Species‐specific fish larvae drift in anthropogenically constructed riparian zones on the Vienna impoundment of the River Danube, Austria: Species occurrence, frequencies, and seasonal patterns based on DNA barcoding

As a result of river regulations over several centuries, followed by restoration measures in recent decades, most of the River Danube shoreline is man‐made, primarily riprap, but some reconstructed gravel banks and riparian side arms. We investigated the effects of these different structures on fish larval dispersal over a 20‐km stretch in Vienna via the use of drift nets. The habitats examined were created 18 years ago when the impoundment of the Danube hydropower station Vienna/Freudenau was constructed. About 15,000 fish larvae were trapped, and a subsample was determined to species level by DNA barcoding. In total, 26 different species were detected, including 10 species that are endangered or in danger of extinction. When species composition was considered, cyprinids become dominant at sites downstream of gravel bars, whereas in riprap sections, the majority of the larvae consist of invasive Gobiidae. Side arm habitats provide spawning and nursery grounds for additional species. Furthermore, clear species‐related seasonal patterns were observed with peak densities and multiple spawning periods of some species being recorded. The largest peak of Percidae occurred in the first half of May, followed by Cyprinidae at the end of May and Gobiidae in mid‐June.

The construction of the run-of-river power station Kraftwerk Wien/ Freudenau between 1992 and 1998 was the last large-scale river engineering works undertaken on the Austrian Danube and included several environmental compensatory measures. The previously straight shoreline was reconstructed by creating backwaters, coves, gravel banks, and pools. Subsequently, further attempts have been made to restore the shorelines and provide ecologically functional habitats. These restored sections provide habitats for a wide range of fish species and different life stages (Straif, Waidbacher, Spolwind, Schönbauer, & Bretschko, 2003).
Within the present study, the artificial shoreline configurations were sampled with drift nets to evaluate their contributions to fish larval dispersal in the River Danube. Species composition of early life stages of fish indicates spawning ground quality within the upstream sections of the river (Humphries & Lake, 2000;Pavlov, 1994). This knowledge is of exceptional importance, as functional spawning grounds and nursery habitats are considered to be limiting factors for riverine fish populations in the contemporary River Danube (Jungwirth, Haidvogel, Hohensinner, Waidbacher, & Zauner, 2014;Jungwirth, Haidvogl, Moog, Muhar, & Schmutz, 2003). In view of these conditions, a clear species identification is essential but also challenging because during the early life history of fish, morphology changes quickly and significantly during development (Balon, 1981) from preflexion larvae to postflexion through to the prejuvenile stage. As a result misidentification of species is likely for both rare and common taxa (Ko et al., 2013). In the last decade, DNA barcoding has become the method of choice for definition of different groups of biota (Hebert & Gregory, 2005). DNA barcoding uses a short genetic marker in an organism's DNA to identify it as belonging to a particular species. Thus, DNA barcoding was chosen for identification of fish larvae as it is currently the most reliable and reproducible method (Pegg, Sinclair, Briskey, & Aspden, 2006;Ward, Zemlak, Innes, Last, & Hebert, 2005). The current study is the first time DNA barcoding has been used to confirm the identification of River Danube fish larvae to species level in Austria.
However, relatively few surveys have been conducted to demonstrate the functionality of such an approach worldwide Geist & Hawkins, 2016;Lechner et al., 2013;Palmer et al., 2005;Pander & Geist, 2013;Pander, Mueller, Knott, Egg, & Geist, 2017). In view of these conditions, the present study focuses on species-specific fish larval drift associated with different shore structures in a highly modified section of the River Danube, upstream of the hydropower plant (hpp) Freudenau/Vienna. Three different constructed shoreline configurations: gravel bars, riparian side arms, and monotonous riprap sections, were studied over 2 years to gain information about (a) the functioning of spawning areas upstream of the sampling points; (b) species-specific differences in their contributions to fish larval dispersal; and (c) seasonal variation in drift densities.
A detailed description of the location and site is given by Chovanec, Schiemer, Waidbacher, and Spolwind (2002), Straif et al. (2003), and Waidbacher et al. (2016). The two gravel bars, "Donauinsel" (no. 1) and "Hügelland" (no. 2), have also been anthropogenically constructed. The riparian shoreline is fixed with riprap, whereas additional submerged riprap prevents the gravel bar from being eroded into the main channel ( Figure 1d). The gravel bar furthest upstream (no. 3) was initiated by the construction of a groin field. The two riprap sections examined (nos 4 and 5) are located in the central impoundment upstream of the side arms. Riprap no. 6 is located in the upper part of the impoundment, and no. 7 is located in an almost free-flowing section.

