Drawdown flushing in a chain of reservoirs—Effects on grayling populations and implications for sediment management

Abstract We used an information theoretic approach to assess the effects of an ecologically adjusted sediment management scheme on grayling (Thymallus thymallus L. 1758) populations. Additionally to reservoir operation, candidate models included a variety of parameters and processes that may influence grayling populations such as flow, temperature, density dependence, and bird predation. Population parameters analyzed included total densities, young of the year numbers, and larval densities. These analyses were supplemented by a characterization of sediments and sedimentation patterns in the reach. Investigations were carried out in six sites affected by flushing and in one control site. A total of thirteen flushing operations have been undertaken within the study period leading to considerable remobilization of fine sediments and gravel. Due to seasonal and hydrological restrictions, not every flood could be used for flushing. These limitations led to an interrupted management throughout the chain of reservoirs as well as to long time intervals between flushing events with possible effects on spawning habitat quality. None of the investigated population parameters was affected by flushing, and thus, the study generally supports the current reservoir management scheme. Our analyses revealed the magnitude and timing of high water events, temperature, and density‐dependent effects, that is, population densities the year before, as the most influential variables for grayling population dynamics in the investigated stretch. The siltation of reservoirs is a significant problem for reservoir storage, flood protection, river deltas, and coastal zones. Its management—which is inevitable to safeguard river deltas and secure flood protection—poses also the challenge to safeguard riverine ecosystems below reservoirs. Based on our experience, we propose a periodic flushing regime in concordance with the hydrograph thereby mimicking the timing, magnitude, frequency, and duration of natural SSC pulses and gravel transport. This flushing regime minimizes adverse downstream environmental impacts and maximizes benefits.


| INTRODUC TI ON
Globally, more than 25% of sediment flux is trapped in artificial impoundments (Vörösmarty et al., 2003). Although humans are simultaneously increasing the river transport of sediment through soil erosion activities, the net result is a global reduction in sediment flux by about 1.4 BT/year over prehuman loads with subsequent impacts on coastal ecosystems (Syvitski, Vörösmarty, Kettner, & Green, 2005). For the European Alps, Hinderer, Kastowski, Kamelger, Bartolini, and Schlunegger (2013) estimated that only 45% of the sediments mobilized in headwaters are exported out of the Alps, most sediments being trapped in artificial reservoirs. From this global perspective, it becomes clear that releasing sediments from reservoirs is necessary to safeguard and protect river deltas and other coastal systems.
The adverse ecological consequences of flushing releases are related both to the increase of suspended sediment concentration (SSC) during the removal operation and to the modification of riverine habitats following deposition of the flushed fine material. Studies on the effects of fine sediment, both suspended and deposited, are numerous and have been reviewed by several authors (e.g., Kemp, Sear, Collins, Naden, & Jones, 2011). It is generally acknowledged that the direct impact differs between species-with salmonids being most vulnerable-and varies with SSC and exposure time.
Deposition of the flushed fine material may also have a significant impact on fish populations. Generally, an increase in fine sediments was found to decrease hatching success in salmonids through reducing intragravel O 2 concentrations (Greig, Sear, & Carling, 2005;Jensen, Steel, Fullerton, & Pess, 2009). As grayling (Thymallus thymallus L., 1758)-contrary to other salmonids-do not bury their eggs but lodge them at depths of 2 to 3 cm into the substrate (Northcote, 1993), this relationship may not be so pronounced in this species.
An aspect mostly ignored when assessing flushing operations is the delivery of coarse substrate during flushing. Grayling has very specific requirements on spawning substrate, preferring mediumsized gravel (20 to 50 mm, Mouton et al., 2008). Sediment supply downstream of dams is greatly reduced, often leading to coarsening of the substrate and armoring (review in Bednarek, 2001). During flushing, these processes may be mitigated and ultimately leading to habitat improvement (Kondolf & Matthews, 1991).
Historically, flushing operations have often been undertaken without consideration of ecological impacts. This led partially to high mortality in fish populations in consequence of flushing (e.g., Hesse & Newcomb, 1982). However, the increase in environmental awareness has led to the recognition that the reservoir management includes a responsibility to protect the natural resources that depend on water. As a result, considerable effort has been invested in developing approaches to lessen the damaging effects of flushing operations including guidance values for SSC, seasonal restrictions, and postflushing with clear water.
The aim of the current paper was to assess the effects of an ecologically adjusted reservoir management on Thymallus thymallus populations. Our assessment is based on several population parameters including total and juvenile densities as well as their rate of change. This allows an in-depth analysis of factors governing population dynamics, including density dependence and factors only operative at individual development stages or ages. Additionally, we incorporated possibly confounding factors such as predation, temperature, and hydrology into our analyses. These analyses are accompanied by a characterization of the flushing events with respect to sediment balance, SSC, and oxygen concentration, as well as a description of the sediment up-and downstream of reservoirs regarding its suitability for spawning.

