Dr Fabio Lepori, Department of Ecology and Environmental Science, Umeå University, SE-901 87 Umeå, Sweden (e-mail email@example.com).
1Increasing degradation of ecological conditions in streams because of human activities has prompted widespread restoration attempts; however, the ecological consequences of restoration remain poorly understood. We explored the effects of restoration through placement of boulders into the channel in the Ume River catchment in northern Sweden, where tributary streams were extensively channelized to facilitate the transport of timber in the 19th and early 20th centuries. Retentiveness and breakdown of coarse particulate organic matter (CPOM), two key ecological functions in low-order streams most likely to be affected by channelization, were compared between restored, channelized and unimpacted reference stream sites.
2Artificial leaves were used to assess short-term CPOM retentiveness, while CPOM breakdown was estimated as the mass loss of alder (Alnus spp.) leaf packs placed in coarse mesh litter bags. Also, the taxonomic richness, abundance, biomass and evenness of the leaf-eating invertebrates (shredders) on the retrieved leaf material were quantified. Detailed field measurements were carried out to identify geomorphological and hydraulic controls of CPOM retentiveness and breakdown at the study sites.
3CPOM retentiveness reflected most strongly the density of boulders and submerged woody debris at the study sites. Restored sites were on average twice as retentive as channelized sites and significantly more retentive than reference sites when discharge was controlled.
4Current velocity at bank-full flow was the single most important predictor of CPOM mass loss, implying that mechanical fragmentation was substantial during high flows; other apparent controls of CPOM breakdown included water temperature and shredder abundance. CPOM mass loss was similar between restored and reference sites. However, breakdown was slightly faster at most channelized sites, consistent with higher hydraulic stress during high flow conditions. The shredder assemblages that colonized the litter bags were similar in richness, abundance, biomass and evenness between treatments.
5Synthesis and applications. In channelized forest streams, low retentiveness and fast mechanical fragmentation during high flows contribute to the rapid depletion of benthic CPOM following leaf abscission in autumn, thereby weakening the heterotrophic energy pathways that probably support much of the biological production in these systems. Our results illustrate that restoration by replacement of boulders can successfully reverse these impacts of channelization and thus contribute to the efficient ecological functioning of impacted streams.
In Europe and elsewhere, overexploitation of watercourses has degraded habitat conditions, reduced biodiversity and undermined ecosystem services such as supply of clean water and removal of pollutants (Malmqvist & Rundle 2002). Over the last few decades, however, growing public sensitivity to the consequences of environmental degradation has driven attempts to restore impacted systems (Ormerod 2003). Although in most cases restoration schemes have ecological goals and the practice of restoration has established a growing alliance with ecological research (Lake 2001), stringent evaluations of the ecological consequences of restoration schemes remain surprisingly scarce (Bash & Ryan 2002). This problem clearly warrants attention, because the ecological assessment of restoration work is necessary for management reasons and can provide valuable opportunities for improving our understanding of how ecosystems work (Bradshaw 1996).
Among the most widespread of human impacts on running water systems, channelization affects aquatic biota directly through changes in habitat and indirectly through changes in nutrients and energy dynamics (Petersen & Petersen 1991; Haapala & Muotka 1998; Negishi, Inoue & Nunokawa 2002). Streams have often been channelized for flood protection, navigation and land drainage. In boreal Scandinavia and North America, streams were also extensively channelized in the 19th and 20th centuries for timber transport, with up to 33 000 km being affected in Sweden alone (Törnlund & Östlund 2002). In Scandinavia, despite early concern over potential damage to fish and other biota, systematic attempts to rehabilitate channelized streams and rivers were implemented only after timber floating ended in the 1980s. Restoration projects have principally targeted the enhancement of structural heterogeneity of stream channels through the replacement of boulders and the removal of in-stream constructions (Muotka et al. 2002).
