Hydropeaking impairs upstream salmonid spawning habitats in a restored Danube tributary

Hydropower is considered an important form of renewable energy, often involving hydropeaking. While the effects of hydropeaking on aquatic communities in areas downstream the dam are well understood, there is a lack of studies investigating potential impacts on tributaries located further upstream. In this study, we tested the effects of hydropeaking operations on upstream tributaries in a restored area of the Danube River, with a focus on the periods of backlog and release of water (up‐ramping and down‐ramping, respectively) during the filling and release of the reservoir. We used brown trout egg and larval mortality, linked to hydraulic, sedimentary and physiochemical changes in spawning grounds as an indicator. We compared hydropeaking‐affected versus non‐affected sites in upstream tributaries using HydroEcoSedimentary Tools (HESTs) loaded with clean gravels and brown trout eggs. Egg and larval mortalities were significantly higher in the hydropeaking‐affected site with more than 80% egg mortality and almost 100% larval mortality compared to values of 55–63% and 80–85%, respectively, in non‐affected sites. Spawning ground quality was significantly altered in the hydropeaking‐affected site, where the highest mortalities were observed. Overall, duration of time periods with flow velocities close to zero were a key variable, potentially decreasing oxygen supply for eggs and larvae. Such periods of close to zero flow velocities were driven by backlog periods during the filling of the reservoir, revealing that such events can severely impair ecological integrity of spawning sites in tributaries upstream of dams by slowing the flows in upstream tributaries. Such altered processes can reduce fish population recruitment and need to be considered in future restoration projects.


| INTRODUCTION
Increasing energy demand and energy decarbonization drive the construction of new hydropower plants in Eastern Europe and developing countries in Asia, Africa and South America (Lehner, Czisch, & Vassolo, 2005;Zarfl, Lumsdon, Berlekamp, Tydecks, & Tockner, 2015). In central Europe, where hydropower has had a long tradition and where its potential is almost fully exploited, power plants are being optimized, changing to more effective turbine technologies (Sari et al., 2018) or operation modes (Wagner, Hauer, & Habersack, 2019) that increase the implementation of hydropeaking regimes (Smokorowski, 2021). The need for the operators to run the power plant as efficient as possible makes it attractive for them to use hydropeaking as an operational tool to feed peaks of energy demand and maximise energy production revenue (Venus, Smialek, Pander, Harby, & Geist, 2020). Sudden stops and starts of the turbines to store and release water efficiently typically results in hydropeaking.
Upstream effects of hydropeaking are less well understood. In the headwater area directly upstream of the power plant, hydropeaking potentially impairs fish habitat by increasing water depth and reducing current speed, leading to more uniform conditions that favour generalist or limnophilic aquatic species and communities (Greimel et al., 2018;Mueller, Pander, & Geist, 2011). During the water storage phase, when the turbines stop, a water backlog may be formed, reducing flow velocity. Low flow velocities may increase fine sediment deposition and potentially affect ecologically relevant processes in the interstitial zone. Most relevant is streambed clogging impairing spawning ground quality (Hauer, Holzapfel, Tonolla, Habersack, & Zolezzi, 2019), altering streambed communities (Bondar-Kunze, Kasper, & Hein, 2021;Elgueta et al., 2021;Salmaso, Servanzi, Crosa, Quadroni, & Espa, 2021), and potentially also affecting the river productivity (Greimel et al., 2018;Mueller et al., 2011). An understudied aspect of these processes is the potential effect that such hydropeakinginduced backlogs may have on aquatic habitat of upstream tributaries.
This lack of knowledge is surprising given the key role tributaries play as aquatic habitats for spawning, larval development and juvenile growth of many riverine fish species such as rheophilic salmonids and cyprinids (Naus & Reid Adams, 2018;Pracheil, Pegg, & Mestl, 2009).
In this study, we aim at tackling the lack of understanding of upstream effects of hydropeaking, with a focus on tributaries and hydropeaking-induced backlog events. We assessed the upstream effects of hydropeaking of the power plant Ingolstadt in the European River Danube on a structurally restored tributary located within a large floodplain restoration area (Stammel et al., 2012), which is known to provide several spawning grounds for lithophilic fish along its course (Pander et al., 2015b;Pander, Knott, Mueller, & Geist, 2019;Pander, Mueller, & Geist, 2018). We hypothesized that a backlog resulting from hydropeaking operations in the main river will impair spawning ground quality in the upstream tributary. In particular, the changed hydrograph of the tributary will have severe consequences on water depth, current speed, fine sediment deposition and oxygen supply, leading to unfavourable conditions in the interstitial spaces of the spawning grounds and reducing the recruitment success of brown trout in affected sites compared to sites in the same system not affected by hydropeaking.

