Introduction

A key question in the field of food web ecology is how spatial environmental heterogeneity drives diverse food resources and food web structures within a limited space (Wissel & Fry, 2005; Leigh & Sheldon, 2009; Sereda et al., 2012; Kaymak et al., 2018). Riverine reach is a typical template that has large spatial variations in geomorphologic and hydrologic conditions among habitat patches (e.g., riffle, pool, backwater and pond) (Yarnell et al., 2006; Heino, 2013). These morphological and hydrological variations may influence the patterns of material and energy flow among habitats within reaches, sustaining spatial heterogeneity in the abundance and composition of food sources (Wanner et al., 2002). Assemblage trophic structure varies spatially in association with longitudinal and lateral gradients of geomorphology, environmental conditions, and disturbance regimes in rivers (Hoeinghaus et al., 2007; East et al., 2017). Some habitats include highly stored terrestrial detritus and epilithon in backwaters and ponds on sandy bars (Nakajima et al., 2006; Flinn et al., 2008), abundant coarse particulate organic matter deposited in riparian ponds (Langhans et al., 2013), high biomass of phytoplankton in water bodies of medium age (2 to 10 days) (Hein et al., 2003), and high biomass of periphyton in lentic habitats (Biggs & Close, 1989). However, little is known about their ecological consequences for food web structure and energy flow at the community’s consumer levels.

Dam reservoirs may artificially alter the spatial distribution of food sources and community food web structure at dam outlet reaches (Hoffsten, 1999; Doi et al., 2008; Helmus et al., 2013; Wellard Kelly et al., 2013; Martinez et al., 2013; Four et al., 2019; Murphy et al., 2019). A typical phenomenon often observed below dams is an increased supply of suspended fine particulate organic matter (SFPOM) as a consequence of high primary production in reservoirs, which is especially notable in eutrophic lakes (Voelz & Ward, 1996). The high load of SFPOM sometimes leads to a critical shift of the main trophic origins of stream ecosystems from natural sources (e.g., attached algae and terrestrial particulate organic matter) to reservoir-derived plankton and also a shift of the functional feeding groups (FFGs) composition of macroinvertebrates to a structure more dominated by filter-feeders (Sheldon & Oswood, 1997; Kobayashi et al., 2011).

Although the trophic influence of dam plankton on downstream consumers has been widely tested and validated, previous tests were conducted only in lotic habitats or riffles (Watanabe & Omura, 2007; Doi et al., 2008; Tagliaferro et al., 2013). To our knowledge, the trophic effect of dams on lentic habitats, such as backwaters and isolated pond (IP) on sandy bars or on riparian terraces, has not been tested (Malard et al., 2002; Takemon, 2007). Therefore, supplementary tests that compare different habitat types in dam outlet reaches may provide better insights into the role of spatial environmental heterogeneity in the formation of spatially varying trophic structures.

Carbon and nitrogen stable isotope analysis is a widely used approach to estimate trophic contributions of different potential food sources to aquatic animals (Phillips et al., 2005; Leberfinger et al., 2011), assuming a monotonic change of isotopic properties through the anabolic process (Rounick et al., 1982; Reid et al., 2008). Autochthonous (e.g., epilithon) and allochthonous (e.g., terrestrial litter) organic matters are the two main sources of trophic contributions in river ecosystems. They have different isotopic signatures due to their different photosynthetic mechanisms and activities (Finlay et al., 2002). In dam outlet reaches, lentic plankton produced in reservoirs could be an additional food source. Dam plankton tends to show distinct isotopic signatures [i.e., low carbon stable isotope ratio (− 28‰) and high nitrogen stable isotope ratio (12‰)] compared with the in situ sources produced in rivers (Ock & Takemon, 2014). Based on the reservoir-specific isotopic properties of plankton, several isotopic studies found high trophic contributions of drifting dam plankton to downstream macroinvertebrate communities (Richardson & Mackay, 1991; Monaghan et al., 2001; Watanabe & Omura, 2007; Doi et al., 2008; Mercado-Silva et al., 2009). However, these previous tests were conducted only in riffles, and spatial heterogeneity among different habitat types was out of their focus (but see Ock & Takemon (2014) for the difference between riffle and pool).

In this study, using a carbon and nitrogen stable isotope mixing model (Finlay et al., 2002; Reid et al., 2008; Molina et al., 2011; Kominoski et al., 2012), we evaluated the trophic contributions of reservoir-derived plankton, epilithon and leaves to stream macroinvertebrates at four dam outlet reaches in the Natori River catchment, northeastern Japan, along with two reference reaches. The effects of dams on communities and habitat components along the Natori River have been described in previous studies (Watanabe & Omura, 2007; Takahashi et al., 2012). The number of filter-feeders feeding on particulate organic matter increased downstream of the dam. Lentic habitat areas were decreased compared to upstream. The reference rivers located upstream from the dams are near-natural rivers surrounded by forests and agricultural lands. The dams along this river are the primary multipurpose dams in this watershed and are appropriate for assessing the impact of dams on downstream habitats. We compared the trophic contributions of reservoir-derived plankton among four different habitat types co-occurring within the reaches. Our hypothesis was that the differences among habitats in the composition of food sources induce spatially heterogeneous patterns in the trophic contributions of reservoir-derived plankton. The trophic effect of dam plankton on macroinvertebrates in lentic habitats may be mitigated by low rates of migration of dam plankton and/or accumulation of allochthonous and autochthonous organic matter of stream origin.