A crucial, yet still controversial, issue in restoration ecology concerns the indicators of success (Bradshaw 1996; Ehrenfeld 2000). To date, assessments of restoration success in streams have focused mostly on fish, particularly the abundance of and habitat quality for salmonids (Huusko & Yrjänä 1997; Zika & Peter 2002; Pretty et al. 2003). However, restoration in a broad ecological sense primarily entails the recovery of the processes by which ecosystems work, including resource dynamics and associated biological production (Palmer, Ambrose & Poff 1997; Richardson & Hinch 1998). From this perspective, which emphasizes overarching functions over single structural ecosystem elements, ecological processes represent natural indicators for evaluation of restoration. Because not all ecological processes can be monitored simultaneously, suitable subsets must be selected depending on the characteristics of the system to be restored, the original cause of degradation, and the practicality of quantification.
Small, forested streams are typically heterotrophic ecosystems in that secondary production depends on inputs of leaf litter and other allochthonous coarse particulate organic matter (CPOM) from the riparian vegetation rather than primary production within the stream (Vannote et al. 1980). Large inputs of CPOM do not, however, guarantee a high local availability of this resource to the stream biota. The accumulation of benthic CPOM will strongly depend on retentiveness, i.e. the ability of a channel to trap drifting CPOM into its streambed (Cummins et al. 1984; Speaker, Moore & Gregory 1984; Prochazka, Steward & Davies 1991). Only a fraction of the retained detritus is converted onto biomass by the stream biota. The largest part is broken down by a combination of other biotic and abiotic factors into fine particulate organic matter (FPOM), most of which becomes entrained and is exported to downstream systems (Wallace, Webster & Cuffney 1982; Webster & Benfield 1986). Therefore, CPOM breakdown driven by physical fragmentation causes local loss of base resources that partly offsets the effects of retentiveness; efficient utilization of CPOM by the stream biota thus depends on the balance between these contrasting processes (Webster et al. 1994). Given the contextual importance of CPOM, such balance constitutes a suitable end-point in the restoration of forested streams, especially where the original impact, as in the case of channelization, probably affects both processes involved.
This present study appraises the consequences of the restoration from channelization of typical second to fourth order forested streams in northern Sweden. The main aim was to evaluate the effects of restoration on stream ecosystem functioning by comparing CPOM breakdown and retentiveness between restored, channelized and reference stream sites. In addition, we considered which environmental factors best explained any differences in function between sites. We predicted that restoration would enhance retentiveness, and increase the relative strength of biotic over physical CPOM breakdown, leading to an overall increased efficiency in CPOM utilization.
study sites and experimental design
Assessing the ecological effects of restoration entails two distinct comparisons (Barmuta 2002). First, restored sites should be compared with their impacted (control) conditions to reveal whether the restoration produced any ecological change. Secondly, restored sites should be gauged against the target conditions, in this study identified as unimpacted streams within the same geographical area, to assess whether the restoration met its objectives.