| Study site
The Mur River is one of the largest rivers in Austria, with a total length of 444 km. It rises in the Hohe Tauern National Park (1,898 m above sea level) and about 325 km are within Austria. The Austrian catchment comprises approximately 10,341 km 2 .
The Mur River has a nivo-pluvial hydrological regime with low discharge in winter, elevated flows during the snow melting period in spring and precipitation-caused flood events in summer and early autumn. High water levels predominantly occur in summer (July, August). Extreme high water events (ca. HQ5) were observed in 1989, 2002, 2005, and 2012 (Supporting Information Data S1).
Although affected by several hydropower plants (HPPs) and partly embanked, the upper reach is still one of the least impacted large rivers in Austria (Brilly, Sraj, Vidmar, Horvat, & Koprivsek, 2012;Matulla, Schmutz, Melcher, Gerersdorfer, & Haas, 2007) and is considered one of the highest rated grayling sites in Austria from a fisherman's perspective. A compilation of salmonid density and biomass in 25 river stretches in Austria identified the Upper Mur River as the one with the second highest values concerning grayling biomass and density (Kaufmann et al., 1991). Recreational fisheries management includes stocking with brown trout (Salmo trutta fario) and rainbow trout (Oncorhynchus mykiss). Exploitation rates are usually low as most stretches are privately owned and fishing is restricted by license.
The study site comprises a section of 133 km and is located be- HPP Fisching is a run-of-river diversion hydropower plant, that is, water from the main river is diverted at r-km 317. The residual flow section extends to the location of the powerhouse at r-km 320.7. All HPPs have appropriate flood control devices, which can be used for flushing. HPP Murau and HPP Judenburg have overflow weirs and the sluice gates are used to drawdown the water level. All the other HPPs have radial gates (Table 1). During drawdown, the top of the sills of the radial gates are close to the river bed, so gravel can pass the weir efficiently.
The section can be divided into stretches with differing hydromorphic characteristics. The most natural stretch is the one between Unzmarkt and Judenburg (mean hydromorphological status = 1.7, Table 2), and the most altered one lies between St. Georgen and Murau (mean hydromorphological status = 3.0, Table 2).
The uppermost stretch is not affected by drawdown flushing and was thus used as a control in our analyses.

| Sampling design
In this study, the results of several investigations addressing fish ecology and reservoir management have been integrated. Most of the data have been gathered during monitoring programs to support authorities and the operators of the power plants in reservoir management and its assessment. A list of the data sources used in the study is provided in Supporting Information Data S3.
F I G U R E 1 Study site between Stadl (r-km 403) and Leoben (r-km 270) with power plants (HPP), the different fishing stretches (black), and location of sediment samples (I-VI). The insert shows the location of the study site in Austria Our main objective was to assess the ecologically adjusted sediment management scheme. Such an assessment is not straightforward, as confounding effects (e.g., high water levels associated with high sediment load) may blur the results, especially in short-term investigations. To disentangle these effects and to analyze population dynamics, long-term investigations are necessary. To gain further insight into the processes structuring the grayling populations, we supplemented the fish data with information on river morphology, hydrology, temperature, and predator densities.

