Short-Term Impacts of Remeandering Restoration Efforts on Fish Community Structure in a Fourth-Order Stream

Channel reconfiguration is a common but debated method used to restore streams, often causing disturbance and producing subsequent negative impacts on biota. Here, we report results from short-term assessment (i.e., one and three years’ post-restoration) of habitat variables (e.g., reach depth, substrate, and canopy cover) and fish community composition and structure (using electrofishing surveys; e.g., proportion of juveniles and tolerant fishes) from a 675 m section of Eagle Creek (Portage County, OH, USA) restored using channel remeandering in August 2013. Mesohabitat analysis was not conducted as part of this study. Sites upstream and downstream of restoration efforts were also monitored. Surveys were completed in 10 separate 50 m stretches: one upstream control site, three new channel sites, two old channel sites, and three downstream sites. Following restoration, fish communities in downstream sites became more similar to new channel sites and diverged from the upstream control site over time, as reflected in increased proportions of juvenile and tolerant fishes. Shifts in fish communities were not explained by habitat variables. Diversity was significantly lower in new channel sites post-restoration than in the upstream control, while downstream sites remained similarly high in diversity compared to the upstream control site over time. Overall, in the short-term, new channel colonizing communities were unable to recover to reflect upstream community composition and structure, and fish communities downstream of restoration were negatively impacted.


Introduction
To improve stream health and function due to anthropogenic degradation, restoration efforts are now common across the U.S. with more than US $1 billion spent on projects each year [1,2]. The majority of stream restoration projects have one or more goals targeted towards increasing biodiversity, stabilizing the channel, improving riparian habitat, improving water quality, and creating in-stream habitat for biota, and most efforts utilize in-stream hydromorphic changes (e.g., addition of boulders, other sediments, and log jams) and channel hydromorphic alteration/reconfiguration (e.g., raising or lowering the bed to reconnect to the floodplain, creating new meanders, and lateral movement of the channel) along with riparian restoration with riparian plantings to accomplish project goals [3]. Although the practice of stream restoration is widespread, both short-and long-term monitoring to evaluate restoration goals are generally lacking [1,[3][4][5]. However, short-term monitoring efforts have increased in recent years [6] and are largely due to implementation by the U.S. Army Corps of Engineers and U.S. Environmental Protection Agency of the 2008 Compensatory Mitigation for decline with an increase in tolerant fishes in both the newly reconfigured channel and in downstream areas. Although the goals of the restoration project in this study are targeted towards channel stability and reconnection with the floodplain, increased biodiversity should be key outcomes of a successful stream restoration project.

Field and Survey Site Descriptions
Hiram College's Eagle Creek restoration site (HCECRS) is a 152-acre plot of land located within the Mahoning River watershed at the confluence of Silver Creek and Eagle Creek in Hiram Township (Portage County, OH, USA, 41 • 28 62" N, 81 • 12 09" W; Figure 1). Land use in the Upper Mahoning River basin is predominantly agriculture (62.7% = cropland and pasture land) along with forested land (13.9%) and urban/residential use (10.2%) [24]. Between 2006 and 2007, this site was heavily logged. It is bordered directly upstream by residential uses and agriculture, causing Eagle Creek to have heavily eroded banks, high turbidity, unstable riffles, and poor canopy cover throughout a substantial portion of the reach. Eagle Creek is a fourth-ordered stream, and sediment ranges from clay and sand to artificially placed boulders. A total of nine sites were surveyed in this study, each consisting of a 50 m stretch of stream. Sites included an upstream control site, two sites within the old, abandoned channel (referred to as the old channel throughout), three sites within the newly, reconfigured channel (referred to as the new channel throughout), and three sites downstream of the reconfigured channel (referred to as downstream sites throughout; Figures 1 and 2). In old channel sites, measurements (fish and habitat variables) were only taken during 2013 (pre-restoration) as it quickly converted into a stagnant oxbow following diversion of water into the new channel.
Water 2017, 9,546 3 of 15 targeted towards channel stability and reconnection with the floodplain, increased biodiversity should be key outcomes of a successful stream restoration project.