To enable the first of these comparisons, seven pairs of restored and still channelized (control) sites were chosen, each pair located on a different stream in the Ume River system, northern Sweden (Fig. 1 and Table 1). Within each stream, channelized and restored sites were separated by 200–3680 m reaches (average 1353 m), all including deep pools, to minimize their direct spatial influence. The paired design was chosen to factor out most background variation, for example in stream size, water chemistry, temperature, flow and CPOM inputs, between treatments. Moreover, the control sites were chosen to match the physical characteristics of the paired restored sites as closely as possible, except for the effects of restoration. All streams except two were located in different subcatchments. The latter, Staggbäcken and Abmobäcken, were > 10 km apart and separated by a lake. At all sites, the channelization comprised multiple modifications of the stream channel (Törnlund & Östlund 2002), of which two were most apparent. First, all large obstructions, including boulders, large woody debris and bedrock outcrops, were removed by the successive use of black powder, dynamite and bulldozers. Secondly, the boulders dredged from the channel were used to build stone walls along the stream banks, the main purposes of which were to narrow and deepen the streams, and cut off backwaters where drifting logs could become stranded. The restoration of channelized streams in Nordic countries has often been implemented to enhance salmonid fisheries (Muotka & Laasonen 2002); in our study streams, however, the primary goal was to reverse the general ecological impacts caused by the channelization by recreating the natural state (Professor C. Nilsson, Umeå University, personal communication). The restoration methods, which were consistent across all study streams, involved the use of bulldozers to tear down the stone walls and replace most boulders (approximately 80–90%), typically 0·5–1 m in size, into the channel. The most noticeable effects of restoration on stream morphology included an increase in the density of protruding boulders, a widening of the channel and a decrease in current velocity (Table 1). Large woody debris was not added as part of the restoration activities, but was more abundant in restored stream reaches due to increased entrapment by the introduced boulders. All restoration work was carried out between 3 and 8 years prior to this study, allowing a reasonable amount of time for the biota to respond to the changed environmental conditions.
Table 1. Habitat characteristics (mean and range) of the study streams. Measurements were taken during summer base-flow conditions
Channelized (n = 7)
Restored (n = 7)
Reference (n = 5)
Stream order (range)
Current velocity (m s−1)
Discharge (m3 s−1)
Median particle size (cm)
Moss cover (%)
Alkalinity (µeq L−1)
Conductivity (µS cm−1)
To enable the second type of comparison, five reference sites were selected on unimpacted streams within the same area; these streams were independent from each other and from the restored and control sites (Fig. 1 and Table 1). Although the reference streams were not channelized, they were also not truly pristine. For instance, due to forestry practices, they were practically devoid of large woody debris (Liljaniemi et al. 2002), a habitat alteration with its own wide-ranging ecological implications (Smock, Metzler & Gladden 1989).
At all streams, sampling sites were established along uniform, wadable reaches. All sites measured 50 m, a length both practical and commensurate with the scale of the restoration, which affects riffles (tens to hundreds of metres long) separated by tranquil reaches that have been left untouched. All study streams drained boreal coniferous forests dominated by Scots pine Pinus sylvestris L. and Norway spruce Picea abies L. (Petersen, Gislason & Vought 1995). Deciduous trees, including willow Salix spp., birch Betula spp. and grey alder Alnus incana L., fringed the streams and supplied most input of leaf litter (Haapala & Muotka 1998). Litter inputs in northern Scandinavian streams are low (e.g. 4–200 g m−2), and are mostly contributed between August and October (Petersen, Gislason & Vought 1995; Haapala & Muotka 1998). Aquatic macrophytes are often limited to mosses, including Fontinalis spp., Schistidium agassizii Sull. & Lesq., Bryum spp. and Blindia acuta Bruch & Schimp. (Petersen, Gislason & Vought 1995; Englund, Jonsson & Malmqvist 1997). Draining crystalline catchments (Precambrian gneiss) topped by organic-rich soils, all streams had oligotrophic and slightly humic waters (dissolved organic carbon (DOC) 5–15 mg L−1) of low conductivity (15–32 µS cm−1) (Table 1). Discharge varied seasonally following a nival regime, with runoff from snowmelt peaking in April–June. Ice typically covered the streams from November to April. Forestry, including clear-cutting, ditching and harrowing, is the main land use in these catchments (Petersen, Gislason & Vought 1995).