| Discharge and temperature
Discharge and water temperature are routinely collected at the gauge Fisching. The gauge Fisching is situated in the middle of the study site, and values are representative for the whole stretch. We used mean daily values for statistical analyses. A description of the hydrograph and the temperature regime is given in Supporting Information Data S1.

| Sediment retention and its removal during flushing operations
In the reservoirs Bodendorf and Fisching, topographical data are routinely collected using state of the art 3° single-beam echosounder with GPS reference. Additionally, grain size distributions were analyzed once for three different strata in each impoundment ( over a given period. Together with the information on the grain size distribution, the retention and removal of total sediment as well as the retention and removal of gravel (>2 mm grain size) were calculated.

| Sediment data
In

| Cormorant data
Data on the total cormorant abundances in Styria were provided

| Statistical analyses
Statistical analyses were carried out in SPSS and in R 3.0.3. (R core Team, 2014) using the R packages glmulti (Calcagno, 2013) and MuMIN (Barton, 2016 First analyses were performed with the data from the uppermost stretch that is not affected by HPPs (termed "control" or "upstream St. Georgen" throughout the text), to identify the variables supposedly regulating grayling dynamics under "natural" conditions. In a second set of analyses, all data were analyzed.
Reduced data sets (i.e., data, where at least two consecutive years were available) were used to detect density dependence in population regulation. Individuals aged ≥3+ were designated as potential spawners (=adults).

| Reservoir sedimentation & flushing operations
The yearly sedimentation rates were about 10% of the original storage volumes for both reservoirs (Figure 4). Additionally to fine sediments, considerable amounts of gravel and coarse sand (grain size > 2 mm) were remobilized. This bed load can be estimated to account for about 22% of total sediments in Fisching and 46% in Bodendorf.
The duration of the flushing events varied between 14.5 and 68.0 hr. Mean SSC varied between 0.6 and 9.7 g/L. Maximum values were mostly below 8 g/L, only once values exceeded 15 g/L for a limited time (Table 5).
At the three flushing events in Bodendorf where oxygen concentration was measured, no indication of oxygen depletion was evident (Table 5). In 1999, oxygen concentration was always above 9.3 mg/L (mean: 10.2 ± 0.4 SD). In 2002, mean oxygen concentration was 14.3 +1.8 SD mg/L with a minimum of 9.1 mg/L. In 2004, oxygen concentration varied between 7.7 and 13.8 mg/L (mean: 9.6 ± 1.0 SD) and oxygen saturation was always above 68%. At Fisching, data from 1999 also indicate sufficient oxygen supply during the drawdown (mean: 11.4 ± 0.8 SD, MIN = 9.9 mg/L).

| Total grayling densities
Densities of grayling varied considerably between the different sections and years. Maximal total densities (>3,000 grayling ha −1 ) were observed in the residual flow reach downstream of the HPP Fisching, generally low densities in the stretch between St. Georgen and Murau ( Figure 5).
In the control reach, total densities were related to the yearly mean temperature and to the mean discharge. In the analysis of all data, only stretch and mean temperature had an effect on total densities. The analysis of the rate of population change revealed winter temperature, the densities in the year before sampling, and maximum discharge as the most influential variables. None of the best ranked models showed a negative effect of flushing (Table 6, Supporting Information Data S6).

| YOY PAS surveys
The density of larvae at the first sampling date each year was negatively related to the maximum discharge in spring. The population change rate of fish larvae between two consecutive sampling events was negatively related to the discharge within this period and positively related to mean larval size. Autumnal densities were best explained by discharge and temperature. None of the best ranked models showed an effect of flushing on autumnal YOY abundance (

| YOY electrofishing surveys
Autumnal abundance of YOY varied considerably between stretches.
Highest densities were observed in the residual flow reach, and lowest densities were found in the stretch between St. Georgen and Murau. Generally, strong year classes coincided with dry years, F I G U R E 4 Trend in total sediment volume (zero = start of operation), volume difference between echo sounding surveys, and output of coarse bed load for Bodendorf and Fisching reservoirs weak ones with wet years which were often accompanied by flushing ( Figure 6).
The analyses of the control site revealed hydrology and temperature as the most influential variables. The same was true for the complete data set. Of all tested variables, only site, temperature, maximum discharge, and the month/day when this discharge occurred were included in the best models. Temperature had a positive effect, and Q max negatively affected autumnal YOY abundance.
A flood had a less negative effect when it occurred late in the year.
The analysis with a reduced data set, including potential spawners as a predictor, showed evidence for Q max and stretch to influence autumnal YOY abundance. None of the analyses included flushing in the best models according to the AIC c criterion (