Materials and methods

Study sites

Field surveys and sample collection were conducted in six reaches in two basins in the Natori River catchment in Miyagi Prefecture, northeastern Japan (Fig. 1). Each basin has one dam in the upstream section, the Kamafusa (K) and Ohkura (O) dams. Four reaches were selected from the downstream (D) reaches of dams (KD1, KD2, KD3 and OD), and two reaches were selected from upstream (U) reaches of the same dams as references (KU and OU). The three reaches below the Kamafusa dam are longitudinally located along a corridor with different water distances of 0.6 km (KD1), 2.7 km (KD2) and 6.3 km (KD3) from the dam. The Ohkura dam has one downstream study reach (OD) located 2.0 km below the dam. Three sampling sites were selected in the lower reaches of the Kamafusa Dam, between the dam and confluence of tributaries for easy access to the riverside. One sampling site was selected in the lower reach of the Ohkura dam, where access to the riverside was difficult due to the surrounding valley. In the section under study, there was no large confluence of a tributary with the main channels. Field surveys were conducted during low flow in each river (KD, 2.3–2.4 m3 s−1; OD, 7.4–10.1 m3 s−1). At the Kamafusa dam, there was a high flow in June (12.2 m3 s−1) due to the typhoon season, and the flow rate was maintained at 3.5 ± 0.7 m3 s−1 during the other months. Below the Okura dam, a high flow was observed in spring and summer (14.3 and 12.9 m3 s−1), and the flow rate was maintained at 9.4 ± 2.4 m3 s−1 during the other months. Although there is a small mountain runoff flow into the downstream reaches of the dams, most of the water in the study reaches is discharged from the dam reservoirs. The primary land-use types around the six study reaches are agricultural areas, forest, and wasteland, and the channels have a completely open canopy. Only OU located in the forest area and the river banks were relatively covered by canopy but open above the channel. The Kamafusa and Ohkura dams have 45.5 and 82.0 m vertical and 177.0 and 323.0 m wide lengths and water storage capacities of 3.9 × 107 and 2.5 × 107 m3, respectively. High concentrations of phytoplankton are often observed in both dam reservoirs, especially during early summer (June and July).

Fig. 1
figure 1

Maps of the study area depicting land-use (a) and habitat maps at each site (b). KU Kamafusa dam upstream; KD1–KD3 Kamafusa dam downstream 1–3; OU Ohkura dam upstream; OD Ohkura dam downstream. The insets show the location of the Natori River catchment (in a square) within Japan. Habitat maps were classified into four habitat types: lotic habitat, bar-head lentic habitat, bar-tail lentic habitat, and isolated pond

Full size image

We investigated spatial habitat structure in the six reaches once for each reach at KU and KD2 on 2 July 2008 and at KD1, KD3, OU and OD on 2 October 2008. We identified up to four habitat types for each reach: lotic (riffle and pool), bar-head (BH) lentic habitat, bar-tail (BT) lentic habitat (backwater) and isolated pond (IP) on sandy bars using a high-precision GPS (ProMark3, Thales, France). Geographic positions were recorded at 2-s intervals while walking along contours of the habitat and later post-corrected using base station data to obtain a precision of 0.1 m or less. Water surface area, frequency and shape complexity of each habitat type in a reach were calculated using ImageJ v.1.42 software (NIH, Bethesda, MD, USA).

We collected quantitative samples of macroinvertebrate communities from each habitat type found in the reaches using a Surber sampler (30 × 30 cm, 250-μm mesh) with three random replicates per habitat type per reach once during 20 October and 18 November 2008. The samples were preserved in 99.5% ethanol and separated from detritus and sediment debris in the laboratory under a microscope with × 150 magnification for identifying organisms. Organisms were identified to the lowest taxonomic level possible (mostly at the species level) based on Kawai & Tanida (2005) and were assigned to one of five FFGs (filter-feeders, collector-gatherers, scrapers, shredders and predators) and six life types (LTs) (net spinners, crawlers, attachers, burrowers, swimmers, gliders) based on the classification of Takemon (2005) modified from Merritt and Cummins (1996) for Japanese freshwater macroinvertebrates. Macroinvertebrate samples for stable isotope analysis were collected qualitatively at each habitat, and collected samples were immediately cryopreserved after identification. In the field, we also measured the water velocity and depth in each habitat type using a current meter (VP-201, Kenek, Tokyo, Japan) and a ruler, respectively, with five replicates per habitat type per reach. Water quality parameters (pH, EC, DO, and water temperature) were measured with a multiparameter water quality checker (U-50, HORIBA, Japan). In the laboratory, water samples were analyzed for total nitrogen (T-N) and total phosphorous (T-P) concentrations using an auto-analyzer (AA3, Technicon, USA).

Epilithon (dominated by attached algae), SFPOM (< 1.0 mm), benthic fine particulate organic matter (BFPOM), benthic coarse particulate organic matter (> 1.0 mm) (BCPOM) ,and leaves were collected from three randomly selected locations per habitat type per reach on the same day of collection of macroinvertebrates. Epilithon was collected from stones using a toothbrush, washed in pure water, and filtered onto precombusted Whatman GF/F glass filters (0.7-µm nominal pore size). SFPOM was collected by filtering 1 to 4 l of surface river water through a 1.0-mm sieve onto a GF/F filter. BFPOM was collected by placing a plastic tube sampler (diameter, 25 cm; depth, 50 cm) on the riverbed, disturbing the benthic material in the sampler by hand, collecting the turbid water and processing the sample by the same pretreatment as for SFPOM. Leaves were collected from several different plant species in each reach. BCPOM was collected using Surber nets together with the macroinvertebrates. Plankton was collected from the reservoirs at three levels (0, 3, and 10 m from the water surface) near the water intakes of the dams and processed by the same pretreatment as for SFPOM in river water. Finally, nine replicates of the epilithon, SFPOM, BFPOM, BCPOM, and dam plankton samples were prepared for carbon (n = 3) and nitrogen (n = 3) stable isotope analyses and for ash-free dry mass (AFDM) analysis (n = 3) (Flinn 2008), and six replicates of leaf samples were prepared for carbon (n = 3) and nitrogen (n = 3) stable isotope analysis. All samples were stored at − 20 °C prior to further processing.

