table 3). (a) Abundance of ‘major’ species (all fish excluding minnow, bullhead, stickleback, spined and stone loach, which were recorded only as present/absent in most depletion samples). (b) Total species richness in each reach (all fish species). Results from Spearman's rank correlation are given.

have the schemes improved the physical and hydraulic environment?

In general, the rehabilitation schemes have been successful in increasing flow heterogeneity and the variability of depth in manipulated reaches. There was some evidence that fish responded positively at sites where there was a large relative change in mean flow velocity between control and manipulated reaches. However, the fish fauna did not vary significantly between treatments, suggesting that this response was not strong enough to translate into a significant increase over the whole reach. It is likely that physical conditions were not altered over a sufficient area to have a significant positive effect on the fish fauna. There were no significant relationships between the fish fauna and the measures of flow and depth heterogeneity used in this study, indicating that physical rehabilitation of a river reach may not lead directly to biological rehabilitation. This raises questions about using physical criteria for measuring the success of schemes in terms of their ecological rehabilitation. Similarly, Muotka & Laasonen (2002) found that rehabilitation schemes increased substrate heterogeneity to the levels found in natural streams, but detritus retention and invertebrate abundance remained significantly lower than natural streams, indicating incomplete ecological rehabilitation. Our results, and those of Muotka & Laasonen (2002), have important implications for the rehabilitation and management of rivers because an assumption underlying river rehabilitation has been that increasing physical diversity improves biological diversity (Harper & Everard 1998; Newson & Newson 2000). While we do not refute this assumption outright, we believe that more consideration should be given to the scale and appropriateness of different rehabilitation projects for a given river.

are the restoration schemes appropriate?

There has been much scientific debate on the difficulties involved in rehabilitating aquatic systems to a previous state, or even defining a realistic target (Haslam 1996; Schouten 1996; Dobson & Cariss 1999). The schemes described in this study are, at best, aimed at enhancing the present environment rather than restoring it to a former condition. While in most cases this is the only practical approach, a possible reason for the failure of schemes to improve fish assemblages significantly may be that they are inappropriate for lowland rivers. In the past, most research into habitat improvement for fish has centred on salmonids, which tend to be found in high gradient, geomorphologically active, gravel-bed rivers. In such rivers, the use of flow deflectors and similar structures can increase fish abundance and biomass through the creation of deeper scour pools, greater pool area and higher mean depth at low flows (House & Boehne 1985; Riley & Fausch 1995; Linlokken 1997; Shields et al. 1998). In channelized lowland rivers, depth is unlikely to be a limiting factor for fish populations but, rather, flow deflectors have been used to create areas of faster flow and encourage sinuosity in the flow pattern of straightened channels. Similarly, artificial riffles were originally used to enhance salmonid spawning habitat in North America (Avery 1996) and have been adopted for use in British lowland rivers to benefit lithophilous spawners such as chub Leuciscus cephalus (L.) and dace L. Leuciscus (L.). In North America, artificial riffles have been successful in some cases but high failure rates have also been reported, indicating that the technique may only be useful in certain conditions (Avery 1996).

Bayley, O’Hara & Steel (2000) discuss the use of instream structures in the rehabilitation of low-gradient rivers and suggest that off-channel and marginal habitats are relatively more important in such systems. Vegetation plays a greater role in providing habitat complexity in lowland rivers, compared to high gradient rivers where channel morphology is naturally more variable. Furthermore, floodplain habitats may have a crucial role in the ecological functioning of lowland rivers (Gore & Shields 1995). Vegetated margins and backwaters connected to the main channel are important habitats for lowland river fishes, especially juveniles (Garner 1996, 1997; Copp 1997b; Lusk et al. 2001). The creation of shallow sloping banks and off-channel bays has been shown to benefit fish in a lowland river (Langler & Smith 2001), although little research has been conducted in this area. Despite the paucity of empirical data, it is likely that such habitats are more appropriate for low-gradient rivers dominated by fish that use vegetated areas for spawning, nursery/refugia and feeding habitats (Mann 1996). Similarly, restoring floodplain connectivity is beneficial to the riverine biota, as well as riparian plants and animals (Audsen, Sutherland & James 2001; Robertson, Bacon & Heagney 2001). Therefore, the creation of off-channel, marginal and floodplain habitats may be a better strategy for rehabilitating lowland rivers. River managers in the UK and elsewhere already create habitats such as backwaters and wet ledges/berms along river margins (Holmes 1998), and the biological benefits of such schemes need to be assessed more thoroughly.

does the fish fauna have the potential for increased density or richness?