Field and Survey Site Descriptions
Hiram College's Eagle Creek restoration site (HCECRS) is a 152-acre plot of land located within the Mahoning River watershed at the confluence of Silver Creek and Eagle Creek in Hiram Township (Portage County, OH, USA, 41°28′62″ N, 81°12′09″ W; Figure 1). Land use in the Upper Mahoning River basin is predominantly agriculture (62.7% = cropland and pasture land) along with forested land (13.9%) and urban/residential use (10.2%) [24]. Between 2006 and 2007, this site was heavily logged. It is bordered directly upstream by residential uses and agriculture, causing Eagle Creek to have heavily eroded banks, high turbidity, unstable riffles, and poor canopy cover throughout a substantial portion of the reach. Eagle Creek is a fourth-ordered stream, and sediment ranges from clay and sand to artificially placed boulders. A total of nine sites were surveyed in this study, each consisting of a 50 m stretch of stream. Sites included an upstream control site, two sites within the old, abandoned channel (referred to as the old channel throughout), three sites within the newly, reconfigured channel (referred to as the new channel throughout), and three sites downstream of the reconfigured channel (referred to as downstream sites throughout; Figures 1 and 2). In old channel sites, measurements (fish and habitat variables) were only taken during 2013 (pre-restoration) as it quickly converted into a stagnant oxbow following diversion of water into the new channel.

Restoration Process and Goals
In 2012, approximately 497 linear meters of Eagle Creek was reconfigured and diverted into an excavated meandering channel extending 675 linear meters. The active channel was constructed to be wider than the old channel and upstream and downstream areas to allow movement of the wetted channel over time. Artificial boulders were placed in the new channel to maintain channel grade elevation, dimension, and a thalweg and to provide habitat structure. No other sediments (i.e., gravel or cobble) were placed in the new channel, and natural colonization of particles from upstream has been allowed to fill in with time. Following excavation of the new channel, native trees were planted and native grasses and forbs were seeded along the banks and within the floodplain with the goals of decreasing erosion and sedimentation and to provide future canopy cover to decrease light intensity and in-stream temperatures. Due to severe flooding during the summer of 2012, water was not diverted into the new channel until August 2013, and, subsequently, streamside vegetation had an additional year to help better stabilize the banks. The major goals of this project were to: (1) re-engage the floodplain and decrease flooding downstream, (2) decrease erosion and siltation within the stream, and (3) to alter the thermal regime by decreasing water temperatures via long-term recolonization of forest. No in-stream habitat improvement for fishes has been done. Restoration design and implementation were completed by a local consulting firm.

Measurement of Habitat Variables
Average substrate size was measured in each 50 m stretch using Wolman pebble counts [25] (for a full description of methods see [26]) once during each summer. Depth (cm) and current velocity (m/s 2 ) were measured at 15 random locations within each sampling site using a wading rod and a Hach ® flow meter (Model FH950.0, Loveland, CO, USA). All other variables, with the exception of wetted channel width (measured along six transects with a measuring tape), were measured in each

Restoration Process and Goals
In 2012, approximately 497 linear meters of Eagle Creek was reconfigured and diverted into an excavated meandering channel extending 675 linear meters. The active channel was constructed to be wider than the old channel and upstream and downstream areas to allow movement of the wetted channel over time. Artificial boulders were placed in the new channel to maintain channel grade elevation, dimension, and a thalweg and to provide habitat structure. No other sediments (i.e., gravel or cobble) were placed in the new channel, and natural colonization of particles from upstream has been allowed to fill in with time. Following excavation of the new channel, native trees were planted and native grasses and forbs were seeded along the banks and within the floodplain with the goals of decreasing erosion and sedimentation and to provide future canopy cover to decrease light intensity and in-stream temperatures. Due to severe flooding during the summer of 2012, water was not diverted into the new channel until August 2013, and, subsequently, streamside vegetation had an additional year to help better stabilize the banks. The major goals of this project were to: (1) re-engage the floodplain and decrease flooding downstream; (2) decrease erosion and siltation within the stream; and (3) to alter the thermal regime by decreasing water temperatures via long-term recolonization of forest. No in-stream habitat improvement for fishes has been done. Restoration design and implementation were completed by a local consulting firm.