At each study site, several habitat characteristics were measured along 10 equidistant transects during summer base-flow conditions (see Appendix 1). Water depth, current velocity near bed and at 0·6 × depth, percentage total moss cover and percentage Fontinalis spp. cover were systematically measured at increments of 1/6th of the stream width, while substratum size (intermediate axis) was measured from particles collected at increments of 1/10th of the width. Fontinalis spp. were also quantified separately from other mosses as their long stems are more likely to influence the retentiveness of CPOM than the other aquatic bryophytes found at the study sites. The percentage cover of each of eight habitat types (defined in Appendix 1) was measured by the line intercept method (Gordon, McMahon & Finlayson 1992). Current velocity at bank-full flow was estimated using Manning's equation (Gordon, McMahon & Finlayson 1992); the bank-full hydraulic radius was derived from depth measurements taken every 0·5 m along one representative transect, while Manning's n was estimated visually following Cowan (1956). Embeddedness was rated using the scale 1 (< 5%), 2 (5–24%), 3 (25–49%), 4 (50–74%) and 5 (75–100%). All pieces of woody debris with diameter > 3 cm occurring at the study sites were counted and the volume was calculated from their diameter assuming cylindrical shape.
Water samples collected in late summer, also during base flow, were analysed for pH, alkalinity, conductivity, total inorganic nitrogen by the laboratory of the Swedish University of Agricultural Sciences, Umeå Sweden (see Appendix 2).
Retentiveness was estimated as the percentage of CPOM particles retained out of a known number of particles released (Lamberti & Gregory 1996). We used plastic strips (4·2 × 5·2 cm) as CPOM particle analogues because they are easily located in streams and cannot be mistaken for naturally present leaves. These artificial leaves have similar flexibility and buoyancy to water-soaked, senescent alder leaves (cf. Young, Kovalak & Del Signore 1978; Speaker et al. 1988; Prochazka, Steward & Davies 1991; Larrañaga et al. 2003); moreover, their size reflects an average alder leaf (i.e. 22·1 ± 10·6 cm2, n = 56; F. Lepori, personal observations).
Before the artificial leaves were released, a seine net (1·8 m tall, mesh size 20 mm, 0·20 mm thread) fitted with a lead line was stretched across the stream at the downstream end of each study site. Five-hundred leaves were evenly scattered across the width of the wetted channel at the upstream end of the site for 10 min. Three hours after the release, when virtually all the leaves released had settled, non-retained leaves were collected from the seine net and counted. The percentage of leaves retained, i.e. the percentage of leaves that did not reach the seine net, was used as the index of retentiveness in the present study.
Discharge immediately upstream of the study site was measured during each experimental release by measuring depth and mean (0·6 × depth) water velocity at 0·5-m intervals along one transect.
The study of retentiveness was conducted between 19 August and 28 September 2003 to coincide with the leaf abscission period in the study area.
CPOM breakdown rate was quantified as the mass loss of introduced alder leaf packs exposed at the study sites for 41 days (Jonsson, Malmqvist & Hoffsten 2001). Senescent leaves collected in late August were air-dried for 2 weeks and placed in 4·0 ± 0·1-g portions in bags made of plastic garden netting (12 × 18 cm; 8 mm mesh size), coarse enough to allow invertebrate colonization. These litter bags were individually labelled, and the mass of their initial leaf content was recorded to the nearest 0·01 g. In mid-September 2003, 10 litter bags were placed at each study site, tethered in pairs to iron stakes driven into the streambed. The litter bags were broadly dispersed within each site, avoiding areas shallower than 0·2 m, to minimize risks of desiccation, and deeper than c. 1 m. Water velocity was measured in front of each litter bag at the beginning and the end of the study with a flow meter (C2, A. Ott, Kempten, Germany). At the end of the exposure period, the litter bags were collected, sealed individually in plastic bags and deep-frozen. In the laboratory, the fragments of alder leaves retrieved from the litter bags were oven dried for 48 h at 50 °C to determine dry mass, after silt and invertebrates had been thoroughly rinsed off over a 250-µm sieve. The initial mass was corrected for handling losses (estimated at 5·6%) and initial moisture content (Benfield 1996).