Stable isotope analysis

For stable isotope analysis, macroinvertebrates and potential food sources were acidified with 1 mol l−1 HCl to remove carbonate, and the remaining material was rinsed with distilled water and kept in a freezer (Walters et al., 2007). All samples were freeze-dried and homogenized before stable isotope analysis. The samples, ranging from 0.5 to 1.5 mg dry weight for macroinvertebrates and from 1 to 2 mg (exceptionally ~ 20 mg for BFPOM) dry weight for the potential food sources, were weighed into tin capsules. Carbon and nitrogen isotope ratios (13C/12C and 15N/14N) were measured using an elemental analyzer (NA2500, CE Instruments, USA) coupled to a continuous flow mass spectrometer (Finnigan MAT, Delta Plus, Thermo Fisher, USA). Stable isotope ratios were evaluated in δ notation as the deviation from standards (Pee Dee belemnite for δ13C and atmospheric nitrogen for δ15N), calculated as δ13C or δ15N = [ (Rsample/Rstandard) − 1] × 103, where R is 13C/12C or 15N/14N, respectively. Typical precision of the analyses was ± 0.5‰ for δ15N and ± 0.2‰ for δ13C.

A Bayesian mixing model of stable isotope analysis in R (SIAR) (Parnell et al., 2010) was used to calculate the relative contribution of each potential food source to the diet of macroinvertebrates. This Bayesian model allows the incorporation of uncertainty and variation in parameters such as trophic enrichment factors. We did not incorporate prior information into the mixture model because the prior information for the dietary proportions was vague. We ran this model using δ13C and δ15N of macroinvertebrates, SFPOM, and BFPOM found in four habitats of the dam outlet reach to assess intraspecific spatial variability in diets, a component of potential food source spatial variability. The potential food sources used in the three-source model in the dam outlet reaches were δ13C and δ15N of epilithon, leaves and reservoir-derived plankton. The carbon and nitrogen isotope enrichments in the model were set to + 0.4‰ (McCutchan et al., 2003) and 3.4‰ (Post, 2002), respectively. The SIAR mixing model was run for 500,000 iterations, discarding the first 50,000 samples. Before performing the model calculations, we confirmed significant differences in mean isotopic values among the three potential food sources using one-way analysis of variance (ANOVA). In the two upstream reaches of dams (KU and OU), we alternatively used a two-source mixing model (Peterson & Fry, 1987; Molina et al., 2011) to determine the relative contribution of epilithon to the diet of consumers (f) against leaves: f = [ (δ13Cconsumer − δ13Cleaves)/ (δ13Cepilithon − δ13Cleaves)] × 100, where δ13Cconsumer is the δ13C of each consumer taxon, and δ13Cepilithon and δ13Cleaves are the δ13C values of epilithon and terrestrial plants. When f was less than 0 in the calculation, 0 was used as f.

Statistical analysis

The mean levels of hydraulic variables (depth and current velocity), amounts of the potential food sources (epilithon, SFPOM, BFPOM, and BCPOM), taxon richness, total abundance and biomass of macroinvertebrates, and percentages of trophic contributions of potential food sources (dam plankton, epilithon, and leaves) in SFPOM, BFPOM, and macroinvertebrates were compared among the four habitat types (lotic habitat, BH, BT, and IP) using one-way ANOVA followed by the Tukey–Kramer test for multiple pairwise comparisons. Two-way ANOVA was performed using habitat types and FFGs as factors for the relative biomass of macroinvertebrates and trophic contributions of reservoir-derived plankton to macroinvertebrates. The interaction term was used to assess whether trophic responses for FFGs occurred between habitat types. To determine whether the trophic contribution of macroinvertebrates was structured according to SFPOM composition among habitats, we tested the correlation between the proportion of reservoir-derived plankton in SFPOM and the trophic contribution of each LT and FFG (Pearson correlation). ANOVA, Tukey–Kramer test, and correlation test were performed using R version 3.4.4 (R Core Team 2021). For each habitat, we visualized the multivariate patterns using multidimensional scaling (nMDS) ordination analysis on the basis of a Morisita–Horn dissimilarity matrix. The index focuses on species-by-species abundance comparisons and is not strongly sensitive to species richness and sample sizes (Chao et al., 2006). From visualized nMDS patterns, we established groups for an ANOSIM (analysis of similarities) with the Morisita–Horn index using the R vegan package.

To examine the spatial distribution of the effects of reservoir-derived plankton on species in greater detail, we constructed a cluster map of the trophic contribution of reservoir-derived plankton for each LT and FFG across all sampling sites using the heatmap clustering command from the R pheatmap package. Information about the distance from the dam and the distribution of flow velocity at each habitat also was included.

Results

Habitat structure, hydraulic, and food source conditions

Lotic and bar-head (BH) lentic habitats commonly occurred throughout the six study reaches. Bar-tail (BT) lentic habitat occurred at Kamafusa dam upstream (KU), Kamafusa dam downstream 3 (KD3), and Ohkura dam downstream (OD) and isolated pond (IP) occurred at KD1 and KD3. IP occurred only at dam outlet reaches (Fig. 1). Throughout the six reaches, lotic habitat always had the highest proportion of water area in the reaches (mean, 93.8%; range, 88.5–96.9%), while lentic habitats (BH, BT, and IP) accounted for a minor proportion of water area (mean 3.4%; range, 1.4–11.5%). There was no significant difference in lotic habitat area and lentic habitat area between the upstream and downstream reaches of the dams.