Aquatic communities are naturally resilient to environmental disturbance (Giller & Myers 1996) and are able to recover following the rehabilitation of degraded habitats. However, the potential for the fish assemblage to respond to habitat improvement depends, firstly, on whether water quality is sufficient to support a diverse fish assemblage and, secondly, whether existing populations can disperse to, and exploit, the improved habitat. The failure of fish to respond positively to habitat rehabilitation in this study may have reflected one or both of these factors.

Most of our study sites showed clear signs of nutrient enrichment, and certain sites had particularly impoverished fish assemblages and low BMWP scores (e.g. the rivers Witham, Lark and Hiz). In addition, these sites had extensive growths of filamentous algae (Cladophora spp.) consistent with heavy enrichment (Mason 1996). These observations suggest that water quality may have been a limiting factor for fishes in these rivers. However, even when sites with few fish were ignored, there was little evidence from the remaining sites that the rehabilitation schemes significantly benefited the fish assemblage. While the remaining sites were also enriched to some extent, their rehabilitation potential was deemed to be higher because existing fish assemblages were more abundant and species rich, and therefore more capable of exploiting the rehabilitated reaches. We cannot discount the idea that water quality was limiting at these sites too, but suggest that other factors may influence the fish response.

The role of physical isolation in influencing biological recovery following habitat rehabilitation has been discussed for invertebrates (Fuchs & Statzner 1990; Wiberg-Larsen 1999). While little attention has been paid to fish, the same theory applies. For invertebrates, it has been shown that the more isolated the rehabilitation scheme, the slower the recovery rate (Fuchs & Statzner 1990; Hansen, Friberg & Baattrup-Pedersen 1999). Similarly, the response of fish to habitat rehabilitation will depend upon the source of available colonists. For example, in North American rivers, dispersal on a relatively large scale is the main mechanism through which trout (Salvelinus fontinalis and Salmo trutta) increase after habitat enhancement (Gowan & Fausch 1996). While coarse fishes may move considerable distances (Lucas & Batley 1996; Lucas et al. 2000), it is possible that the response to rehabilitation at some of our study sites was limited by the impoverished assemblage within reach of the rehabilitation scheme. This problem may have been exacerbated by the small scale of the schemes investigated in this study; they ranged from 150 m to 1000 m in length, while the length of river with degraded habitat was often many times greater. It is possible that such small-scale schemes can be effective only when they are close to high quality reaches supporting a greater number of potential colonists. Furthermore, the biological response even to large-scale rehabilitation projects has been moderate (at least in terms of invertebrates; see Biggs et al. 1998; Friberg et al. 1998) so it may be unrealistic to expect dramatic increases from small-scale schemes. This highlights the need to prioritize schemes adjacent to reaches where rehabilitation potential is greater as a more cost-effective approach than targeting isolated sites along impoverished rivers.

This study has demonstrated that, in many cases, the addition of flow deflectors and artificial riffles in lowland, low-gradient rivers had a minimal effect on fish assemblages. This weak response may have been because the schemes were inappropriate in design and scale, and that the fish populations lacked the potential for increase by such measures. We recommend that rehabilitation strategies for lowland rivers are revised and stress that techniques devised for higher gradient rivers may be inappropriate for many lowland rivers, where wetland, marginal and off-channel habitats may have more impact (Bayley et al. 2000). The schemes investigated here are typical of the earlier rehabilitation efforts made in Britain and there has since been considerable development in the field of river rehabilitation. However, similar schemes are still being implemented and detailed post-project appraisal is often a neglected aspect of most rehabilitation works. More extensive and directed biological monitoring is essential to improve our understanding and to enable future improvements in the design of river rehabilitation schemes and the selection of sites with greater potential for success.

Acknowledgements

This research was undertaken as part of a project on the ‘Restoration of Lowland River Habitats’ funded by EPSRC (grant number GR/L95014/01) awarded to Profs R.D. Hey and A.G. Hildrew, and Drs C. Smith and P.R. Wormleaton. We thank the Environment Agency, particularly Mike Atkinson (Anglia Central), Nick Bromidge (Anglia North) and Bob Preston (Thames North-west) for their help with site selection, provision of site information and electrofishing data.

Ancillary