Data on invertebrate abundance, biomass, richness and evenness were generated to study the biotic components of litter decomposition (cf. Appendix 2). Invertebrates retrieved from the litter bags were stored in 70% ethanol; shredders (for definition see Jonsson, Malmqvist & Hoffsten 2001) were subsequently identified to species and counted. Also, the body length (Capniidae and Lepidostomatidae) or the head capsule width (other taxa) of all shredders was measured to the nearest 0·1 mm to estimate the dry mass from published mass–length equations (Capniidae and Lepidostomatidae: Smock 1980; other taxa: Meyer 1989). The species evenness of the shredder community was estimated using the probability of interspecific encounter (PIE; Hurlbert 1972), which is the probability that two individuals randomly drawn from a sample belong to different species:
where N is the total number of individuals in the sample and Ni is the number of individuals of the ith species in the sample. Low values of PIE indicate uneven communities dominated by one or few species.
CPOM retentiveness, breakdown rate, shredder richness, evenness (PIE), abundance and biomass were compared between restored and channelized sites using a randomized block analysis of variance (anova). Comparisons between restored and reference sites used one-way anovas in most tests. However, because the reference sites were slightly smaller than the restored sites (Table 1), and discharge is well-known to influence CPOM retentiveness in streams (Prochazka, Steward & Davies 1991; Larrañaga et al. 2003), CPOM retentiveness at restored and reference sites was compared using analysis of covariance (ancova) to account for the difference in discharge between these treatments.
Partial least-squares (PLS) regression analysis was carried out using the software Simca-P 10·0 (Umetrics AB, Umeå, Sweden) to identify potential controlling factors of CPOM retentiveness and breakdown at the study sites. Similar in purpose to multiple regression, PLS can contend better with intercorrelated predictors, as it combines sets of predictors into one or several independent components that best explain the dependent variables (Zhang, Malmqvist & Englund 1998; Eriksson et al. 1999). Variables and sites are attributed loadings and scores, respectively, which are interpreted in the same way as their analogues in principal component analysis (PCA). The overall relevance of individual predictors to the PLS model is expressed by variable importance in the projection (VIP) values, with VIP > 0·7 indicating important predictors (Eriksson et al. 1999). Only those components that explained more than 10% of the variation of the dependent variables (R2 > 0·1) were considered in this study. Variables with skewed distributions were log transformed to attain normality according to Kolmogorov–Smirnov tests.
During the leaf-retentiveness study, local rainstorms caused large differences in hydraulic conditions among streams, with discharge at the study sites ranging from one to 14 times higher than summer base flow (mean 3·7), depending on site. In spite of this variation, restored sites were on average more than twice as retentive as the channelized ones (68% vs. 30%; anova, d.f. = 1, F = 45·35, P = 0·001; Fig. 2). Moreover, differences in retentiveness were remarkably consistent among the seven different pairs of restored and channelized sites (anova, d.f. = 6, F= 1·16, P= 0·445).
CPOM retentiveness was similar between restored (68%) and reference sites (65%; Fig. 2). However, the restored sites were significantly more retentive than the reference sites when the effect of differences in discharge between streams was removed by ancova (d.f. = 1, F= 7·96, P= 0·020). In this comparison, the negative effect of discharge on retentiveness was highly significant (ancova, d.f. = 1, F= 15·54, P= 0·003).
The PLS analysis extracted two components with R2 > 0·1, collectively explaining 88% of the variation in retentiveness between sites (Table 2). The first PLS component loaded positively on woody debris density and percentage of run with boulders, and negatively on percentage run and bank-full current velocity (see Appendix 1 for definition of habitat types). This component therefore reflected the density of large-scale roughness elements (boulders and woody debris), which in turn control hydraulic resistance and current velocity. The interpretation of the second PLS component, which loaded negatively on mean depth and percentage of backwater habitat, and positively on near-bed mean velocity, sorting and woody debris density, was less clear.