Measurement of Habitat Variables
Average substrate size was measured in each 50 m stretch using Wolman pebble counts [25] (for a full description of methods see [26]) once during each summer. Depth (cm) and current velocity (m/s 2 ) were measured at 15 random locations within each sampling site using a wading rod and a Hach ® flow meter (Model FH950.0, Loveland, CO, USA). All other variables, with the exception of wetted channel width (measured along six transects with a measuring tape), were measured in each 50 m stretch at five cross-sectional transects separated 10 m apart. Dissolved oxygen (mg/L and % saturation), temperature ( • C), and pH was measured at the center of each transect using handheld meters (dissolved oxygen and temperature: YSI Professional Series ® , Model Pro2030, Yellow Springs Instruments, Yellow Springs, OH, USA, pH: EcoTestr pH2 ® , Eutech Instruments, Vernon Hills, IL, USA). Percent canopy cover was estimated once during each summer with measurements taken along each transect at the edge of each stream bank (left-and right-side) and directly in the center of the channel (n = 15 per 50 m stretch) using a GRS TM densitometer (Geographic Resource Solutions, Arcata, CA, USA). All habitat variables, with the exception of average substrate size and percent canopy cover, were measured once during June 2013 and four times per year throughout June-July (at two week intervals) in 2014 and 2016.

Fish Surveys
All fishes were properly handled and collected with recommendations from the American Association for Laboratory Animal Science and approval from Hiram College's Institutional Animal Care and Use Committee (IACUC protocol number 13-05 and 16-03). For field collection, an Ohio Department of Natural Resources collecting permit was obtained (Wild Animal Permit #17-128). Surveys took place prior to restoration during early July in 2013 in the upstream control site, the two sites within the old channel, and the three sites downstream of the reconfigured channel, while sampling occurred post-restoration in the upstream control site, the three sites within the reconfigured channel, and the three sites downstream of the reconfigured channel in early July in both 2014 and 2016. All sampling efforts took place during normal summer baseflow conditions. To survey fishes, two full passes of electrofishing were completed in each 50 m stretch using a backpack Electrofisher (LR-24 Electrofisher, Smith-Root TM , Vancouver, Washington, DC, USA) and two dip nets (3 mm mesh) while walking in an upstream zig-zag pattern. Prior to sampling, block nets (2 mm mesh) were stretched across the entire width of the stream both upstream and downstream to prevent movement of fishes into and out of sampling sites. All fishes collected were identified to species and measured for total length (TL, cm) using a standard fish measuring board. At the completion of both passes of electrofishing, all fishes were returned live to the center of the sampling site.

Data Analysis
Differences in habitat abiotic variables (i.e., depth, substrate grain size, and canopy cover), Shannon-Wiener diversity, and species richness were analyzed using separate ANCOVAs for each response variable with channel type and year included as explanatory variables. Pair-wise comparisons were performed using Tukey's honest significant difference (HSD). All analyses were performed in R 3.3.2 [27].
Constraints on fish community composition were examined using redundancy analysis (RDA) of a Hellinger distance matrix of fish communities at each sampling site and year (for a description of RDA, see [28,29]). Hellinger distance standardizes community composition data by column sum (the abundance of each taxa across all sites), reducing sensitivity of analyses to differences in scale and accounting for absences in taxa across sites [29][30][31]. The maximal RDA model, which included channel type, year, stream depth, stream width, stream velocity, proportion juvenile fish, site identification, canopy cover, dissolved oxygen, pH, and substrate grain size as constraining variables, was simplified to a parsimonious model using backward and forward stepwise model reduction with the step function in R 3.3.2 [27,32], producing a simplified model that included channel type, year, dissolved oxygen, and canopy cover as contributing constraints. Statistical significance of the reduced RDA model, the RDA axes, and constraints were evaluated using permutation ANOVA with 1000 permutations. RDA and associated analyses were performed using the "vegan" package in R 3.3.2 [27,33].
Individual fish were assigned an age class (juvenile or adult) according to records in Trautman, 1981 [34] and to one of the following tolerance categories: intolerant (includes fishes designated moderately intolerant), intermediate (fishes classified as being intermediate in tolerance were omitted from the regression), or tolerant (includes fishes designated as moderately tolerant) from Grabarkiewicz and Davis, 2008 [35]. Binomial logistic regressions were separately fitted to age class and fish tolerance responses, where juveniles and fish classified as tolerant were treated as "successes" in the analyses and year and channel type were used as predictors. An analysis of deviance, goodness-of-fit test was used to test for significance of the independent variables in the logistic regression models (i.e., channel type, year, and the interaction between them). Logistic regressions and associated analyses were performed with function glm in R 3.3.2 [27].

