Figure legends

Figure 1. Overview map of the three study sites. White dots show the water releases of the hydropower plant (KUB) and of the retention basins (KWO and KLL). HP and RF reaches are identified with white and yellow rectangles, respectively. White arrows show the flow direction. Source background Orthophotos: Swissimage © Swisstopo.
Figure 2. Experimental setup. (a) Schematic sampling design for HP (top panel) and RF reaches (bottom panel). (b) Chronological sampling illustrated as example of the three HP scenarios (SC1, SC2, SC3) at day one in the HP reach of the Sitter. Drift samples were collected separately at base flow (B) and during each HP phase: up-ramping (UR), first and second peak (P1, P2), down-ramping (DR). (c) HP scenarios. Water level (blue solid line) and water temperature (red solid line) measured in the HP reaches as well as water temperature measured in the RF reaches (red dotted line). Grey squares indicate sampling period at day one and two. Please note different scales on y-axes.
Figure 3. Drift intensity (ind./m2min) across all reaches (‘ALL’) and separated by reach. RF: residual flow reach; B: base flow; HP: during hydropeaking (i.e., entire HP scenario: UR, P1, P2 and DR phase). Boxplots show the 25th and 75th percentiles, median (solid line in the box), mean (dashed line in the box), whiskers (10th and 90th percentiles) and outliers (white dots). Numbers indicate the sample size, whereas letters show group affiliation according to post hoc tests with Bonferroni correction.
Figure 4. Relationship between drift intensity (ind./m2min) and stranding density (ind./m2) based on linear regression models across all three HP reaches (‘ALL’) and separated by reach (Sitter: black square; Hasliaare: dark grey circle; Linth: light grey triangle). (a) Drift data of an entire HP scenario: UR, P1, P2 and DR phase. (b) Only drift data of the UR phase. Solid lines: linear regressions; dashed black lines in the left plots: 95% confidence intervals. To better meet assumptions of normality and homogeneity of variances, all data were log transformed (X+1) prior to computing regressions. Although, in (b) the data for ‘ALL’ and Hasliaare are not normally distributed. R: coefficient of multiple correlation.
Figure 5. Biplot of Canonical Correspondence Analysis (CCA) based on (a) drift propensity and (b) stranding propensity ofthe most common taxa in the three HP reaches. Left panels: confidence ellipses (95% confidence limit, standard deviation) are fitted on the CCA plots to depict differences between river reaches. Right panels: distribution of the species scores of the most common taxa (the larger the circle the higher the propensity). Asterisks indicate taxa with abundances > 1% in the drift/stranding but not in the benthic samples (Appendix C in Data S1). Factor fitting shows the statistical differences in drift/stranding propensity between HP reaches (Sitter vs Hasliaare vs Linth), sampling day (day 1 vs day 2) and HP scenario (SC 1 vs SC2 vs SC3). Significance levels: *0.05, **0.01, ***0.001.
Figure 6. Benthic density (ind./m2) across all three RF respectively all three HP reaches(‘ALL_RF’, ‘ALL_HP’) and separated by reach. Numbers indicate the sample size, whereas letters show group affiliation according to Mann-Whitney tests.
Figure 7. NMDS ordination of the three RF and three HP reaches based on benthic densities (ind./m2) of the most common taxa (relative abundance > 1%, Appendix C in Data S1). 95% confidence ellipses of the standard deviations of site scores of the three rivers (Sitter, Hasliaare, Linth) within the two respective reaches (RF: solid ellipses; HP: dashed ellipses) are depicted. Taxa that most contributed to dissimilarity (p < 0.05) are fitted on the NMDS plot and indicated with black dots and taxon name.
Figure 8. Benthic density (ind./m2) across all three RF respectively all three HP reaches separated by the two traits ‘interstitial/surface’ and ‘lentic/lotic’. Numbers indicate the sample size, whereas letters show group affiliation according to Mann-Whitney tests.