The inﬂuence of discharge, current speed, and development on the downstream dispersal of larval nase ( Chondrostoma nasus ) in the River Danube 1

: We investigated the mode (active versus passive) of larval downstream dispersal and its inﬂuencing factors in the nase carp ( Chondrostoma nasus ). Marked larvae (early and later stages), together with equivalent numbers of passive particles, were released in the main channel of the River Danube (Austria) at different ﬂow (low, high) and current (over-critical, under-critical) conditions. Larvae and particles were recaptured with stationary nets at varying distances from release. We assumed that differences in the spatial dispersal patterns between larvae and particles were due to ﬁsh activity. We hypothesized that river discharge, developmental stage, current speed, and distance from release would inﬂuence these differences. We found that activity was independent of developmental stage or current speed at release, although activity was higher during low ﬂow conditions. It may be that larvae deliberately enter the current during low ﬂow, because the hydraulic conditions facilitate active dispersal. Furthermore, activity was greatest near the release site. This might be due to an intrinsically greater activity when ﬁsh are placed into novel surroundings or a result of rheoreaction. The discharge-dependent dispersal patterns observed represent an important ecological link between ﬂow and recruitment and demonstrate the importance of inshore conditions for the early life stages of ﬁsh in large rivers, especially with regard to river modiﬁcation and restoration schemes. Résumé : Nous avons étudié le mode (actif ou passif) de dispersion des larves vers l’aval et les facteurs qui l’inﬂuencent chez le hotu ( Chondrostoma nasus ). Des larves marquées (stades précoces et plus avancés) et un nombre équivalent de particules passives ont été relâchées dans le chenal principal du ﬂeuve Danube (Autriche) dans différentes conditions d’écoulement (débit faible, débit élevé) et de courant (supercritique, sous-critique). Des larves et des particules ont été capturées dans des ﬁlets stationnaires a` différentes distances du lieu de lâcher. Nous sommes partis de l’hypothèse que les différences des motifs de dispersion spatiale entre les larves et les particules étaient dues a` l’activité des poissons. Nous avons postulé que le débit du ﬂeuve, le stade de développement, la vitesse du courant et la distance par rapport au lieu de lâcher inﬂuenceraient ces différences. Nous avons constatéquel’activité étaitindépendantedustadededéveloppementoudelavitesseducourantaulieudelâcher,l’activitéétant toutefois plus forte durant des conditions de faible débit. Il est possible que les larves entrent délibérément dans le courant quand le débit est faible, parce que ces conditions hydrauliques facilitent la dispersion active. En outre, l’activité était la plus forte près du lieu de lâcher. Cela pourrait être dû a` une activité intrinsèquement plus forte quand les poissons sont placés dans un environnement nouveau ou en raison d’une rhéoréaction. Les motifs de dispersion dépendant du débit observés constituent un lien écologique important entre le débit et le recrutement et démontrent l’importance des conditions près des berges pour les stades précoces des poissons dans les grandes rivières, notamment en ce qui concerne les projets de modiﬁcation et de restauration de ces cours d’eau. [Traduit par la Rédaction]


Introduction
Movement is a fundamental process in the lives of virtually all organisms (Nathan et al. 2008).Therefore, understanding the temporal and spatial patterns of movement of individuals, species, and populations is central to effective conservation and management (Baguette et al. 2012).The movement (or dispersal) of early life history stages of fishes (i.e., eggs, free embryos, larvae) from spawning sites to appropriate rearing habitats affects the survival and fitness of the individual and, consequently, year-class strength, probability of completion of the life cycle, and population health (Daewel et al. 2011;Hinrichsen et al. 2012).The dispersal of eggs can generally be described using passively drifting particles (Dudley and Platania 2007) or 3D hydrodynamic models (Peck et al. 2009;Pacariz et al. 2014).For a long time, the same was also assumed for larvae, because of their presumed poorly developed behavioural, physical, and sensory capabilities (Brown and Armstrong 1985;Roberts 1997).Indeed, the early life stages of fishes are often referred to as "ichthyoplankton" (Muth and Schmulbach 1984;Bialetzki et al. 1999;Martin and Paller 2008), with all the implications of passivity that this term brings.