| Site description
The Ottheinrichbach (OHB) River is located in the middle section of a large scale floodplain restoration project at the upper River Danube (mean annual discharge = 333 m 3 Ás À1 ) in Bavaria, south-Germany (Fischer & Cyffka, 2014;Pander et al., 2015b;Stammel et al., 2012;48 44 0 55,7 N, 11 16 0 35,7 O; Figure 1a). It was constructed to restore the impaired fish migration within the Danube due to the construction of the power plant Bergheim, to increase groundwater dynamics, to provide additional habitat for keystone organisms and to re-wet one of the last remaining fluvial forests along this large European River (Pander et al., 2015b, Pander, Mueller, Knott, Egg, & Geist, 2017, Pander et al., 2018, Pander et al., 2019Stammel et al., 2012;Stammel, Fischer, Gelhaus, & Cyffka, 2016). Due to its location in spatial proximity to the hydropower plants Ingolstadt and Bergheim and its important function as a fish nursery (Pander et al., 2015b), the OHB represents an ideal model stream for tributaries located along heavily modified water bodies affected by hydropower generation. Due to hydropeaking of the Ingolstadt Plant, the confluence OHB-Danube is subject to severe flow fluctuations largely driven by the power plants' backlog ( Figure 2). The backlog effect reaches from 700 to 1,200 m into the OHB, depending on the discharge of the River Danube, potentially impairing one of the spawning grounds that were newly built in the year 2010 during the floodplain restoration ( Figure 3). The OHB has a variable discharge between 0.5 and 5 m 3 Ás À1 and can be additionally flooded with another 25 m 3 Ás À1 to mimic natural dynamics of flood scenarios that were historically present in the alluvial forest (Fischer & Cyffka, 2014;Pander et al., 2019;Stammel et al., 2012).
In the OHB, three sites known to function as spawning grounds for lithophilic fish species were chosen for active bioindication with trout eggs (Pander & Geist, 2013).

| Hydroecological assessment
In order to assess the potential effects of the hydropeaking-driven backlog, we compared the quality of the spawning grounds between F I G U R E 1 Illustration of the study area within Europe (a) and location of the three investigated spawning sites within a large scale floodplain restoration at the Danube River (b) including a schematic top-view on the study area between the power plants Bergheim and Ingolstadt with an indication of the backlog caused by the power plant Ingolstadt in side view. Before exposure, each HEST was pre-filled with sediment truncated at the 6.3 mm fraction. We followed granulometry curves from the OHB's natural streambed material, described in Pander et al. (2015b), to mimic the corresponding sediment sizes at each site.
Brown trout eggs were purchased from a local fish farm (Forellenhof Nadler, Eching, Germany). Brown trout eggs were inserted in the eye-F I G U R E 2 Hydrograph of the Danube's discharge during the exposure of the brown trout eggs in comparison to water level changes in the hydropoweraffected site 3 (a) as well as for representative medium (b) and low (c) discharge conditions during the egg exposure. Light blue arrows in panel c indicate flow direction towards the Danube (OHB discharges regular into the Danube); blue arrows indicate flow direction from the Danube upstream into the OHB ( (Pander & Geist, 2010) and interstitial water quality (Pander, Schnell, Sternecker, & Geist, 2009). After receiving the eggs from the fish farm, they were acclimatized for 30 min including a temperature adjustment of 2 C in the laboratory of the Aquatic Systems Biology Unit, Freising, Germany. Dead eggs were subsequently removed, and the remaining live eggs were distributed into the HESTs. Each HEST compartment (T and B) was loaded with 30 eggs that were randomly picked and placed by forming a small pit in the HESTs, which was then carefully covered with 5 cm of substrate. The full HESTs were kept submerged in a large tank with aerated water, so the quality of the eggs was preserved during transport until their installation in the river the next day (see Casas-Mulet et al., 2021 for further details).
On 18th December 2018, 27 HESTs were installed at each site into the streambed of the OHB. A hole big enough to fit each of the HESTs was dug with a common spade, keeping disturbance of the surrounding streambed to a minimum. Each HEST was then inserted so that its top was even with the gravel surface, and the river´s natural coarse sediment was used to fill up the gap to the same level. The HESTs were installed in closed mode and after fines were settled, they were opened to start the experiment (Casas-Mulet et al., 2021).
After 33 days of exposure, all HESTs were retrieved and immediately assessed in the laboratory for egg and larvae survival. The HEST compartments were opened carefully, data loggers and larger stones were removed, and the sediment content was emptied into a tray, rinsed with water and examined for larvae and eggs. A distinction was made between live and dead eggs as well as between live and dead larvae,