Mean water depth (F3,166 = 24.7, P < 0.001) and mean current velocity (F3,164 = 105.5, P < 0.001) (ANOVA) differed significantly among the four habitat types in all sites. Multiple pairwise comparison tests found significantly shallower water depths in the BH (mean = 0.13 m) than in the lotic habitat (0.32 m) upstream, and the deepest water depth was observed in the BT (0.46 m) downstream (P < 0.001, Tukey–Kramer test) (Table 1). The mean current velocity was significantly higher in the lotic habitat (0.51 m s−1) than in the three lentic habitat types (BH, 0.02 m s−1; BT, 0.01 m s−1; IP, 0.00 m s−1; P < 0.001, Tukey–Kramer test). Current velocity was also significantly higher in lotic habitats upstream and downstream. Although there were no significant differences in pH, EC, and water temperature among habitats in all sites, dissolved oxygen was lower in IP than in other downstream habitats (Table 1). Nutrients were not significantly different among habitats, and T-N in the reservoirs was relatively high (Table 1).

Table 1 Mean (SD) of environmental variables, potential food sources, and macroinvertebrate community structure in each habitat type among upstream and downstream reaches and reservoirs
Full size table

Differences among habitats in the availability of food sources were also evident. The mean biomass (AFDM) of potential food sources was significantly different among the four habitat types (SFPOM, F3,91 = 7.554, P < 0.001; BFPOM, F3,82 = 19.421, P < 0.001; BCPOM, F3,58 = 11.717, P < 0.001), except for epilithon (F3,73 = 2.241, P > 0.05) in the lower reaches. Only BCPOM biomass was significantly different among upstream habitats (F2,12 = 9.958, P < 0.01) (Table 1). There were no significant differences in the mean biomass of each organic matter type in each habitat between upstream reaches that differed in canopy cover (respectively, P > 0.05, t-test). Multiple pairwise comparison tests found significantly larger mean biomasses of SFPOM, BFPOM, and BCPOM in the three lentic habitat types than in the lotic habitat (P < 0.05, Tukey–Kramer test). Downstream, SFPOM biomass was higher in IP and BT than in the lotic habitat (P < 0.01, Tukey–Kramer test), and BH was in between (P > 0.05, Tukey–Kramer test). BFPOM biomass was highest in the BT (P < 0.001, Tukey–Kramer test), and BCPOM biomass was highest in the IP (P < 0.01, Tukey–Kramer test). Three habitat types, but not the IP, showed significant increases in mean SFPOM biomass from the upstream to the downstream of the dam (lotic, from 1.2 to 5.8 mg l−1; BH, from 2.0 to 16.8 mg l−1; BT, from 2.4 to 35.5 mg l−1; P < 0.05, t-test). BFPOM biomass increased only in the BT (from 37.9 to 154.2 mg l−1, P < 0.01), and epilithon and BCPOM biomass did not change between the upstream and downstream of the dam.

Macroinvertebrate community

We collected a total of 5547 macroinvertebrates and identified 125 taxa from the 17 habitats among the six reaches. The majority of taxa were of the orders Trichoptera (40 taxa), Ephemeroptera (40 taxa), Plecoptera (11 taxa), and Diptera (11 taxa). Taxon richness ranged widely from 11 to 42 taxa among the 17 habitats, with significant differences between the lotic (mean = 32.7) and lentic habitat types in all sites (BH, 20.1; BT, 15.0; IP, 18.5; P < 0.05, Tukey–Kramer test). Taxon richness was not significantly different between upstream reaches and downstream reaches of each habitat (P > 0.05). Total abundance and total biomass were also higher in the lotic habitat than in the lentic habitat types, except for BH and BT at KU, where Chironomus sp. (70 and 921 individuals 0.27 m−2, respectively) were highly abundant, and IP at KB3, where Cloeon dipterum (Linnaeus, 1761) (967 individuals 0.27 m−2) was highly abundant. Mean abundance and biomass in lotic habitats were higher downstream than upstream, in contrast to those in BT (Table 1). Among the 125 taxa, 37 and 39 taxa occurred only in lentic or lotic habitats, respectively. Antocha sp. was commonly found throughout all 17 habitats with high abundance. The most frequently occurring taxa in the lotic habitats were Macrostemum radiatum (McLachlan, 1862), Cincticostella okumai (McLachlan, 1875), and Baetiella sp., whereas Hydropsyche setensis (Iwata, 1927) occurred throughout all six lotic habitats.

Habitat heterogeneity resulted in the uneven distribution of FFGs and LTs. Downstream, the mean proportion biomass of filter-feeders [i.e., Stenopsyche marmorata (Navas, 1920), Hydropsyche orientalis (Martynov, 1934), Hydropsyche albicephala (Tanida, 1986), Ephemera strigata (Eaton, 1892), and Antocha sp.] dominated in lotic habitats, and the scrapers, predators, collector-gatherers, and shredders, in that order, had progressively smaller populations (Table 1). Both BH and BT, at the upper and lower sides of the sand bar, respectively, showed similar relative abundance for all functional groups except for shredders. By contrast, IPs distant from a core stream primarily hosted collector-gatherers and scrapers. The proportion of FFGs in upstream reaches was different from that in downstream reaches, as scrapers and predators were dominant in lotic habitats (Table 1). For LTs of macroinvertebrates, relative biomass differed significantly among the habitats in all sites (F = 12,65 2.04; P < 0.001). Net spinners comprised the highest population proportion in the downstream lotic habitat (86.2%), whereas crawlers and gliders comprised the highest proportion upstream. There was no significant difference in the other types among habitats at both reaches, but the crawlers were relatively common among them. The interactive effect of habitat heterogeneity and LTs on relative abundance was attributable to the marked increase in specific species in the habitat distribution of a habitat scale.