Table 2. Loadings of a PLS analysis predicting CPOM retentiveness from geomorphological and hydraulic variables at 19 stream sites in northern Sweden. Variables are ranked after variable importance in the prediction values (VIP). Predictors with a positive sign contribute positively to CPOM retentiveness. Loadings ≥ 0·3 are in bold. Only predictors with VIP > 0·7 are included in the table (see definitions of symbols and full set of predictors in Appendix 1)
Component 1 (R2 = 75%)
Component 2 (R2 = 13%)
U, current velocity; U-sd, current velocity (standard deviation); U-av, current velocity (average); D84, 84th percentile of the particle size distribution; D50, 50th percentile of the particle size distribution.
% run with boulders
0·6 × depth U-sd
0·6 × depth U-av
The scores of the restored and channelized sites were most distinct along the first PLS component, with the restored sites having higher densities of woody debris, higher percentage of run with boulder habitat, lower percentage of run habitat, and slower bank-full current velocities (Fig. 3). The scores of the reference sites largely overlapped those of the restored sites along the first two PLS components (Fig. 3).
The breakdown study started under low-flow conditions, but rainstorms in the middle of the exposure time led to a spate, lasting approximately 1 week, during which water stage rose to bank-full level at all sites. During the exposure, the leaf packs lost 40–85% of their initial dry weight, with most losing roughly half of their initial weight (Fig. 4). In six out of the seven pairs of restored and channelized sites, CPOM mass loss was lower at restored sites, but the difference was not significant (mean 2·0 vs. 2·3 g; anova, d.f. = 1, F= 2·11, P= 0·197; Fig. 4). CPOM mass loss was not significantly different between restored and reference sites (mean 2·0 vs. 2·3 g; anova, d.f. = 1, F= 0·45, P = 0·516; Fig. 4).
The first two components extracted by PLS explained 58% of the variation in CPOM mass loss between sites (Table 3). The first component loaded positively on bank-full current velocity and temperature, and negatively on PIE, indicating that faster CPOM breakdown was associated with uneven shredder assemblages. The second component reflected mostly shredder abundance, temperature and shredder richness. Shredder biomass, pH, nutrient status and current velocity in front of the litter bags (measured during low flow) were not strongly associated with CPOM mass loss.
Table 3. Loadings of a PLS analysis predicting CPOM mass loss from hydraulic, chemical and invertebrate variables at 19 stream sites in northern Sweden. Variables are ranked after variable importance in the prediction values (VIP). Predictors with a positive sign contribute positively to CPOM breakdown. Loadings ≥ 0·3 are in bold. Only predictors with VIP > 0·7 are included in the Table (see definitions of symbols and full set of predictors in Appendix 2)
Component 1 (R2 = 42%)
Component 2 (R2 = 16%)
Pairs of restored and channelized sites had consistent differences in PLS scores along the first component, indicating that restored sites were associated with slower bank-full velocities (Fig. 5). Most reference sites had relatively high scores along the second PLS component, probably reflecting high shredder abundance and richness (Fig. 6).
The species richness, number per litter bag, evenness and biomass per litter bag of the shredders collected from the remaining leaf material were not significantly distinct between restored and channelized sites, or between restored and reference sites (anovas, results not shown), but some noticeable patterns emerged from the data (Fig. 6). For instance, shredder assemblages at reference sites tended to be more abundant, richer in species and more even relative to channelized and restored sites. Compared with channelized sites, restored sites tended to have sparser, but slightly more diverse (higher richness and evenness), shredder assemblages.
Stream ecology theory states that forested streams are energetically reliant on leaf litter, and the pulsed contribution of this base resource to streams often leads to seasonal depletion, forcing detritivore invertebrates through resource bottlenecks (Richardson 1991). Northern boreal streams are particularly prone to early CPOM depletion, because they receive low inputs of leaves that are processed rapidly in the water (Malmqvist & Oberle 1995). In this context, further reduction of benthic CPOM because of channelization might have severe consequences for invertebrate production. Indeed, in a Finnish stream channelized for timber floating, benthic CPOM was noticeable only during a short period following leaf abscission (Haapala & Muotka 1998).
