Results

Short-term response of macroinvertebrate (drift, stranding) to hydropeaking
Across all reaches, drift intensity was 2.5 times higher during HP (12.5 ± 0.9 ind./m2min) compared to base flow (4.9 ± 0.8 ind./m2min, p< 0.001) and approximately 1.5 times higher than in the RF reaches (7.4 ± 0.8 ind./m2min, p < 0.001, ‘ALL’ in Figure 3). Drift intensity in the RF reaches was also significantly higher than at base flow in the HP reaches (p < 0.05). At the river level, significantly higher drift intensities were found during HP compared to base flow of the Sitter (p < 0.01), Hasliaare (p < 0.001) and Linth (p < 0.01) as well as compared to the RF reaches of the Sitter (p < 0.001) and Hasliaare (p < 0.01). The Linth showed a different pattern with higher drift intensity in the RF reach compared to HP (p = 0.06) and to base flow (p < 0.001).
Across all HP reaches, drift intensities during all HP phases (UR, P1, P2, DR) were significantly higher than during base flow (p ≤ 0.001, Appendix D in Data S1). Moreover, drift intensity in the UR phase was significantly higher compared to the DR phase (p < 0.01). The highest average drift intensity was found for the UR phase (17.1 ± 1.7 ind./m2min) and the first part of the peak phase (P1; 13.7 ± 1.3 ind./m2min) which showed drift intensities 3.5 times and approximately three times higher, respectively, than during base flow (4.9 ± 0.8 ind./m2min).
Across all HP reaches, and considering the entire HP scenario (i.e., UR, P1, P2 and DR phase), stranding density was not significantly related to drift intensity (‘ALL’ in Figure 4a). However, if considering only the UR phase, a positive significant relationship was found (R = 0.368, p < 0.01, ‘ALL’ in Figure 4b). At the reach level, a positive significant relationship was found in the Hasliaare and in the Sitter (but only for the UR phase) but not in Linth. In general, stranding was much less pronounced than drift and many stranding samples contained only few individuals (Appendix C in Data S1).
The CCA analyses revealed that taxa drift and stranding composition significantly differed between the three rivers (drift: r2 = 0.23, p < 0.01; stranding: r2 = 0.32, p < 0.01; Figure 5). Additionally, but to a lesser extent, the lateral sampling location (day 1 vs day 2: higher peak flow velocities at day 2, see Table 1; drift: r2 = 0.16, p < 0.01; stranding: r2 = 0.27, p < 0.01) significantly contributed to the differences. Drift composition in the Sitter and Linth were more similar than in the Hasliaare (Figure 5a, confidence ellipses). The HP scenario also significantly contributed to the drift (r2 = 0.16, p = 0.05) but not to the stranding differences. 43.5% and 40.8% of the total variation of drift and stranding taxa distributions, respectively, can be explained by the selected environmental variables. The main environmental variables that could affect the taxa drift propensity to the greatest extent were the flow velocity near the surface (v100 ; F = 12.5, p < 0.001), mean up-ramping rate (URmean; F = 10.4, p < 0.001), turbidity (NTU ; F = 6.4, p < 0.001), and water temperature (T ; F = 6.0, p < 0.001) which explained 8.0, 9.8, 9.0, and 8.5% of the variation, respectively (Figure 5a). Froude number (Fr ) also contributed to the explained variation (8.2%) but not significantly (F = 1.7, p = 0.119). The main environmental variables that could affect the taxa stranding propensity to the greatest extent were the flow ratio (Qpeak/Qbase ; F = 9.7, p < 0.001), max. down-ramping rate (DRmax ; F = 8.8, p < 0.001), water temperature (T ; F = 5.5, p < 0.001), turbidity (NTU ; F = 4.1, p < 0.001) and flow velocity near the surface (v100 ; F = 3.7, p < 0.01), which explained 7.3, 8.8, 8.2, 8.1, and 8.3 % of the variation, respectively (Figure 5b).
Across all HP reaches, Rhyacophilidae and Limnephilidae showed high propensity to drift and strand, whereas Oligochaeta and Heptageniidae showed the lowest propensity (‘ALL’ in Table 2). Simulidae also showed high propensity to drift and Perlodidae to strand. However, drift and stranding propensity varied considerably between reaches. Limnephilidae and Simuliidae showed considerable propensity to drift in all three HP reaches, whereas Rhyacophilidae only in the Hasliaare. Limnephilidae, Rhyacophilidae and Nemouridae showed considerable propensity to strand only in the Hasliaare, whereas Empididae and Perlodidae in the Linth. Almost all taxa showed highest propensity to drift and strand in the Hasliaare and lowest propensity to drift and strand in the Sitter.
Long-term response of macroinvertebrate (density and community composition) to hydropeaking
Average benthic density was five times higher in the Hasliaare RF reach (1146 ± 229 ind./m2) compared to the HP reach (227 ± 23 ind./m2, p < 0.001), whereas in the Sitter a contrasting trend was observed with average density approximately 3.5 times higher in the HP reach (1838 ± 357 ind./m2) than in the RF reach (493 ± 97 ind./m2, p < 0.001, Figure 6). The density variability of the Sitter HP samples was the largest of all reaches. No significant differences were found for the Linth and considering all reaches together (‘ALL’ in Figure 6).
The NMDS analysis showed that benthic community composition was significantly different between the six reaches (p = 0.001), whereby it was more similar in the three RF reaches than in the three HP reaches (Figure 7, overlap vs no overlap of the confidence ellipses; lower R2 and F values in Appendix E in Data S1). The Hasliaare showed the largest dissimilarity between RF and HP reaches, whereas the other two rivers grouped more together, indicating more considerable consistency in benthic community composition.
Four taxa (Heptageniidae, Chironomidae, Baetidae and Leuctridae) remarkably contributed to differences in community composition among RF and HP reaches of all three rivers (Table 3). Among these taxa, Heptageniidae, Chironomidae and Baetidae in the Sitter, and Chironomidae, Baetidae and Leuctridae in the Linth cumulatively contributed for approximately 70% of the dissimilarities. In the Hasliaare two other taxa, Limnephilidae and Nemouridae, accounted for 41% of the differences among RF and HP reaches, and Chironomidae additionally contributed for approximately 15%.
Across all reaches, Leuctridae, Nemouridae and Limnephilidae showed significantly higher benthic densities in the RF than in the HP reaches (p < 0.001, Appendix E in Data S1). In contrast, Simuliidae showed significantly higher benthic densities in the HP reaches (p < 0.05). However, differences between HP and RF reaches varied widely among taxa and river (Appendix E in Data S1).
On average, lentic taxa were 5.5 times more abundant in the benthic samples of the RF reaches (191.2 ± 47.7 ind./m2) compared to the HP reaches (35.0 ± 6.1 ind./m2; p < 0.001), whereas lotic taxa were two times more abundant in the HP reaches (437.8 ± 105.3 ind./m2) than in the RF reaches (222.2 ± 35.1 ind./m2), but the difference was not significant (‘ALL’ in Figure 8). Differences between RF and HP reaches for taxa classified as surface dwelling or interstitial were also not statistically different.