The ordination of macroinvertebrate community structures each habitat by NMDS showed two different community compositions (lotic habitat group and lentic habitat group) from overlap of confidence interval ellipses (Fig. 2). The composition of the macroinvertebrate communities differed among the four locations (lotic habitat, lentic habitat, OU, IP at KD3), as indicated by the significant ANOSIM R statistics (Global R = 0.76, P < 0.01). Macroinvertebrate communities from lotic habitats, except for OU, differed from those in lentic habitat groups. The community structure of the lotic habitat of OU was different from that downstream, whereas that of the lotic habitat of KU and KD1–3 was similar, although the similarity decreased from upstream to downstream. The community structure of BH in both rivers was also similar to the longitudinal changes in the community structure of lotic habitats.

Fig. 2
figure 2

Multidimensional scaling (MDS) ordination analysis of similarity (ANOSIM) results of macroinvertebrate assemblages in each habitat at the study sites. The gray polygons represent the classification based on ANOSIM results. The dashed circles represent the 95% confidence intervals for lotic habitat group and lentic habitat group. Lotic lotic habitat; BH bar-head lentic habitat; BT bar-tail lentic habitat; IP isolated pond; KU Kamafusa dam upstream; KD1–KD3 Kamafusa dam downstream 1–3; OU Ohkura dam upstream; OD Ohkura dam downstream. Closed symbols represent upstream habitats; open symbols represent downstream habitats. Four symbols for each site correspond to four habitat types

Full size image

In dam outlet reaches, the mean proportional abundance of filter-feeders in the macroinvertebrate community was significantly higher in lotic habitats than in lentic habitats (F12,40 = 3.088; P < 0.01, two-way ANOVA; P < 0.01, Tukey–Kramer test), whereas this was not true for the other FFGs (P > 0.05). The spatial variation in the proportion of filter-feeders among the six lotic habitats was significantly and positively correlated with the variation in SFPOM concentration (r = 0.954, P < 0.01). We did not detect any other significant correlations between the proportion of FFGs and the biomass of SFPOM, BFPOM ,and BCPOM in any habitat type.

Carbon and nitrogen isotope signatures

The carbon (δ13C) and nitrogen (δ15N) isotope signatures varied among potential food sources in all habitat types within reaches (P < 0.01, ANOVA) (Supplementary Material Appendix 1, Fig. S1). Epilithon had the highest δ13C (− 22.4 ± 3.4‰) and δ15N (4.1 ± 2.3‰) values, while leaves had the lowest δ13C (− 29.8 ± 1.4‰) and δ15N (0.5 ± 0.2‰) values in each habitat. Dam plankton collected from the reservoirs had low δ13C (− 28.2 ± 0.5‰) and high δ15N (4.9 ± 0.7‰) values. In the dam outlet reaches, δ13C and δ15N values for SFPOM and BFPOM fell within the range of values for epilithon, leaves and dam plankton, allowing the three-source mixing model to be run. All three lentic habitat types had significantly lower mean proportions of dam plankton in SFPOM than the lotic habitat (P < 0.05, Tukey–Kramer test) (Fig. 3d). On the other hand, the mean proportion of epilithon in SFPOM was higher in the BH and IP than in the lotic habitat (P < 0.01, Tukey–Kramer test). BFPOM had higher proportions of dam plankton only in the IP (P < 0.05, Tukey–Kramer test) (Fig. 3e). In upstream reaches of dams, there were no significant differences among habitat types in the proportion of the two sources (epilithon and leaves) in SFPOM and BFPOM (Fig. 3a and b).

Fig. 3
figure 3

source models. Lotic lotic habitat; BH bar-head lentic habitat; BT bar-tail lentic habitat; IP isolated pond. In each box plot, the horizontal line crossing the box is the median, the cross indicates the mean, 25–75 percentile range (box) and 5–95 percentile range (whiskers). Outliers are shown by circles above and below whiskers

Comparison of mean trophic contributions of potential food sources in suspended fine particulate organic matter (a) and (d), benthic fine particulate organic matter (b) and (e), and the macroinvertebrate community (c) and (f) among four habitat types, calculated using two- or three-

Full size image

For stable isotope analysis of macroinvertebrates, we selected 91 taxa with > 0.2 mg dry mass. The δ13C and δ15N values of macroinvertebrates fell within the habitat-specific ranges among the three potential food sources (epilithon, leaves, and dam plankton) in the dam outlet reaches (Supplementary Material Appendix 1, Fig. S1). The trophic contributions of dam plankton to the macroinvertebrate community estimated by the three-source mixing model were significantly different among the four habitat types (F3,397 = 23.462, P < 0.001) and among FFGs (F4,396 = 35.695, P < 0.001) (Fig. 3f). Lotic habitats had a significantly higher contribution from dam plankton (mean, 25.8%) than did the BH (16.4%), BT (7.6%) and IP (12.5%) (P < 0.001). The contribution of leaves in the lotic habitat was significantly higher than that in the BH and IP in the dam outlet reaches (P < 0.05). Epilithon was estimated to be a principal food source throughout the habitat types downstream from dams. In the upper reaches of the dam, the trophic contribution of epilithon was dominant (88.2–97.1%), regardless of habitat type (Fig. 3c).