Consistent with the few existing similar studies, our results support the prediction that the restoration of channelized reaches through placement of boulders enhances CPOM retentiveness, and therefore reduces losses of food resources to downstream systems (Muotka & Laasonen 2002; Negishi & Richardson 2003). Mechanisms of CPOM retention in streams vary temporally with discharge conditions (Maridet et al. 1995); crucially, however, the difference in retentiveness between the restored and the channelized sites in this study was consistent over a wide range of hydrological conditions. Considering that exports of leaf litter from stream reaches typically peak during spates (Malmqvist, Nilsson & Svensson 1978), the differential retentiveness between channelized and restored reaches during high flows is clearly relevant to the long-term dynamics of benthic CPOM at the study sites, and probably at restored sites in general.
In streams, the direct trapping of CPOM particles by channel obstructions in runs and riffles has been considered a more important mechanism of CPOM retentiveness than the deposition in dead zones in the channel (Speaker, Moore & Gregory 1984; Webster et al. 1994). Although depositional areas were infrequent at the study sites, our results support this view, showing that retentiveness was strongly associated with the density of protruding boulders and woody debris in the channel. Besides being retentive structures themselves, boulders and woody debris also promote hydraulic complexity, causing CPOM particles to drift with more complex trajectories, thereby increasing their probability of being trapped by other obstructions. Our results emphasize the benefits of stream rehabilitation practices involving boulders or woody debris additions to the resource dynamics of running water systems.
In Finland, channelized streams rehabilitated with techniques similar to those used in Sweden remained significantly less retentive than natural streams for at least 3 years after the restoration work, purportedly because aquatic mosses, a key retentive feature in those systems, had been negatively affected (Muotka & Laasonen 2002). In contrast, in our case, the retentiveness of the restored sites substantially exceeded that of the reference streams. Such a difference was probably unrelated to differences in moss cover, as moss cover was not an important control of retentiveness at our sites. Given the comparative nature of the present study, we cannot exclude the possibility that important differences between these treatments predated the channelization. However, the high CPOM retentiveness at the restored sites might also indicate some artificiality in the density or the distribution of retentive structures. For instance, medium-sized rocks (0·1–1 m) tended to be more abundant in restored than in natural streams, as many originated from the blasting of boulders or bedrock during the channelization work (F. Lepori & D. Palm, personal observations). This difference, which probably contributes to CPOM retentiveness, shows that restoration might not recreate perfectly natural conditions, as some impacts of channelization are irreversible.
Rates of CPOM decomposition in streams often correlate positively with shredder benthic densities (Webster & Benfield 1986 and references therein), implying that the colonization and subsequent processing of CPOM are determined by the local availability of detritivores. In turn, several authors have argued that shredder densities depend on the channel structural complexity and ability to retain inputs of detritus (Rounik & Winterbourn 1983; Hildrew et al. 1991; Prochazka, Steward & Davies 1991); channel retentiveness might therefore indirectly affect rates of CPOM decomposition (Gelroth & Marzolf 1978; Rounik & Winterbourn 1983). Such a bottom-up effect, however, was not confirmed by the present study. The higher retentiveness at restored relative to channelized sites was not accompanied by faster CPOM breakdown; moreover, the abundance of the shredders colonizing the litter bags did not differ between the treatments.