Canopy Cover, Depth, and Substrate Type
Regardless of year, canopy cover (ANOVA, F 2,312 = 54.55, p < 0.0001), water depth (ANOVA, F 2,897 = 88.48, p < 0.0001), and substrate size (ANOVA, F 2,2097 = 54.27, p < 0.0001) were higher in both the upstream control site and downstream sites than the new channel for all three parameters (all Tukey's HSD, p < 0.05; see Table 1 for means and dispersion). Percent canopy cover was, however, higher in the upstream control site than that in both the downstream sites and new channel sites, whereas water depth was greater in the downstream control sites compared to that in both the upstream control site and new channel sites. Substrate size was statistically similar between the upstream control site and downstream sites (Tukey's HSD > 0.05), and no differences were detected in canopy cover, depth, or substrate size across years (p > 0.08).

Effects on Fish Diversity and Community Composition
A total of 22 species were collected across all sites within Eagle Creek over the entire study (Supplementary Materials Table S1). The most common species captured across sites were Semotilus atromaculatus creek chub, Etheostoma flabellare fantail darter, Etheostoma nigrum Johnny darter, Notropis buccatus silverjaw minnow, Rhinichthys obtusus western blacknose dace, and Catostomus comersonii white sucker (Table S1). All common species were captured in the newly reconfigured channel one-year post-restoration with the exception of western blacknose dace, which was collected 3-years post-restoration only (i.e., 2016, Table S1). The channel type significantly affected both Shannon-Wiener diversity (H; ANOVA, F 3,13 = 8.66, p = 0.0020) and species richness (F 3,13 = 11.55, p = 0.0006), and median H and species richness were higher in the upstream control site and downstream sites than that in both new channel and old channel sites (Figure 3). Median H was also higher in new channel sites than that in old channel sites, but median species richness was similar across new and old channel sites (Figure 3). There was no significant effect of year (H: F 1,13 = 1.02, p = 0.3321; species richness: F 1,13 = 1.52, p = 0.2399) and no significant interaction effect between channel type and year for both H and species richness (H: F 2,13 = 0.26, p = 0.7744; species richness: F 1,13 = 3.08, p = 0.0801).  RDA axes 1, 2, and 3 were statistically significant and explained 49.7%, 27.6%, and 11.9% of the variance in fish community composition by site, respectively ( Table 2). Channel type, year, canopy cover, and dissolved oxygen were maintained in the reduced RDA model, and a permutation ANOVA on the reduced RDA model indicated that channel type and year were statistically significant factors impacting fish communities ( Table 2). In contrast, canopy cover and dissolved oxygen (DO) were not statistically significant (Table 2). Year was more strongly correlated with RDA axis 1 than to RDA axis 2 (Table 2, Figure 4).
Old channel site fish community composition differed from all other channel types, and fish communities in new channel sites did not overlap with upstream sites (Figure 4). Prior to restoration in 2013, fish community composition in sites downstream of the newly reconfigured channel were similar to the upstream control site; however, following restoration, fish communities in downstream sites became more similar to the new channel sites and diverged from the upstream site over time. In contrast to the change in fish community composition observed in downstream sites, upstream site fish communities remained relatively stable across time, as indicated by a relatively small shift along RDA axis 1 in upstream site communities. RDA axes 1, 2, and 3 were statistically significant and explained 49.7%, 27.6%, and 11.9% of the variance in fish community composition by site, respectively ( Table 2). Channel type, year, canopy cover, and dissolved oxygen were maintained in the reduced RDA model, and a permutation ANOVA on the reduced RDA model indicated that channel type and year were statistically significant factors impacting fish communities ( Table 2). In contrast, canopy cover and dissolved oxygen (DO) were not statistically significant (Table 2). Year was more strongly correlated with RDA axis 1 than to RDA axis 2 (Table 2, Figure 4).
Old channel site fish community composition differed from all other channel types, and fish communities in new channel sites did not overlap with upstream sites (Figure 4). Prior to restoration in 2013, fish community composition in sites downstream of the newly reconfigured channel were similar to the upstream control site; however, following restoration, fish communities in downstream sites became more similar to the new channel sites and diverged from the upstream site over time.
In contrast to the change in fish community composition observed in downstream sites, upstream site fish communities remained relatively stable across time, as indicated by a relatively small shift along RDA axis 1 in upstream site communities.