In dam outlet reaches, filter-feeders also consumed more dam plankton than the other FFGs (filter-feeders, 37.5 ± 25.2%; collector-gatherers, 15.2 ± 13.3%; scrapers, 10.9 ± 13.3%; shredders, 5.4 ± 7.0%; predators, 15.0 ± 11.4%, Tukey–Kramer test, P < 0.001) (Supplementary Material Appendix 1, Table S1). The epilithon contribution of collector-gatherers and scrapers was significantly higher than that of filter-feeders and shredders, and the leaves contribution of predators was higher than that of filter-feeders, collector-gatherers, and scrapers. On the other hand, all FFGs utilized epilithon as a predominant food resource throughout upstream of the dams (Supplementary Material Appendix 1, Table S1). Regardless of habitat type, filter-feeders were affected by dam plankton in the dam outlet reaches; particularly, the magnitude of this effect was much greater in trophic contributions at lotic and IP (Fig. 4). Additionally, the variability of the trophic contribution was also larger in filter-feeders. No significant differences arose when the contribution of dam-derived plankton was compared to filter-feeders among habitats (P > 0.05) or when scrapers were compared to shredders. However, collector-gatherers were significantly higher at lotic and BH habitats, and predators were significantly lower in BT (Tukey–Kramer test, P < 0.01). There were no significant differences between lotic and BH habitats for all FFGs.

Fig. 4
figure 4

Boxplot of trophic contributions of reservoir-derived plankton to each functional feeding group (FFG) among habitat types. In each box plot, the horizontal line crossing the box is the median, the cross indicates the mean, and circles represent individual macroinvertebrates. Outliers are shown by circles above whiskers. Superscripts a, b indicate statistically significant differences between the habitats (Tukey–Kramer test, P < 0.05). N.S. no significance

Full size image

Filter-feeders are classified into four LTs: net spinners (i.e., Hydropsyche orientalis, Hydropsyche setensis, Stenopsyche marmorata), burrowers [Ephemera strigata, Ephemera japonica (McLachlan, 1875), Ephemera orientalis (McLachlan, 1875), attachers (Molanna moesta (Banks, 1906)] and swimmers [Isonychia japonica (Ulmer, 1919)], and filter-feeders had a significantly higher trophic contribution of dam plankton than the other LTs (P < 0.01; t-test). For evaluating in more detail the effect of SFPOM on the trophic contribution of dam plankton to macroinvertebrates, species were classified according to LT and FFG, and the correlation between the proportion of dam plankton of SFPOM and the trophic contribution of dam plankton were measured (Fig. 5). Spatial variation was observed in the contribution of dam plankton to net spinners. Filter-feeders were significantly positively correlated with the proportion of dam plankton in SFPOM in a lotic habitat (r = 0.62, P < 0.01). However, there was no significant correlation between them in lentic habitats (P > 0.05) (Fig. 5a). Regardless of habitat type, other LTs of filter-feeders were positively correlated with the proportion of dam-derived plankton of SFPOM (r = 0.64, P < 0.01) (Fig. 5b). Collector-gatherers, feeding mainly on deposited FPOM, were positively correlated with all LTs and dam plankton in SFPOM (r = 0.31, P < 0.05) (Fig. 5c). There also was a significant positive correlation between the predators and dam plankton in SFPOM (r = 0.50, P < 0.001) (Fig. 5d).

Fig. 5
figure 5

Relationships between trophic contributions of reservoir-derived plankton to a net spinners-filter-feeders, b other filter-feeders, c collector-gatherers, d predators and percentage of reservoir-derived plankton in suspended fine particulate organic matter (SFPOM) each habitat. The dotted line indicates a significant correlation (a shows the correlation for only lotic habitat). BH bar-head lentic habitat; BT bar-tail lentic habitat; IP isolated pond; at attachers; bu burrowers; cr crawlers; ns net spinners; sw swimmers. White symbols, lotic habitat; light gray symbols, bar-head lentic habitat; dark gray symbols, bar-tail lentic habitat; black symbols, isolated pond; cross symbol indicate glider-predators in bar-tail lentic habitat

Full size image

To complement the spatial variation of the effect of a dam environment to macroinvertebrates, we sorted the trophic contribution of reservoir-derived plankton, which allows for more precise profiling. From these trophic contribution data, we constructed a heatmap to reveal the spatial variability of the contribution of dam plankton to each LT and FFG (Fig. 6). Except for attachers, filter-feeders were strongly affected by dam plankton in all habitats of KD1 and in the lotic habitats of other reaches, with the net spinners and filter-feeders being particularly strongly affected in KD1. Groups other than filter-feeders had a lower dam plankton contribution even within the lotic habitat. Most of the organisms living in lentic habitats are not affected by the dam effect. This result became more pronounced with increase in distance from the dam site.

Fig. 6
figure 6

Heatmap and dendrogram of the trophic contribution of dam-derived plankton to macroinvertebrate life types and functional feeding groups present in different sampling sites. Right side heatmaps present current velocity and distance from dam sites at each sampling site. KD1–3 Kamafusa dam downstream 1–3; OD Ohkura dam downstream; lotic lotic habitat; BH bar-head lentic habitat; BT bar-tail lentic habitat; IP isolated pond; at attachers; bu burrowers; cr crawlers; ns net spinners; sw swimmers; FF filter-feeders; CG collector-gatherers; SC scrapers; SH shredders; PR predators