| Field sampling
Early life stages of fish were sampled continuously from April to July 2013 on 39 occasions and on 24 occasions in 2014, approximately twice a week. Drift nets similar to Zitek, Schmutz, Unfer, and Ploner (2004) with two equilateral triangles with a side length of 37 cm per sampling device were used. Each net had a mesh size of 1 mm, a mouth opening of 592.8 cm 2 and a total length of 550 cm. The last 50 cm were detachable to allow the caught larvae to be emptied. Drift nets were exposed for 24 hr on each occasion.

| Identification of early life stage fishes and DNA barcoding
Fish larvae are difficult to morphologically determine to genus or species level (Ko et al., 2013). Therefore, trapped fish larvae were divided into "similarity groups" based on simple morphological features (e.g., body shape, pigmentation, and head proportion) by using specific keys and literature (Balinsky, 1948;Pinder, 2001;Spindler, 1988;Urho, 1996). A total of five groups linked to the family taxonomic level were defined: Gobiidae/Cottidae (drop shape in dorsal view; often big roundish ventral fin visible), Cyprinidae p. (pigmented; fragile, elongated, body shape), Cyprinidae n.p. (not pigmented; fragile, elongated body shape), Percidae (more massive body shape, often two dorsal fins visible), and Undefined (e.g., damaged and fragments of larvae).
Fish were caught in all larval stages, and juveniles were excluded. Following processing, a sample of 671 larvae was analysed using DNA barcoding to species level (Hebert, Cywinska, & Ball, 2003). The selection criteria were the abundance of each "similarity group" recorded at a sampling site for each calendar week and the potential number of species per group/family. The latter represents the proportion of potential species hidden in the five "similarity groups" (Cyprinidae p.

| Data analyses
The mtDNA verified information of the 671 individual species identifications was then proportionally calculated for the entire dataset of the 14,555 individuals caught from the sample sites across the morphological group affiliation. This was done to all seasonal and spatial aspects of species-specific patterns to be analysed (Tukey, 1977). All data presented in the result section are based on the genetic verified species level information. For the asymmetric confidence ranges (Figures 2 and 3), a random sample of all genetically analysed specimens was selected for the calculations (α = 0.1). The affiliation to guild follows Schiemer and Waidbacher (1992) and Zauner and Eberstaller (1999) and was slightly expanded for N. melanostomus, which is considered as a speleophilic species (Kottelat & Freyhof, 2007).
Cyprinidae (13-20%) and Percidae (7-13%) occurred less frequently in catches. However, there were also minor differences between the sampling sites along one shoreline ( Figure 2; Table 2). The riprap "free proportions of Gobiidae. Over 50% of the larvae still comprise cyprinids, mostly Nase, Barbel, and Asp, around one third were Round Gobies, and a small proportion of Percidae was recorded.
In the side arm habitats, very few Gobiidae and Cottidae were caught. The majority or individuals recorded were Cyprinids. Roach were dominant in "Habitat C," where remarkably some Carp (Cyprinus carpio) and Pike (Esox lucius) larvae were also recorded; the latter being