The earlier assumptions of passive dispersal by fish larvae were overturned by marine studies, revealing that larvae are capable of detecting fine-scale hydraulic, chemical, and auditory cues, and exhibit sophisticated behaviour (e.g., directional swimming), relevant for effective dispersal (e.g., travel distance or encounter rates with suitable nursery areas; Staaterman et al. 2012;Mouritsen et al. 2013).Reef fish larvae, for example, orientate towards odours and sounds of their natal reef and use directional currents to get there (Paris and Cowen 2004;Leis et al. 2011;Paris et al. 2013).Observations of larvae moving along hydraulic gradients (Stoll and Beeck 2012), adapting swimming and drifting speeds to the prevailing flow regime (Hogan and Mora 2005) or regulating depth distribution according to currents (Kunze et al. 2013), further emphasize an active response to the hydrodynamic environment.The interactive effects of temporally unstable hydrodynamics and ontogenetically variable larval behaviour are increasingly being used in oceanic dispersal models to gain more realistic simulation results (Vikebo et al. 2011;Sponaugle et al. 2012;Nolasco et al. 2013).
Comparable studies in rivers and streams are scarce, but understanding of patterns and processes are improving rapidly.The dispersal of young fish in lotic systems is primarily mediated by the current and is termed drift (Pavlov 1994).Most riverine species spawn demersal eggs (Balon 1975), and initial drift peaks occur when recently hatched individuals emerge from the substrate and enter the current (Persat and Olivier 1995).Ideally, these fish reach suitable inshore settlement areas where environmental conditions match rearing requirements (Schiemer et al. 2002).Subsequent drifting events between adjacent nursery habitats serve to optimize growth and minimize predation or competition (Pavlov 1994;Humphries 2005).Generally, drift in rivers follows a circadian rhythm, peaking after dusk (Johnston et al. 1995;D'Amours et al. 2001).The nocturnal drift entry of young fishes is supposed to rely on active behavioural decisions (Reichard et al. 2002;Boujold et al. 2004;Sonny et al. 2006).On the other hand, there is still a general perception that the transport of free embryos and larvae per se is a passive process, caused by a negligible swimming performance and overriding environmental forces (see Lechner et al. 2016).The critical swimming speeds that young fishes can maintain increase with body size and developmental stage (Flore and Keckeis 1998;Flore 2001;Kopf et al. 2014), but are limited to values <0.15 m•s −1 during the early ontogeny of many freshwater species (see Wolter and Arlinghaus 2003).Accordingly, the average flow velocities in rivers are considered over-critical for free embryos and larvae (i.e., the individuals cannot hold position and are washed downstream; Bardonnet 2001;Wolter and Sukhodolov 2008).This process is intensified during elevated discharge due to the associated increase in current speed, turbulence, and turbidity (Harvey 1987;Bischoff and Wolter 2001;Reichard and Jurajda 2004), potentially resulting in high population losses (Mion et al. 1998;Peirson et al. 2008).
In light of recent findings, however, equating riverine larvae being transported downstream to passive particles is not appropriate.Drift surveys in the Austrian Danube showed that travel paths of larvae exhibit family-specific correlations with hydraulic forces, even at over-critical flows (Lechner et al. 2014a), and clearly deviate from those of modelled passive particles (Schludermann et al. 2012).Consequently, passive transport models in rivers overestimate by a long way the covered distance of drifting larvae (Braaten et al. 2012).It seems reasonable to conclude from this recent evidence that active larval drifting behaviour in rivers has evolved to facilitate movement, minimize mortality while drifting, and fulfil physiological and other requirements through locating and settling in appropriate rearing habitats (Robinson et al. 1998).
The overarching objective of this experimental study was to expand our understanding of the nature of the mode of driftingactive versus passiveof young fishes in running waters.Specifically, we aimed to clarify the influence of river discharge, current velocity, and developmental stage (and related swimming performance) on the dispersal pathways of drifting larvae.This experiment is the first attempt to directly compare larval dispersal with passive diffusion in their natural environment.The results are expected to influence the design of the first individual-based larval dispersal model in rivers (see Glas et al. 2017), as well as future river management measures.
