Introduction

Many rivers and streams have been channelized for various purposes across the world. Channelization reduces the heterogeneity of physical habitats in streams and rivers into a homogeneous one. Numerous studies have shown that a loss of habitat diversity caused by channelization has decreased the abundance and taxon richness of macroinvertebrates (e.g., Quinn et al. 1992; Bis et al. 2000; Negishi et al. 2002). Alteration of physical habitats by channelization is the most significant threat to biodiversity and river ecosystem function worldwide (Allan and Flecker 1993; Rosemberg et al. 2000; Nakamura et al. 2002; Nakamura and Yamada 2005). River restoration and rehabilitation to restore biodiversity and ecosystem function appear to be a popular arena of political, social, and scientific investment in Europe and North America. Several restoration programs have recently been conducted in streams and rivers in Far East Asia, such as in South Korea (Shin and Lee 2006) and Japan (Nakamura 2003).

Few other lotic ecosystems have been more altered by human impacts than lowland rivers of temperate regions, because their floodplains have been the center of human activity (Nilsson et al. 2005). In addition, channelization of lowland rivers in these regions has a long history, and therefore, little is known of their original condition. Lack of knowledge on the reference condition complicates the restoration of lowland rivers. However, many restoration projects have endeavored to restore ecosystem processes in lowland rivers (e.g., Brookes and Shields 1996; Waal et al. 1998).

The lower segment of the Shibetsu River in northern Japan historically meandered through its floodplain. Channelization work between the 1950s and 1970s for flood control and farmland reclamation eliminated the physical complexity of in-stream habitats and caused extirpation of native fish species in the Shibetsu River (Hirai and Kuga 2005). Growing interest in river health in Japan has led to a program to reconstruct meanders in the Shibetsu River to restore the river ecosystem (Nakamura 2003). Because of lack of experience in meander restoration in Japan, a small-scale pilot project was launched to test construction methods and examine the response of aquatic biota, about 8.5 km upstream from the river mouth.

Although the number of studies on the responses of stream and river ecosystems to river restoration has increased (Ormerod 2004), the number dealing with lowland rivers is limited. Thus, approaches to restoring lowland river ecosystems have been limited. The objective of this study is to synthesize the findings of previous basic and applied works, which contributes to the success of future restoration projects in lowland rivers. In this paper, we focus on macroinvertebrates, which play an important role in the structure and function of lotic ecosystems (e.g., Covich et al. 1999; Allan and Castillo 2007) and are frequently used as indicators of the ecological status of rivers to evaluate the efficacy of river restoration (Karr and Chu 1999). We initially review the geomorphic and hydraulic characteristics of lowland rivers and their traits as macroinvertebrate habitats. We then describe the responses of macroinvertebrate communities to river restoration in the lower segment of the Shibetsu River, and finally discuss approaches to restoring macroinvertebrate communities in lowland rivers.

Characteristics of lowland rivers as macroinvertebrate habitats

In general, lowland rivers have higher-order channels within an unconstrained valley segment and exhibit less turbulence than mountain streams with a higher gradient (Brussock et al. 1985; Bisson et al. 2006). Reaches of lowland rivers have little longitudinal change in their physical environment, compared with features such as riffle-pool sequences in mountain streams. In contrast to longitudinal change, lateral variations such as asymmetrical cross-section are greater in lowland rivers, because point bars are formed at meander bends by meandering flows. This reach type consists of low-gradient, meandering channels with a predominantly fine substrates (Church 2002). Sediment movement occurs not only during floods, but also in base-flow conditions in thalweg of lowland rivers. This continuous transport of sediments forming an unstable streambed is one of the main characteristics of lowland rivers.