Effects on Juvenile and Tolerant Fish
The main effects of channel type and year, and the interaction between channel type and year each affected the probability of encountering of juvenile fish (Table 3). In the upstream control site, the probability of encountering juvenile fish remained similar across years and was low compared to both old channel and new channel sites ( Figure 5). In contrast, the probability of encountering juvenile fish in downstream sites was similar to that of the upstream control site in 2013, but increased

Effects on Juvenile and Tolerant Fish
The main effects of channel type and year, and the interaction between channel type and year each affected the probability of encountering of juvenile fish (Table 3). In the upstream control site, the probability of encountering juvenile fish remained similar across years and was low compared to both old channel and new channel sites ( Figure 5). In contrast, the probability of encountering juvenile fish in downstream sites was similar to that of the upstream control site in 2013, but increased in 2014 and 2016, matching the relatively high probability of encountering juvenile fish in the new channel sites ( Figure 5). in 2014 and 2016, matching the relatively high probability of encountering juvenile fish in the new channel sites ( Figure 5).  Of the six most common species captured across the study, only one is considered intolerant (fantail darter) while three are considered tolerant fishes (creek chub, western blacknose dace, and white sucker). Similar to the juvenile fish results, the main effects of channel type and year, and the interaction between channel type and year, were all statistically significant factors affecting the probability of encountering tolerant fish (Table 3). Although the probability of encountering tolerant fish was similar across sites in 2013 and remained similar in the upstream control site across all years, probabilities increased in downstream sites with year ( Figure 5). In new channel sites, the probability of encountering tolerant fish was lower than both the upstream control and downstream sites in 2014, Of the six most common species captured across the study, only one is considered intolerant (fantail darter) while three are considered tolerant fishes (creek chub, western blacknose dace, and white sucker). Similar to the juvenile fish results, the main effects of channel type and year, and the interaction between channel type and year, were all statistically significant factors affecting the probability of encountering tolerant fish (Table 3). Although the probability of encountering tolerant fish was similar across sites in 2013 and remained similar in the upstream control site across all years, probabilities increased in downstream sites with year ( Figure 5). In new channel sites, the probability of encountering tolerant fish was lower than both the upstream control and downstream sites in 2014, but increased and was higher than the upstream control site and similar to pre-restoration old channel sites by 2016 ( Figure 5).