| Data analysis
Total percent mortality was calculated as the difference between the surviving eggs or larvae and the initially loaded egg number for each of the HESTs retrieved from the three sites. For univariate multiplegroup comparisons of egg and larval mortality as well as abiotic habitat variables, each dataset was tested for normal distribution (Shapiro-Wilk test) and homoscedasticity (Levene test). Since data did not fulfil the criteria for parametric testing, the non-parametric Kruskal-Wallis test was applied to test for significant differences. A subsequent post-hoc Wilcoxon test with Bonferroni correction for multiple comparisons was used to determine whether values differed significantly between HESTs, compartments and/or infiltration directions. Since no significant differences between longitudinal, lateral and vertical HEST measurements were observed, the data was pooled for further investigation. Univariate statistics were carried out using statistical and graphical open-source software R (R Core Team, 2020, version 4.0.3R, www.R-project.org/, last accessed on 27 July 2021).
To visualise differences of HESTs in egg and larval mortality between the three sites, a non-metric multidimensional scaling (NMDS) using PRIMER v7 (Plymouth Marine Laboratory, Plymouth, UK) was plotted. For this multivariate comparison, a resemblance matrix was calculated using the full data resolution of egg and larval mortality from the different retrieval time points of HESTs at the three sites (Clarke, Gorley, Somerfield, & Warwick, 2014). To test for significant differences of mortality rates between HESTs and between top and bottom compartments, one-way analysis of similarities (ANOSIM) based on Bray-Curtis similarities (Bray & Curtis, 1957) calculated from egg and larval mortality data (Clarke, 1993) were computed. To analyse the interaction between the measured abiotic habitat variables and the ordination of HESTs in the NMDS, the abiotic habitat variables were displayed in the NMDS using the overlay function in PRIMER.
To test for relations between egg mortality and minimum oxygen concentration found in the HESTs, the oxygen concentration was used as predictor variable and plotted against the response variable egg mortality. To avoid false-negative conclusions due to potentially non-linear relations, Spearman rank correlation for monotonic trend were computed and a smooth curve was displayed. This analysis was also performed using R (R Core Team, 2020). For all testing's, a significance level of p ≤ .05 (= 95% probability) was applied.

| Brown trout eggs and larval mortality
Univariate comparisons of egg and larval mortality revealed significant differences (Table 2) between the replicates of hydropeaking-affected site 3 and the two non-hydropeaking affected sites 1 and 2 that were T A B L E 1 Abiotic spawning ground characteristics of the assessed sites considered as a reference. In the hydropeaking affected site 3, egg and larval mortalities were significantly higher ( Figure 4, Table 2), with over 80% egg mortality and almost 100% larval mortality compared to the non-hydropeaking affected sites. Egg and larval mortality in the non-hydropeaking affected site 1 (over 63% egg mortality and 85% larval mortality) and site 2 (55% egg mortality and 80% larval mortality) were not significantly different (Table 2). In site 2, we found the lowest egg and larval mortality of the tested spawning grounds ( Figure 4).
In addition, significant differences between top and bottom compartments could only be detected in site 3 (ANOSIM: R = 0.07, p < .05) with 20.5% higher mortality for eggs and 19.6% for larvae in bottom compartments compared to top compartments (Table 4). Correlating abiotic habitat variables with egg mortality revealed that mortality increased sharply with decreasing oxygen supply in the interstitial zone, with more than 90% egg mortality when oxygen values were below 10 mgÁL À1 (Figure 6).

| Abiotic habitat variables at the spawning sites of the OHB
All abiotic habitat variables of sites 1 and 2 differed significantly from those measured in the hydropeaking-affected site 3, except for temperature (Kruskal-Wallis ANOVA p < .001 and subsequently applied Wilcoxon signed-ranks post-hoc test, see Table 1). In addition, site 1 differed significantly from site 2 with lower values of electric conductance, depth and current speed (Table 1). However, the differences for some variables between sites concerning electric conductance, pH and turbidity were small and within the range of values required for successful brown trout development. Mean dissolved oxygen values were highest at site 2 (11.79 mgÁL À1 ) and 2 mgÁL À1 lower at site 3 (Table 1) Figures 2 and 3). In contrast to site 3, the water level at sites 1 and 2 was rather constant during the day and comprised less water level fluctuations during the investigation period (seasonal fluctuation in site 1 = 15 cm and in site 2 = 12 cm). In sites 1 and 2, the flow Letters below the median indicate significant differences between sites direction was always constant and did not change direction. Mean particle diameter was largest at site 2 (2.75 mm) and smallest at site 3 (1.66 mm). Site 2 also comprised the highest infiltration rates (1.12 kgÁd À1 ) of fine sediment, being almost twice as high as at site 3.