In a lowland river in Germany, Brunke et al. (2002) found that the shifting sand habitat was located in the deep midstream floor of the river, where flow velocity and transport capacity for sediment were high. The abundance and richness of macroinvertebrates in the shifting sand habitat was much lower than in other habitat types such as woody debris, roots, and stable sand located at the stream’s margin with low-flow conditions. Macroinvertebrates in the deep water riverbed at midstream (i.e., thalweg) were depauperate in lowland rivers of Canada (Rempel et al. 2000) and Japan (Nakano and Nakamura 2006a). Shear velocity, which affects riverbed stability, increased with increasing water depth, and was negatively related to the density and richness of macroinvertebrates in these lowland rivers (Rempel et al. 1999; Nakamura et al. 2008). These previous studies indicated that riverbed stability is an important factor determining macroinvertebrate distributions in lowland rivers. In contrast to lowland streams and rivers, macroinvertebrate density and richness were highest in the midstream of riffles, where shear velocity was highest in mountain streams, probably because high shear flow plays an important role in maintaining interstitial spaces between coarse substrates that are suitable for macroinvertebrates (Nakano and Nakamura 2006a). This suggests that the difference in the relationship between macroinvertebrate assemblages and shear velocity between mountain streams and lowland rivers is attributable to streambed stability during base-flow conditions.

Woody debris provides various habitats for macroinvertebrates from mountain streams to lowland rivers (Benke and Wallace 2003). Stabilized woody debris can itself be a major habitat, especially in lowland rivers, because of a lack of stable substrates there (e.g., Benke et al. 1984; Rabeni and Hoel 2000). As well as woody debris, rip-rap generally supports a high abundance and diversity of macroinvertebrates in a lowland river (Brunke et al. 2002). Natural stable substrates including woody debris, roots projecting from the bank, and large stones can be major habitats for macroinvertebrates in lowland rivers. Johnson et al. (2003) showed the importance of large woody debris as a macroinvertebrate habitat and indicated negative impacts of woody debris removal on macroinvertebrate communities in low-gradient streams. In addition, bank protection decreases the recruitment of woody debris and large stones from the stream bank. Furthermore, several previous studies have evaluated the impacts of woody debris removal, including increased water velocity (Gregory 1992), decreased channel stability (Bilby 1984; Heede 1985), and enhanced scouring of sediments (Beschta 1979). Thus, bank protection and woody debris removal for transportation enhancement significantly decrease stable substrates and riverbed stability.

Approaches to restoring macroinvertebrates in lowland rivers

The pilot project to restore a meander course in the Shibetsu River was completed in March 2002 (Fig. 1). Several studies have reported the effects of meander restoration on macroinvertebrate assemblages in lowland rivers of Europe (Friberg et al. 1994, 1998; Biggs et al. 1998). However, relationships between changes in the physical environment and the macroinvertebrate assemblage structure are still ambiguous, because none of these studies analyzed changes in hydraulic parameters associated with re-meandering. To evaluate the effects of changes in physical environments by meander restoration on macroinvertebrates, abundance, richness, and distributions of macroinvertebrates were compared among three reaches: a restored meandering reach (restoration reach), a channelized reach (control reach), and a naturally meandering reach (reference reach), from June 2002 to June 2004 (Nakano and Nakamura 2008). The channelized reach was located upstream of the restoration reach of the Shibetsu River. The reference reach was located in the lower segment of the Nishibetsu River, flowing immediately south of the Shibetsu River. Streambed materials in both rivers consisted mainly of sand and gravel, with an average diameter of approximately 12 mm in the Shibetsu River, and 10 mm in the Nishibetsu River. Shifting sands were observed on the streambed for water depths of more than 50 cm in both rivers. The restoration and reference reaches had asymmetrical cross-section at their channel bends and showed high cross-sectional diversity in physical variables. When five sampling events in the study area were carried out in base flow condition (June 2002, November 2002, June 2003, November 2003, and June 2004), total taxa richness across a reach of two meandering reaches was always higher than that of the channelized reach (Nakano et al. 2005; Nakano and Nakamura 2008). Figure 2 shows the cross-sectional distribution of physical variables (water depth and shear velocity) and macroinvertebrate community structure (taxa richness and total density) in June 2004. Sampling points of physical variables were located at nine equidistant positions along a transect established in each reach, numbered from the left bank (the edges of the left and right bank being 0 and 10, respectively). Four macroinvertebrate samples (400 cm2) were collected from a sampling location (1.0 × 1.0 m quadrat) established at each alternate sampling point (with an odd number). Macroinvertebrate abundance and richness were highest at the shallowest habitats in the two meandering reaches (Fig. 2). In addition, almost all taxa observed in the two meandering reaches were concentrated in the shallowest margins of point bars created at the inside of the bend. In contrast with the meandering reaches, the channelized reach showed a trapezoidal cross-section with a uniform water depth, and had no concentration of macroinvertebrates at any cross-sectional locations (Fig. 2). Macroinvertebrate composition has been improved by the meander restoration through the formation of edge habitat with a shallow depth (Nakano and Nakamura 2008). Caddisflies, Padunia forcipata, which are mainly found in habitats with slow flow conditions were abundant in the edge habitat (Nakano and Nakamura 2006b). In contrast, Ceratopogonidae, which probably tolerate unstable streambeds, were the most abundant taxa in the control reach. Shear velocity increased with water depth and had a negative association with macroinvertebrate abundance and richness (Fig. 3). Even in base flow conditions, increases in shear velocity could induce movement of streambed sediment in lowland rivers. These results suggest that the recovery in macroinvertebrate abundance and diversity in the restoration reach occurred through the creation of a stable riverbed suitable for macroinvertebrates around the stream margins with shallow depth (i.e., low shear stress).