Discussion
Overall, remeandering restoration practices in Eagle Creek had negative impacts on fish communities in the short-term (≤three years). Colonizing communities in the new channel were unable to recover to reflect upstream community composition and structure; further, restoration activity negatively impacted fish communities in sites downstream of restoration over time. Since fish community diversity, proportion of juveniles, and proportion of tolerant fishes remained relatively stable in the upstream control site throughout the study, the impact of restoration appears to be the main driver of observed changes in fish communities downstream of the newly constructed channel. Restoration techniques created a relatively homogenous habitat in the new channel that was shallow, dominated by fine sediments, and completely lacked canopy cover. As of year three (2016), deep pools and logjams, habitat features present and stable throughout other sections of Eagle Creek, had not formed in the new channel, and, although riffle formation was visible in a few locations, riffles were not stable features (shifting or were absent following minor rain events).
Although substantial differences in canopy cover, water depth, and substrate size were observed between the new channel and upstream and downstream sites, these and other habitat variables were not correlated directly with changes in fish community composition. However, other habitat variables not measured in this study, such as presence of coarse woody debris, riffle-pool sequences, and embeddedness, may be important drivers of fish communities. Further, habitat variables may have influenced other biotic interactions (e.g., filamentous algae blooms, predator-prey interactions, competitive interactions, macroinvertebrate assemblages, and food availability). For example, loss of canopy cover is not uncommon with channel modification restoration practices (this study, [11,36,37]) and can lead to increased in-stream temperatures and primary productivity [11,38]; all factors that could impact fish communities. Extensive filamentous algae blooms were observed in this system in the new channel and downstream areas (to a lesser degree) following restoration [39] and may have influenced fish community composition and structure. In other systems, even when channel modification shows improvement of habitat variables, substantial changes in biodiversity generally do not occur [3,5,40].
Fish responses in the new channel reflected community composition that was substantially different from pre-restoration communities throughout all of Eagle Creek, and restoration efforts had negative consequences on fish community diversity. In fact, Shannon-Wiener diversity and species richness were substantially lower in the new channel and did not reach pre-or post-restoration levels found in the upstream control and downstream sites. However, Shannon-Wiener diversity in the new channel was greater than pre-restoration levels in the old channel, while species richness was similar; thus, diversity may be approaching levels found in the old channel. Although community composition and structure did not recover in the new channel, all six of the most common species in this system (creek chub, fantail darter, Johnny darter, silverjaw minnow, western blacknose dace, and white sucker) were able to colonize by year-three post-restoration. Of these six species, all but western blacknose dace colonized one-year post-restoration. Since fishes are generally good dispersers [41] and source populations are present upstream of this study site (unpublished data), rapid colonization by mostly tolerant and intermediate fishes was expected in the short-term. Such rapid colonization (within one-year) is common in other restored reaches [10,42,43] and is generally dependent on the availability [42,43] and proximity [43] of source populations with habitat variables having less of an effect [42].
In other stream systems, channel modification restoration practices appear to have inconsistent impacts on fish communities in the short-term, improving in some streams [13,44], showing no response in others, [10,13,45,46], or reflecting negative shifts in the fish community [10,13]. Even when fish community composition and structure shows rapid improvement (within one year), responses can be short-lived, and patterns of recovery can reverse within the short timeframe (three-five years) [47]. The downstream impacts of channel reconfiguration on fish communities are largely unknown, because studies addressing impacts of restoration generally take a before-after control-impact (BACI) design, which compares restored reaches and control reaches that have not been manipulated (but are considered to have had similar disturbances (e.g., channelization) prior to channel restoration). In this study, we reveal that downstream fish community composition shifted post-restoration, overlapping the composition found in new channel sites. However, neither Shannon-Wiener diversity nor species richness of downstream sites were negatively affected by restoration and remained similar to that of the upstream control site and high in comparison to that of the new channel sites. The shift in community composition evident from ordination, but not in other diversity metrics, is likely due to silverjaw minnows and western blacknose dace becoming more common post-restoration. However, our results contradict those found by Schwartz and Herricks reporting slight positive impacts on fish diversity and community composition downstream of restored riffle-pool structures two years' post-restoration [19].