Full size image

Discussion

Our approach comparing trophic structures among different habitats has provided useful insights into the role of spatial environmental heterogeneity in the formation of food web structures in rivers. Environmental heterogeneity was related to community diversity and food resources of macroinvertebrates on each spatial scale (Boyero, 2003; Boyero & Bosch, 2004; Zilli & Marchese, 2011; Pilotto et al., 2016). However, in previous studies, the relationship between heterogeneity and food webs in macroinvertebrates within habitats was not considered. The result of our investigation into the structure and composition (functional feeding groups and life types) of macroinvertebrate communities was significantly different between lotic habitat and lentic habitat. Regarding the lotic habitat groups except for those upstream of Ohkura Dam, macroinvertebrate communities were similar, although abundance and biomass of macroinvertebrates were higher in the group downstream of the dam than upstream, and the percentage of filter-feeders and net spinners were also higher downstream. On the other hand, the variation in macroinvertebrate community structures within lentic habitats were stronger than in lotic habitat, indicating a unique community structure in response to local environmental heterogeneity. In addition, changes in community structure due to the effects of dams were included in the inter-habitat variability and were less sensitive to dam effects than in lotic habitat. Therefore, it is necessary to examine the macroinvertebrate community structure of lotic habitats to assess the response to dam impacts. However, to evaluate the macroinvertebrate community structure of the whole habitat, we need to add a survey of lentic habitats; otherwise, we may underestimate the species diversity.

Using carbon and nitrogen stable isotope analyses, we compared the trophic influence on upstream and downstream macroinvertebrate communities between lentic and lotic habitats. In the reference reaches, most macroinvertebrates used epilithon as a food resource; however, there was no difference in food resource availability among habitats. These results suggest that the direct utilization of epilithon by species and the indirect utilization of FPOM produced by the decomposition of epilithon occur simultaneously in the rivers upstream of the two dams. However, downstream results showed habitat-specific patterns of trophic influence of dam plankton, with less influence of dam plankton in lentic habitat than in lotic habitats. On the other hand, other organic matter may also be utilized as potential food sources for macroinvertebrates. Although we did not collect and identify macrophytes at this study sites because they were absent or barely observed, for example, macrophytes are a potential food resource for macroinvertebrates and fish, and their stable isotope values (δ13C: − 15 to − 32‰, δ15N: 5 to 10‰) overlap with the potential food resource values in this study (Lloret & Marin, 2009; Dalu et al, 2016; Wolter et al., 2018). Therefore, the results of the trophic contribution of macroinvertebrates may contain some uncertainties. It will be necessary to add macrophytes to potential food sources when we analyze food webs in river ecosystems.

We considered the habitat-specific source composition of SFPOM as a potential driver of the reduced trophic impact in lentic habitats. The three-source model analysis of SFPOM indicated a lower proportion of reservoir-derived plankton and higher proportions of epilithon and leaves in lentic habitats than in lotic habitats. The lower proportion of reservoir-derived plankton in lentic SFPOM was most likely reflected in the lower trophic impact on lentic consumers through assimilation of SFPOM, especially on filter-feeders. The habitat-specific composition of SFPOM may be derived from hydraulic and landscape characteristics of lentic habitats. BH located at the leading edges of the bar was not only an area where surface water down-welled into the hyporheic zone but also an area where SFPOM from upstream was accumulated (Boulton et al., 2008). Hence, the contribution of dam plankton in the BH was higher than that in other lentic habitats, but the amounts of epilithon and leaves were higher than in the mainstream, resulting in an increase in river-derived SFPOM. It can be inferred that the impact of the dam was moderated. BT located at the downstream end of the bar area where groundwater upwells is an environment where mainstream water and SFPOM cannot easily enter. The biomass of organic matter in the BT is also higher than in the mainstream, and the interaction between hydraulic, morphological, and ecological processes may have contributed to mitigating dam effects. On the other hand, it is impossible for dam plankton to flow into the IP when it is disconnected from the mainstream during periods of normal water level. After a flood, dam plankton may be left behind in the IP. In addition, the amount of BFPOM in the IP may have increased as a result of plankton production in the water body (Doi, 2009). Therefore, it was suggested that the influence of dam-derived plankton in lentic habitats is due to different hydraulic, morphological, and ecological processes. Furthermore, SFPOM derived from allochthonous and autochthonous sources was produced by decomposition of bacteria and relatively dominant scrapers and shredders that inhabited the lentic habitat (Langhans et al., 2008; Corti et al., 2011; Treplin & Zimmer, 2012; Halvorson et al., 2015). The high amounts of epilithon, BFPOM, and BCPOM observed in lentic habitats can account for these hydraulic, morphological, and biological processes, leading to the higher proportions of epilithon and leaves in lentic SFPOM.

The trophic patterns that are the lower contribution of reservoir-derived plankton in lentic habitats were specifically notable for collector-gatherer, predators, and filter-feeders, except for net spinners. Although the trophic contribution of reservoir-derived plankton on SFPOM was reflected in the trophic contribution of collector-gatherers (Fig. 5c), the BFPOM component was hardly responsive to the trophic structure of collector-gatherers. Dam plankton continually flowing from the dam became trapped and settled out on the riverbed or top of the edge on a sand bar (BH) (Boulton et al., 2008). Hence, collector-gatherers may have assimilated SFPOM with a large percentage of dam plankton in the lotic habitat and BH, as dam planktons are deposited in the surface layer of BFPOM. By contrast, the relatively high reservoir-derived plankton contribution of BFPOM in IP was not reflected in the collector-gatherer trophic structure. Differences in the habitat environment affected the trophic structure of collector-gatherers through SFPOM dynamics. IP consisted of habitats created by a body of water left behind after flooding on sandbars and spring waters, and dam plankton may not exist on the surface layer of BFPOM.