| Flushing events
Based on our analyses, no significant impact of flushing operations on population dynamics of European grayling was evident. If an effect is assumed, this effect was only minimal compared to other sources of variations.
Similarly, Espa et al. (2012) and Espa, Crosa, Gentili, Quadroni, and Petts (2015) found no adverse effect on the fish community of the Adda River (Italy), when flushing operations were ecologically adjusted and maximal sediment concentrations were 2.7 and 4.8 g/L, respectively, during the operation. Gutzmer, King, and Overhue (1996), Gutzmer, King, Overhue, and Chrisp (2002) reported minimal impacts to fish populations after operational adjustments of Spencer hydro sediment management. These results and comparable findings (e.g., Gerster & Rey, 1994) suggest that an ecologically adjusted reservoir management significantly reduces negative impacts on fish populations. Ecological adjustment often includes recommendations for SSC and duration of flushing events, as well as hydrological and seasonal restrictions. These adjustments were also undertaken at the Mur River reservoirs. Additionally, a postflushing with clear water was undertaken within the residual flow reach downstream of HPP Fisching to keep higher shear stresses and thus leading to a faster evacuation of sand and finer fractions. The measured values of SSC (overall mean: 2.3 ± 2.2 g/L; overall max: 6.1 ± 3.8 g/L) for a limited time (mean duration: 42 ± 14 hr; peak < 1 hr) seem to be within the tolerance limits of grayling. In the laboratory, salmonids generally tolerate SSC of >10 g/L for several days and SSC of >1 g/L over a month without significant mortality (e.g., Lake & Hinch, 1999, Michel, Schmidt-Posthaus, & Burkhardt-Holm, 2013. We also found no elevated fine sediment content in the uppermost sediment layers below reservoirs, and thus, long-term effects on spawning habitats due to deposition of the flushed fine material seem to be insignificant.
In contrast to ecological adjusted flushing operations, the ones ignoring ecological consequences generally report much higher SSC values and/or were accompanied by low oxygen concentra- on fish densities in the discussion of their results, but state that these effects are of only marginal importance compared to seasonal variations.
A drawback of many studies is a flawed study design, that is, missing control sites; thus, results are often only indicative.

| Site specific differences
Our analyses revealed significant inter-site variability, although this variability was smaller than the high temporal variability caused by floods. To depict spatial differences thoroughly, a more balanced database would have been necessary. Nevertheless, the data allow individual paired samples comparisons. These analyses indicate higher total densities in the stretches "04 Unzmarkt-Judenburg," "05 RW Fisching" and "06 Fisching-Pregbach" and comparably lower ones in the control site and in stretch 02 "St.
Georgen-Murau." YOY densities tend to be higher in "05 RW  The high total and YOY densities in the residual flow reach "RW Fisching" may be explained by its high hydromorphological status together with a sufficient and "stable" flow; that is, the majority of the natural flow dynamics which may act as disturbances are more or less excluded from the reach. Nevertheless, high flows which induce morphological dynamics are still active allowing a naturelike morphological development.

| Cormorant predation
Our data showed no signs of predatory effects of Great Cormorant on grayling. This is in contrast to many studies blaming cormorants for reducing grayling populations (e.g., Steffens, 2011). Cormorants were also often cited as the most critical threat to grayling populations in Austria as well as in the Mur River (Woschitz & Parthl, 1997).
The reasons for the low impact may be that-although there are night roosts nearby-only few individuals forage in the Upper Mur River, most cormorants feed in the reservoirs of the Lower Mur (e.g., Ringert, 2005). Furthermore, the average size of fish captured and ingested by Great Cormorant is significantly lower than the size at maturity. Cech and Vejrik (2011) reported a mean TL of preyed fish of 13.0 cm and fish ≤20 cm TL comprised more than 90% of the cormorants' diet. This size class comprises to a large extent YOY, which are generally vulnerable to winter stressors (Hurst, 2007) and bird predation seems thus to work within the range of compensatory mortality.