An in situ comparison between the spatial dispersal patterns of marked fish larvae (common nase, Chondrostoma nasus, Cyprinidae) and passively drifting particles was performed in the main stem of the Austrian Danube.The study was conducted in two consecutive years (2011,2012) with different flow regimes (low and high), providing contrasting hydraulic environments.Within each year, early and later larval stages were simultaneously released at two adjacent points, featuring over-critical and undercritical current conditions, together with the same number of particles.The spatial dispersal patterns were surveyed with stationary drift nets at several sampling stations further downstream.We assumed differences in recapture rates of fish and passive particles arose from larval activity, which has been discussed elsewhere (Metaxas 2001), and hypothesized the following: (i) differences in recapture rates of fish and passive particles will be smaller during elevated than low discharge, because the harsh hydraulic conditions at high flows limit the ability of larvae to influence their movement and force larvae to drift more passively; (ii) differences in recapture rates of fish and passive particles will be smaller for earlier than later larval stages, because the limited swimming performance of less-developed individuals form the basis of passive dispersal; (iii) differences in recapture rates of fish and passive particles will be smaller at the over-critical than the under-critical release points, because strong currents, exceeding the larvae's swimming speed, limit the ability of larvae to influence their movement and force larvae to drift more passively; (iv) differences in recapture rates of fish and passive particles will increase with increasing distance from the release points, because larval activity patterns will, with time, increasingly differentiate themselves from purely passive transport.

Study area
The experiment was performed in the main stem of the Austrian Danube, between river kilometres 1890.0 and 1893.8 (Fig. 1).At the study site, the mean values for river width, water depth, and flow velocity were approximately 300 m, 3 m, and 2.5 m•s −1 , respectively.Although located in a national park area (www.donauauen.at),this free-flowing river stretch is characterized by profound man-made alterations.The right shoreline is formed by artificial, steep, and straightened embankments (rip-rap).Additionally, rock wingdikes (groynes), perpendicular to the main flow direction, support navigability and prevent bank erosion.The left shoreline, where this study was conducted, has been revitalized in the course of a river restoration project between 2007 and 2009 (Tritthart et al. 2014).The rip-rap was removed and the original groynes were modified (Fig. 1), enabling a bankside flow along a newly formed gravel bar, which serves as a dispersal corridor for young fish (Lechner et al. 2014b).The modified groynes submerge at a discharge ϳ1200 m 3 •s −1 .
The study area is characterized by a diverse fish fauna, including several endangered species (Schiemer and Spindler 1989), such as the nase carp.The nase, a common rheophilic cyprinid in the Danube, has become a flagship species for the conservation of large European rivers (Schiemer et al. 2002).This is, inter alia, due to its life cycle; recently hatched larvae drift from swift-flowing spawning sites to littoral nursery habitats (Persat and Olivier 1995;Keckeis et al. 1997;Reichard et al. 2001), meaning that river conditions are critical to the survival of this species early in its life.

Larval supply and marking
The acquisition of fish larvae and tagging methodology are described in detail elsewhere (Lechner et al. 2014b).Briefly, adult nase carp were obtained from a natural spawning population by electrofishing and were hand-stripped in the laboratory.The eggs of eight (2012) to ten (2011) females were mixed with the sperm of several males, using the dry method.Fertilized eggs were divided into two oxygenated rearing flumes, with different temperature regimes (cold flume: 11.7 ± 0.7 °C in 2011 and 11.4 ± 0.8 °C in 2012; warm flume: 14.8 ± 3.3 °C in 2011 and 16.8 ± 1.2 °C in 2012).As higher water temperatures accelerate tissue differentiation and growth (Seikai et al. 1986;Pepin 1991), individuals from the warmer flume were more developed (fourth larval stage, L4) and larger (mean standard length: 12.7 ± 1.8 mm in 2011 and 14.3 ± 0.9 mm in 2012) at the day of release than their conspecifics from the cooler flume (second larval stage, L2: 11. 7 ± 1.9 mm in 2011 and 13.4 ±  0.8 mm in 2012).L2 individuals were characterized by a large fin fold, instead of differentiated ventral, dorsal, and anal fins (Fig. 2).Many larvae had a truncated posterior margin of the caudal fin, in which the lepidotrichia started to ossify.According to Penaz (1974), L2 nase already show rheophilia.Although the embryonic fin fold was still present at the fourth larval stage, the ventral, dorsal, and anal fins began to separate within it (Fig. 2).L4 nase had ossified rays in the incised homocercal caudal fin.According to Penaz (1974), these fish display enhanced nimbleness and rapid movement.Swimming performance linearly increases with body length (Flore et al. 2001) and is boosted by the differentiation of fins and the postcranial skeleton (Ott et al. 2012).Accordingly, the spatial drift patterns of later stages should reflect this improvement.