Although the high trophic influence of dam plankton and reservoir-derived FPOM on filter-feeders has been reported in numerous studies (Doi et al., 2008; Power et al., 2013; Tagliaferro et al., 2013; Four et al., 2019), these earlier observations were limited to lotic habitats, where the strongest impact was observed in this study. However, the mean trophic contribution of reservoir-derived plankton was not significantly different among habitats as values of each species were high variation. Net spinners-filter-feeders Hydropsyche albicephala (Tanida, 1986) and Hydropsyche nakaharai (Iwata, 1927) have a high contribution of reservoir-derived plankton at IP in KD1. The common net spinners, caddisfly (Hydropsychidae), inhabit a lotic habitat, riffle, and glide, and the IP is unsuitable for those species. Therefore, these Hydropsychidae may have migrated from a larger stream to an IP during flooding in study sites before sampling, and species were left behind in IP when the water level went down. Because the trophic contribution of filter-feeders to dam effects differed according to LT, it was necessary to evaluate the trophic structure of filter-feeders considering LTs. Directly below the dam, the effect of reservoir-derived plankton also had extended to a lentic habitat due to species migration from the mainstream (Fig. 6).

In addition to the habitat-specific composition of food sources, habitat preference of macroinvertebrates was considered as a potential reason for the reduced trophic contribution of dam plankton in lentic habitats. Filter-feeders generally prefer to live in lotic environments with rapid flow and loose stones and gravel of suitable size for their net-spinning behavior on the riverbed (Georgian & Thorp, 1992). Stabilization of the substrate due to reduced hydraulic and sediment dynamics resulting from dam control is also a driving factor for the abundance of filter-feeders at dam outlet reaches (Oswood, 1979; Hoffsten, 1999; Tszydel et al., 2009). In lentic habitats, the substrate is mainly composed of fine materials (silt and sand) and embedded stones with extremely low water flow. Therefore, the abundance ratio of filter-feeders in the macroinvertebrate community is reduced in lentic habitats, with increases in the ratios of other FFGs, such as collector-gatherers, scrapers and shredders. These other FFGs can feed on sources less mixed with reservoir-derived plankton (e.g., BCPOM and BFPOM) and thus are robust to the high input of dam plankton. As a result of high ratios of these FFGs being potentially robust to the input of dam plankton, the trophic impact on the macroinvertebrate community as a whole may be reduced.

Finally, from the conservation viewpoint, the ecologically important role of lentic habitats is worth mentioning. Although the mean proportion of lentic area in the downstream reaches was less than 5.7% in our study, the distinctive hydro-physical and landscape characteristics of the lentic areas led to heterogeneous trophic conditions within the reaches. In general, the amount of lentic habitat in sand and gravel bars decreased in downstream reaches due to riverbed degradation (Brandt, 2000; Rollet et al., 2014). In this study, the area of lentic habitat was lower in downstream reaches than in upper reaches, except for KD3. Because KD3 was the most distant reach from the dam, it was believed that the lentic habitat was restored by sediment supplied from the river bank and reduction of the impact of the dam (Rollet et al., 2014). In previous studies, the significance of lentic habitats in the formation of biodiversity was often reported (Taniguchi & Tokeshi, 2004; Tews et al., 2004; Warfe et al., 2008) and was applied to conservation. For example, many river restoration works have created lentic spaces to facilitate spatial heterogeneity of species diversity (Wyżga et al., 2012; Van den Brink et al., 2013) and to supply flow refugia to fishes from flood disturbances (Sedell et al., 1990; Milner & Gilvear, 2012). Creation of lentic habitat was basal idea downstream due to improvement of river morphology, and it was shown that it contributed to mitigation of the trophic impact. Our finding of habitat-specific food resources in lentic spaces and their influence on trophic structures of invertebrate communities expanded the existing idea a little further. Lentic water bodies can help sustain adequate and diverse flows of materials in food webs. In particular, in environments where the main food sources are artificially modified from the natural sources, such as in the dam outlet reaches observed in this study, lentic habitats are expected to act as trophic refugia that can mitigate the trophic impact.

Conclusion

This study illustrated that a stable isotope three-source mixing model could be used to evaluate trophic contributions of reservoir-derived plankton, epilithon, and terrestrial leaves to stream macroinvertebrates in lotic and lentic habitats at four dam outlet reaches and two reference reaches. Although the trophic contribution of reservoir-derived plankton of filter-feeders was not significantly different among habitats, we found that populations of collector-gatherers and predators were significantly lower in lentic habitat than in lotic habitat. Additionally, the trophic contributions of SFPOM, which is the primary food source of filter-feeders, were lower in a lentic habitat than a lotic habitat, reducing the trophic impact of the reservoir. In lotic habitats in dam outlet reaches, spatial variation in the contribution of reservoir-derived plankton to net spinners-filter-feeders were significantly positively correlated with the proportion of dam plankton in SFPOM. However, filter-feeders except for net spinners, collector-gatherers, and predators were significantly positively correlated with the proportion of dam plankton in SFPOM among all habitats. Therefore, the trophic contribution of macroinvertebrates shifted to river-derived organic matter in a lentic habitat downstream of a dam, whereas that of net spinners-filter-feeders depended on dam plankton in the lotic habitat. We suggest that the reduced plankton inflow combined with SFPOM production from both epilithons and leaves affects the trophic contribution of macroinvertebrates in lentic habitats. Although lentic habitats, comprising ≤ 5.7% of space, are a rare space for an ecosystem downstream of a dam, a lentic habitat can maintain species diversity and availability of food resources of river origins. In the dam downstream where food sources have been artificially altered, lentic habitats are expected to prevent the uneven distribution of food sources and increase the diversity of trophic contributions for macroinvertebrates.