| Variable recruitment and density dependence
Our analyses did not support hypotheses related to flushing, but were consistent with patterns resulting from variable recruitment and density-dependent effects. Two independent data sets (PAS, YOY) revealed hydrology and temperature as the main physical parameters affecting grayling recruitment. The same result was found by Charles, Mallet, and Persat (2006) in the Ain River. Their population model also revealed hydrology and temperature as the main variables affecting recruitment in European Grayling. Hydrology generally seems to be the main factor governing survival in early life history of salmonids in natural and nature-like alpine streams. Unfer, Hauer, and Lautsch (2011) for example showed that high flows during incubation and emergence were negatively correlated with recruitment success of brown trout in the Ybbs River, a finding also supported by Cattanéo, Lamouroux, Breil, and Capra (2002) in their study on population dynamics of brown trout in 30 French stream reaches.
After the critical emergence and larval period, population dynamics are regulated by density-dependent effects, that is,  Suter (1995) who found strong density dependence in grayling populations of the River Rhine. Density-dependent compensatory processes have also been found in other potamodromous salmonids such as Salmo marmoratus (Vincenzi, Crivelli, Jesensek, Rubin, & Leo, 2007) and Salmo trutta fario (e.g., Richard, Cattanéo, & Rubin, 2015).

| Spawning habitat
A highly important factor, in view of an ecologically adjusted flushing management, is the supply of gravel to the free-flowing river stretches downstream from HPPs, as this enables the formation of new and loose uncolmated gravel substrates, which constitute a major prerequisite for the preservation of a river-type-specific fish and benthic fauna. This management strategy has already been proposed by Habersack (1996)  These limitations led on the one hand to a discontinuous sediment management throughout the chain of reservoirs, on the other hand to extended periods without flushing especially at the HPP Fisching.
To overcome this shortcoming, we propose an adapted management scheme which aims at more frequent flushing. This may be accomplished by relaxing some seasonal and hydrological restrictions to achieve a periodic flushing regime in concordance with the hydrograph thereby mimicking the timing, magnitude, frequency, and duration of natural SSC pulses and gravel transport.
During small floods, we recommend drawdown routing (or sluicing) to minimize sand and silt deposition. Sluicing permits fine sediment to be transported through the reservoir rapidly to avoid sedimentation, and channel erosion may occur locally at the head of the reservoir. The required discharge and drawdown magnitude for sluicing depends on reservoir geometry and hydraulic parameters, in particular bed shear stress and particle size distribution.
During high flows-that is, whenever erosion can be expecteddrawdown flushing involving extensive scouring and resuspending of sediments should be undertaken. In contrast to sluicing, whose aim is to pass sediment without allowing it to deposit, drawdown flushing focuses on scouring and resuspending deposited sediment and transporting it downstream. It involves the complete emptying of the reservoir to freely pass the flushing discharge through the dam without upstream impounding (Kondolf et al., 2014).
Additionally, out-flowing sediment loads should be regulated by reducing drawdown celerity during the transition phase of drawdown and free-flow.
We suggest to dispense flushing from April to June when a drawdown flushing was performed the preceding year. The reason therefore is-although an effect of flushing on YOY was not evident according to model ranking-that eggs and larvae are often noted to be especially sensitive.
This flushing regime minimizes adverse downstream environmental impacts and maximizes benefits: A large proportion of the fines will be transported through the reservoir even at small floods so that the natural pattern of sediment discharge is approached.
During large floods, flushing additionally allows for remobilization and transport of coarse material improving habitat conditions downstream through delivery of gravel.

ACK N OWLED G M ENTS
The analyses are partly based upon data obtained during the to writing of the manuscript, and gave final approval for publication.

DATA ACCE SS I B I LIT Y
Upon acceptance, all data supporting this study will be provided as Supporting Information accompanying this paper or be archived in an appropriate public archive.