| DISCUSSION
Artificially constructed shorelines provide functional spawning grounds in large rivers , which can be assessed by the occurrence of early life stages of fish (Pavlov, 1994). Speciesspecific information is necessary to evaluate their particular contributions to fish larval dispersal within a river. Through the inclusion of species information, for example, it became apparent that gravel bars provide suitable spawning grounds for lithophilic species, whereas the side arms were rich in phyto-and litho/phytophilic species  (Lechner, Schludermann, Keckeis, Humphries, & Tritthart, 2010). However, the present study indicates that differences within one shoreline type may also occur. These differences are even stronger when compared at the species level ( Table 2). suitable habitats for all life stages (Brandner et al., 2015;Roche, Janač, & Jurajda, 2013). Consequently, a potential measure to reduce its abundances is to remove riprap where possible. In addition, such structural alterations affect the hydraulics of inshore areas, which may have dramatic effects on the dispersal and viability of native fish populations (Lechner et al., 2014;Schiemer, Keckeis, & Kamler, 2002).
Our results are in line with other studies (Lechner et al., 2013;Ramler, Ahnelt, Nemeschkal, & Keckeis, 2016) that report that the near natural shores provide substantially more suitable larval habitats for native fish fauna than anthropogenically stabilized shores.
One noteworthy finding is that the abundance of Bighead Goby increased compared with Round Goby from the head of the impoundment (0%) to the central impoundment (75% of Gobiidae). This might be due to changing habitat characteristics such as the reduction of flow velocity, reduced sediment loads, and constant water levels (Jungwirth et al., 2003;Ward & Stanford, 1983), which facilitated and increases the reproductive success of Bighead Goby.
Although the exact origin of the larvae caught remains unclear, our findings still indicate that some larvae drifted long distances and some probably hatched just upstream of our sampling points. An example of long-distance drifters, which were detected, was Whitefish larvae (Coregonus sp.), which were evenly distributed across the entire investigation area. There have been debates regarding the existence of a self-sustaining Whitefish population in the Danube for a long time (Holcik, 2003). In many fish surveys on the River Danube, only adults have been recorded, probably derived from stocking activities in impoundments upstream for recreational fishery purposes (Holcik, 2003;Jungwirth et al., 2014). The repeated capture of Whitefish larvae demonstrates that these spawned in the River Danube, although the evidence for the successful completion of the complete reproductive cycle is questionable as all the larvae were dead and there are no reports of juvenile Whitefish being recorded in the study area.
Within our study sites, there was a gradual increase in the number of species recorded on the course of the river, indicating drift for several kilometres downstream. This is also supported by the results of Gravel bar "Donauinsel," which differed from the sites due to the presence of high proportions of typical riprap species, probably originating from the 10-km-long riprap stretch just upstream of the 2km-long gravel bar. The undercut slopes in this area are characterized by high current speeds and turbulence, which probably exceeds swimming capacities of recently hatched larvae (Webb & Cotel, 2011;Wolter & Sukhodolov, 2008) and therefore results in greater drift distances (Corbett & Powles, 1986). However, the differences between gravel bar and riprap sections located in series are particularly pronounced, indicating nearby sources of the fish larvae. For the purpose of ecological river management, there is a pressing need for further research to determine drift distances of different species, in order to detect spawning grounds so that sites downstream can be designated as protected areas (Lechner, Keckeis, & Humphries, 2016). Regardless of the drifting mode of the fish larvae (Pavlov, 1994;Pavlov, Mikheev, Lupandin, & Skorobogatov, 2008), we found evidence that ( (Lechner et al., 2013;Pander, Mueller, & Geist, 2015;Pander, Mueller, Sacher, & Geist, 2016). The demonstration of the reproductive success of carp (C. carpio) within one of the artificially built side arms is particularly noteworthy, given that self-sustaining wild carp populations in the River Danube are considered particularly rare (Schiemer & Waidbacher, 1998).
In total, the 26 verified species represent nearly half of all species that have been sampled in this area between 2013 and 2015 (Waidbacher et al., 2016). This does not necessarily mean they do not reproduce here, as they may either spawn at other sites or avoid drifting. This may be the case for Bleak and Chub, which are abundant as juveniles and adults (Waidbacher et al., 2016) but are rarely caught or recorded at the larval stage as they have a negative propensity to drift (Reichard & Jurajda, 2007). Further research centred on the River Danube fish fauna is necessary, and the application of a classification proposed by Humphries and King (2003) characterizing the relevance and propensity to drift will improve interpretation of data.  (Der, 1992).
The reproduction and records of larvae of numerous protected and endangered species highlight the importance of these anthropogenically constructed inshore restoration structures.
All of the species recorded displayed a specific drift period. Similarly to the findings of  and Janáč, Šlapanský, Valová, and Jurajda (2013), repeated occurrences of early larval stages in drift were observed. This indicates repeated spawning events for some species as the appearance in drift is directly linked to the timing of reproduction (Brown & Armstrong, 1985). In both years of the investigation, records started with 2 weeks of zero catches, clearly highlighting the start of larval drift in the middle of April.
Seasonality and duration of drifting were generally specific for each species. Most species appeared for 3 to 4 weeks, whereas some displayed an extended drifting periods of up to 8 weeks. Other studies in this area have recorded similar seasonal patterns, even though most of them did not cover the entire drifting season and therefore missed the peaks of either the early drifters (e.g., Bullhead in April) or those last to drift (Lechner et al., 2010, Ramler et al., 2016. In the last 2 weeks of July, only a small number of Chub, Round Goby, and Barbel were recorded, indicating the end of drifting for most species. Other studies have reported drifting periods through to September, especially for the invasive Round Goby (Borcherding et al., 2016, Janáč et al., 2013. The knowledge generated on the seasonal variability of drifting linked to spatial variation can be used to help inform conservation measures. For example, navigation or other activities could be modified in areas were fish reproduction of endangered species occur during their drifting season, as these practices have negative effects on the survival rates of fish larvae (Pavlov et al., 2008;Wolter & Arlinghaus, 2003).

| CONCLUSION
This study indicates that the artificial shoreline areas investigated, riprap, gravel bar, and side arms are potentially used as spawning grounds for riverine fish species. Furthermore, these different shoreline configurations determine the species composition of fish larval dispersal in the River Danube with a species-specific periods of drifting. The relevance of the habitat mitigation measures examined (gravel bar and riparian side arms) highlights the apparent reproductive success of numerous protected and endangered species. The results of this study therefore provide the basis for effective conservation and management of riverine fish populations. Furthermore, the effect of monotonous riprap shorelines on the spatial distribution and potential spread of the invasive Gobiidae is clearly documented.