There are a number of possible causes for the seeming lack of dependence between shredder abundance in the litter bags and retentiveness at our sites. First, at the time of the study detritus availability might not have limited shredder densities because of the recent input, and might not have become limiting until late winter or spring (Malmqvist & Oberle 1995). Secondly, benthic density of shredders might have differed between retentive and non-retentive sites, but any effect on the colonization of the litter bags was offset, or mitigated, by compensatory factors. In retentive stream reaches, for instance, superabundant CPOM might dilute the local pool of shredders (cf. Reice 1991) so that shredder density per individual leaf pack might remain low even at high benthic densities. In contrast, sparse leaf packs could act as resource magnets attracting high numbers of detritivores in poorly retentive locations (Webster & Waide 1982; Benfield & Webster 1985; Dobson 1999). Finally, the shredders colonizing the litter bags might have been recruited from areas outside the boundaries of the restored reaches (typically 50–200 m). In this case, any habitat changes caused by restoration might be at a different spatial scale than that controlling patterns of invertebrate colonization of leaf litter.
Most studies on the controls of litter breakdown in streams have emphasized the role of biotic factors, including shredder activity and microbial conditioning (Webster & Benfield 1986; Graça 2001). However, under the wet, but not atypical, autumnal weather conditions accompanying our study, current velocity at bank-full flow was the single most important predictor of CPOM mass loss. Two main conclusions may be drawn from this result. First, mechanical fragmentation during high flows must have been an important component of leaf litter breakdown in the study streams (Heard et al. 1999). Secondly, the importance of current velocity explains why litter breakdown tended to be slower at restored compared with channelized sites: channelized sites had substantially faster current velocities during high flows because the lack of boulders caused lower hydraulic friction. The weaker association between CPOM mass loss and current velocity during low flow conditions suggests that the relative importance of different controls of CPOM breakdown changes dynamically with flow conditions, with physical abrasion and fragmentation driving breakdown during high flows, and other factors, including biotic ones, becoming more important during low flows.
Because spates frequently occur in autumn and spring in boreal Sweden, two periods when substantial CPOM occurs in streams and heterotrophic energy pathways are important, the slower CPOM breakdown at most of the restored sites is probably of ecological relevance. A reduction of mechanical fragmentation during high flows means that less detritus is lost to downstream systems; as a result, following the autumnal input, CPOM might remain available for biotic processing in larger amounts and for longer (Dobson & Hildrew 1992), creating conditions conducive to efficient CPOM utilization and high secondary production.
Although not clearly distinct between treatments, biotic factors were probably relevant to the differences in CPOM breakdown among individual sites. The associations between CPOM breakdown and temperature (possibly reflecting effects of microbial activity; Irons et al. 1994), shredder abundance and shredder species richness are consistent with previous observations from Sweden (Jonsson, Malmqvist & Hoffsten 2001) and elsewhere (Webster & Benfield 1986; Graça 2001). However, the negative association between evenness and CPOM breakdown rates in this study implies that biotic controls of detritus decomposition in streams might be complex, and involve aspects of community structure that we are only beginning to appreciate (Dangles & Malmqvist 2004). Intriguingly, evenness correlated negatively with bank-full current velocity, suggesting that hydraulic disturbance in locations of low flow resistance might have been instrumental in the uneven invertebrate assemblage structure observed (Negishi, Inoue & Nunokawa 2002). Although circumstantial, these findings indicate the possible existence of far-reaching linkages between habitat conditions, community structure and ecosystem functioning that deserve further investigation.
Although effects on biotic processing were not substantiated, this study illustrates that restoration through placement of boulders into the channel enhances the availability of leaf litter to lotic biota through increased retentiveness and reduced mechanical fragmentation at high flows. Considering the well-documented importance of leaf litter to forest streams, we believe that the restoration of streambed heterogeneity in channelized streams can significantly contribute to the efficient functioning of these systems.
We thank Per Nilsson and Johan Baudou for assistance in the field, Per-Ola Hoffsten for skilful identification of the invertebrates, and Erik Törnlund, Stig Westbergh, Tommy Stenlund and Daniel Holmqvist for help with the selection of the study sites. Christer Nilsson, Niclas Hjerdt, Brendan McKie, Zlatko Petrin, Per-Ola Hoffsten and three anonymous referees provided helpful comments on the manuscript. This research was funded by the Swedish EPA through the Local Investment Program (LIP).