Fig. 1
figure 1

Photograph of the experimental restored meandering reach in the Shibetsu River, northern Japan

Full size image
Fig. 2
figure 2

Cross-sectional distribution of water depth (top row), shear velocity (second row), taxa richness (third row), and total density (bottom row) at three reaches (reference, restoration, and control) in June 2004. The restoration and the control reaches are in the Shibetsu River. The reference reach is in the Nishibetsu River, located immediately south of the Shibetsu River (modified from Nakano and Nakamura 2008). The same lower-case letters indicate no significant difference (Tukey HSD test, p > 0.05)

Full size image
Fig. 3
figure 3

Macroinvertebrates density (a) and taxa richness (b) in relation to shear velocity at each sampling point in the three study reaches. The line corresponds to the regression equation (modified from Nakano and Nakamura 2008)

Full size image

As mentioned above, stable substrates are major habitats for macroinvertebrates in lowland rivers with fine substrates. Thus, large woody debris additions could be another rehabilitation option. Large woody debris additions may stabilize bed materials and provide suitable habitats for macroinvertebrates, although how much bed material could be stabilized with large woody debris are unknown. In our study, bank erosion on the outside of the bend promoted uprooting and falling of riparian trees into the river. Thus, meandering channel flows create two major habitats for macroinvertebrates: the marginal habitat with a stable riverbed at the edges of the point bar (inside of bend), and the wood habitats (trunks, roots, and branches) formed at the meander scroll (outside of the bend) in lowland rivers (Fig. 4). Recovery of the original meandering channel and natural flow regime could be the best way to restore macroinvertebrate assemblages in a channelized lowland river. However, reconstruction of meanders may not be feasible because of social and economic constraints. In particular, setback of artificial dikes to create spaces for a meander channel is one of the most difficult approaches available in the floodplain of lowland rivers. In such a situation, installation of in-stream structures consisting of large woody debris (Erskine and Webb 2003; Brookes et al. 2004; Spanhoff et al. 2006) and boulders (Negishi and Richardson 2003) could serve as an alternative. Nakano and Nakamura (2006b) also demonstrated that placement of groynes created a stable edge habitat and increased the abundance and diversity of macroinvertebrates in a channelized lowland river. We conclude that, in order to recover macroinvertebrate communities in channelized lowland rivers, a variety of restoration methods stabilizing substrates should be implemented.

Fig. 4
figure 4

Two major habitats (stable riverbed formed at the inside of the bend and wood habitat from fallen trees at the outside of the bend) for macroinvertebrates created by meandering flows in lowland rivers

Full size image