All larvae were tagged with a chemical marker (Alizarine Red S; Sigma Aldrich).The application of systematic multiple staining events created unique sequences of fluorescent rings in larval otoliths (Fig. 2).Using this labelling key, a recaptured individual could be allocated to a release point (i.e., inshore release, IR; offshore release, OR) and a developmental stage (L2, L4).The effectiveness of the marking procedure was tested several times on a random basis and found to be 100%.Following each staining event, individuals with identical marks were transferred to sepa- After 34 (2011) to 46 (2012) days in the laboratory, during which larvae were daily fed with nauplii larvae (Great Salt Lake Artemia Cysts, Sanders) and dry food (Vipagran Baby, Sera), fish were prepared for release.Two subsamples of 100 individuals were taken from one cold and one warm flume.These larvae were overdosed with tricaine methanesulfonate (MS-222, Sigma-Aldrich, St. Louis) and then weighed to 0.01 g.The mean mass was used to extrapolate larval counts from the total moist mass within each flume.All remaining larvae were packed into well-oxygenated polyethylene bags and transported to the study site.

Drift experiment
In both years of investigation, experiments were conducted on two consecutive days in May.Four drift sampling stations (SS1-SS4) were installed at increasing distances (65, 290, 400, and 510 m) downstream of the predefined release points (Fig. 1).Exact positions of the SS were mapped with a dGPS device (GS 20, Leica, St. Gallen).Each SS consisted of three parallel nets (inshore (IN), midshore (MID), and offshore (OFF); refer to Fig. 1) set in a row perpendicular to the shoreline.The distances between single nets were variable and adjusted to the prevailing hydraulic conditions at the SS, ensuring the security of equipment and people.The conical-shaped nets (length 1. 5 m, mesh size 500 m) featured a circular mouth (diameter 0.5 m) and a removable plexiglass jar at the cod end (diameter 7.5 cm).They were fixed to stationary iron rods with a mountaineering cord (length 2 m), making them flexible in the current.A mechanical flowmeter (2030R, General Oceanics, Miami), attached to the net opening, measured the volume of filtered water.
Larvae were acclimated to the temperature of the Danube by gradually adding river water to the transport bags for 20 min.
Afterwards, the bags were positioned at the exact locations of release, which were a low-flow inshore area (IR) and a point further offshore in the mainstream (OR).The release sites of both years are shown in Fig. 1.Corresponding values for flow velocity and water depth are given in Table 1.Immediately before release, passive particles (P) (polyethylene pellets, diameter 4 mm, density 0.93-0.95g•cm −3 ; see Fig. 2) were added to the bags equal to the number of contained larvae.Particles were black at the IR and white at the OR.To match the natural circadian rhythm of drift (Zitek et al. 2004), larvae (and particles) were released at sunset, 2015 h.
Drift sampling (deployment of drift nets) at all SS started synchronously with release and was carried out for 30 min and thereafter repeated in hourly intervals over 3-5 h periods.All samples were preserved in 96% ethanol and taken to the laboratory for continuative analysis.

Sample processing and statistical analysis
Each sample was suspended in a water bath to separate larvae and particles from organic flotsam and then manually sorted.All nase larvae were extracted from the samples and checked for reference marks.Therefore, lapilli (located most anteriorly in the pars superior) were dissected and embedded in epoxy resin (Crystalbond, Aremco, New York).Afterwards, these otoliths were ground, polished, and screened under wavelengths of 515-565 nm with an epifluorescence light microscope (Zeiss, Axio Imager M1 with Axio Vision 4.8.2 software for image analyses).A selection of different mark sequences is shown in Fig. 2.