Observed changes in community structure included increased capture of juvenile fishes post-restoration in both the new channel and downstream sites compared to that of the old channel and upstream control sites. These results suggest that the impacts of restoration either support habitats that promote the recruitment of juveniles or had negative consequences for adult fish colonization in the new channel and persistence in downstream areas. High proportions of juveniles in the new channel are likely explained by shallower water present in this stream section, whereas increases in downstream areas are more likely tied to disturbance from restoration, degrading adult persistence. Since stream depth remained high and deep pools/logjams remained intact in downstream areas, changes in age class structure are likely caused by other factors tied to restoration practices (e.g., sedimentation, degradation of upstream adult source populations in the new channel). Restoration techniques that create shallow margins can increase prevalence of juvenile [48][49][50] and small-bodied fishes (especially cyprinids) and appear to be important nursery zones [48]. Shallow margins were common in the new channel and colonizing communities were dominated by both juveniles and small-bodied fishes (e.g., cyprinids and darters) post-restoration, and deep-water fishes (e.g., centrarchids) were rarely captured. Increases in the proportion of juvenile fish may also be explained by (1) limitations in movement of juveniles throughout the new channel and downstream areas as young-of year fishes typically have limited swimming capabilities [51] or (2) increases in filamentous algal cover in this system. Anecdotally, filamentous algae became visually abundant post-restoration, while it was not evident pre-restoration, covering large portions of the new channel and some sections (to a lesser degree) of downstream areas with no detectable change in the upstream control site. Although not directly assessed in this study, an increase in algal cover could promote recruitment of juvenile fishes by providing refugia from predators or supporting increased macroinvertebrate abundance (food resource).
Further changes in community structure are reflected in increased capture of tolerant fishes in both the new channel and downstream sites from year one to three post-restoration, suggesting that new channel fish communities have not recovered and are negatively impacting downstream communities. While the proportion of tolerant fishes in new channel sites was lower compared to upstream control and downstream sites one-year post-restoration, by three years' post-restoration, the proportion of tolerant fishes exceeded levels found in both the upstream control and in that of the old channel sites, and the proportion of tolerant fishes continued to climb in downstream sites. Impacts of re-meandering had similar consequences in an Indiana stream, where shifts from communities dominated by intolerant fishes to a greater percentage of tolerant fishes were observed after restoration efforts [10]. Further, in other systems, even when in-stream habitat structure was improved (i.e., creation of riffle-pool sequences), tolerant species persisted post-restoration, and intolerant cyprinids and darters were absent from restored reaches [19]. In our study system, fantail darters (an intolerant fish) remained common in the upstream control site but became rare or absent by year three post-restoration in the downstream sites and were common in one site but rare in the other two. Further, mottled sculpin (another intolerant fish) never colonized the new channel and was absent from downstream areas by year-three, but again persisted in the upstream control site. Mottled sculpin appear to be especially sensitive to disturbance and can take up to six years to recolonize following channel reconfiguration [52].
Although the goals of natural channel design restoration are to achieve bank and channel stability, biodiversity is expected to improve and ecological function is expected to be restored (i.e., the "Field of Dreams" hypothesis, [53,54]). Unfortunately, this restoration technique does not consider restoration of stream chemistry or biological processes [3,55], and the short-term consequences on biota are largely unclear [5]. Meta-analysis of 91 stream restoration projects suggests that techniques utilizing channel reconstruction and improvement of the riparian zone were less effective in improving fish communities than restoration techniques that use in-stream structures to improve habitat [15]; however, the majority of restoration projects analyzed reported short-term effects, and it has been suggested that a minimum of 10 years may be necessary to evaluate the success or failure of stream restoration projects [56]. Regardless of the short-term impacts within a system, long-term monitoring can show improvement in some fish communities (12-years [57], 17 years [52]).
In addition to the importance of restoration technique to fish community recovery, length of the restored section can impact successful recovery [57,58]. Small, reach-scale restoration projects are likely to be unsuccessful when systems are impacted by heavy anthropogenic land use at the watershed-level [5,59,60], and underlying perturbations are likely to outweigh the benefits of restoration efforts if they do not target the major sources of degradation. The watershed surrounding Eagle Creek (this study) is impacted by row-crop agriculture and residential land-use, and a~8 m tall dam is located approximately 1.5 km downstream, possibly impeding fish movement within this stream system. These watershed-level impacts may ultimately limit or outweigh the future benefits of restoration in this system. Further, bank erosion is common throughout Eagle Creek and sedimentation can limit the recovery of fish communities in restored systems [10]. Since positive responses of the fish community in restored streams may take longer than the timeframe evaluated in this study, long-term monitoring in this system is necessary to evaluate restoration success or failure.