| DISCUSSION
Hydropower is a growing source of renewable energy worldwide with negative effects on aquatic habitats and biological communities, particularly fishes (Auer et al., 2017;Bartoň et al., 2021;Casas-Mulet et al., 2016;Grabowski & Isely, 2007;Greimel et al., 2018;Hauer et al., 2017;Moreira et al., 2019;Puffer et al., 2015;Schmutz et al., 2015). An objective evaluation of the "green" and "red" aspects   Such exchange between open water and interstitial is crucial for oxygen supply and successful egg or larval development (Sternecker, Cowley, & Geist, 2013). It is possible that the stagnant conditions caused by the backlog were responsible for an oxygen deficiency in the interstitial pores, as the subsurface flow was not enough to supply the exposed eggs and larvae with oxygen and to transport the metabolites away. In addition, respiration of organic matter could have led to reduced oxygen availability in the interstitial spaces. At site 3, there was no increased sedimentation of fines detectable compared to site 2 or site 1 where egg and larval development was on average 20%-40% higher. Even though this was not experimentally tested, based on prior literature, we took the assumption that stagnant flow conditions result in an accumulation of a thin layer of the commonly present fines <0.0045 mm on the eggs, hindering oxygen supply and removal of metabolites (Greig, Sear, & Carling, 2005 We acknowledge that this study represents only one season and one hydropeaking affected area, and that, in general, spatial and temporal variability in spawning success can be high. However, this case study helps illustrate that the positive effects of structural river restoration can be limited in highly hydropower-affected environments (Pander & Geist, 2013). If constraints such the one illustrated here in a hydropeaking-imparied tributary cannot be mitigated, restoration success may be lower than expected (Geist & Hawkins, 2016). This is particularly critical for highly specialized riverine fish such as rheophilic salmonids and cyprinids, which are all high on the conservation agenda (Mueller, Bierschenk, Bierschenk, Pander, & Geist, 2020). Such upstream hydropeaking effects can lead to severe bottlenecks in species recruitment, which is a prerequisite for a F I G U R E 6 Correlation plot of the oxygen concentration measured in the HESTs against egg mortality. The smooth curves are locally weighted regression fits (Cleveland & Devlin, 1988) to the data points. LCL = lower critical limit for egg to fry survival according to Rubin and Glimsäter (1996) reviewed in Smialek, Pander, and Geist (2021) [Color figure can be viewed at wileyonlinelibrary.com] successful population development, and is an important factor for the functionality of river restoration.

| CONCLUSION
To the best of our knowledge, this study is the first to provide evidence on the potential ecological upstream effects of hydropeaking, with a focus on fish habitat quality in tributaries. Spawning grounds can be severely affected by hydropeaking operations carried out in the main river, particularly when these comprise strong water level fluctuations linked to water storage and release phases. The negative effects of hydropeaking in this study were expressed in reduced salmonid hatching rates and larval survival. These could be linked to water level fluctuations causing temporarily increased water depth and reduced current speed at the affected spawning grounds, which may have led to the reduced oxygen availability in the hydropeaking affected site. Accounting for such effects is critical for the survival of gravel-spawning early life stages of many fish species. Particularly for target species of conservation such as rheophilic salmonids and cyprinids, such altered processes can result in reduced population recruitment and need to be considered in the design and implementation of restoration projects.

ACKNOWLEDGEMENTS
We would like to thank Catarina Eirich, Julia Reinbeck and Martin Gauger who contributed to the sampling in the Ottheinrichbach. We are also grateful to all the volunteers and student assistants who helped with the installation and retrieval of the HESTs in the field and later sample processing. We thank the AuenZentrum Neuburg-Ingolstadt for providing water level data, and, in particular to B. Cyffka for coordination of the permissions needed from the landowners. We also would like to thank the water authorities, in particular B. Kuegel from the local water authority, for their kind support of the study. This research was partly funded by the Alexander von Humboldt Foundation through a fellowship awarded to R. C.-M., carried out at TUM. This study was also partly supported by the framework of the project AquaKlif in the bayklif network for investigation of regional climate change funded by the Bavarian State Ministry of Science and the Arts. Open access funding enabled and organized by Projekt DEAL.
Open access funding enabled and organized by Projekt DEAL.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.