If the number of potential recaptures in the sample was ≤5, all larvae were checked for marks.Otherwise, subsamples were inspected and the overall number of recaptures was extrapolated by where N RC is the total number of recaptures, N MS is the number of marked individuals in the subsample, N S is the number of larvae in the subsample, and N PR is the number of potential recaptures in the sample.Subsampling was not performed for particles, which were all counted.Then, recapture rates (RR, in %) of larvae and particles from each subgroup (i.e., L2_IR, L2_OR, L4_IR, L4_OR, P_IR, and P_OR) were calculated for the single drift nets: where N RL equals the number of released fish or particles.We regarded a difference (DIFF) in the recapture rate of larvae (RR L ) and particles (RR P ), deriving from the same release points within each net, as a measure of fish activity or behaviour.We used absolute values of DIFF, because negative or positive values indicate fish activity or behaviour relative to passive particles (i.e., preference or avoidance).This procedure required a pretreatment of the data to integrate samples when either larvae or particles were absent from the nets.In such cases, DIFF was always 1, irrespective of the actual difference in the RR of larvae and particles.This information, however, is essential for our study questions.The addition of a constant amount (i.e., 0.001), lower than the smallest value in the data (i.e., 0.007), to all RR constituted a feasible solution to keep the original differences and to relate DIFF to these values.Afterwards, DIFF ranged between 0.000 (same RR of larvae and particles) and 0.998 (large differences in RR) and was positively correlated to the numerical difference in larvae and particle counts within the nets (Spearman rho = 0.67, P < 0.001).
DIFF was calculated as DIFF ϭ Խ (RR L ϩ 0.001) Ϫ (RR P ϩ 0.001) (RR L ϩ 0.001) ϩ (RR P ϩ 0.001) Խ and displayed strong positive correlations with the overall recapture rates (RR L + RR P ) in the nets (which in turn were dependent on the distance between net and release point).
Generalized linear models (GLMs) were developed to test, according to the hypotheses, which factors (year, stage, release, and station) best explain variation in the dependent DIFF.The GLM procedure with a gamma-log-linking function in IBM SPSS was used, and the model selection was based on the lowest value of Akaike's information criterion (AIC).The factors year of investigation (2011,2012), sampling station (1-4), release point (IR, OR), and developmental larval (L) stage (L2, L4) were included as fixed factors, and the filtered volume of each sample (volume) was taken as covariable.
The hydrogeomorphological data (i.e., flow velocity and water depth) come from a fully 3D model (RSim-3D) of the sampling site (Tritthart and Gutknecht 2007).All figures and analysis were done in Arc Gis 10.0, SigmaPlot 12.0 and IBM SPSS 23.

Methodological considerations
We are aware that some of the observed differences might be because of the different characteristics (e.g., buoyancy, shape, etc.) of fish and polyethylene pellets.Downstream-drifting larvae, spread over the water column, are exposed to different regimes of turbulence, shear stress, and current velocity than floating particles (Tritthart et al. 2009).Nevertheless, the surface-current-driven particle transport should accurately reflect the main direction of flow and thereby a spatial (lateral) corridor of passive dispersal.Additionally, differences in the vertical distribution of particles and larvae should be buffered by drift nets, which covered the upper 65% of the water column in most cases (92% in 2011, 90% in 2012).It has been demonstrated that larval drift densities are highest near the surface (Johnston 1997;Gadomski and Barfoot, 1998).Regardless, the potential differences between fish larvae and plastic particles affect all our comparisons among factors (year, stage, release, and station) equally and so could be considered a systematic error at worst.

Environment
The discharge of the River Danube during the sampling in 2011 ranged from 1010 to 1155 m 3 •s −1 and was close to regulated low flow (i.e., the discharge that is exceeded 94% of the time; Fig. 3).All modified groynes were exposed.The re-established bankside flow passed a heterogeneous shoreline, with many shallow bays and debris dams acting as retention zones.Large slack water areas extended downstream beyond the remaining groyne structures.The median current speeds near shore (0-15 m), where larval drift densities are presumably highest (Reichard et al. 2004), were below the maximum sustainable water velocity of released nase (i.e., the current velocity at which fish cannot maintain their position in the water column for more than 2 min; Fig. 3; Flore et al. 2001), which is 0.103 ± 0.005 m•s −1 for the released larvae (total length = 13.0 ± 1.1 mm).
The discharge of the River Danube was distinctly elevated in 2012, relative to its level in 2011, ranging from 2050 to 2250 m 3 •s −1 , which is clearly above mean flow (Fig. 3).All modified groynes Note: See also Fig. 1 for the spatial arrangement and sampling site positions.
were submerged (Fig. 1), and normally lentic downstream were then flowing.The bankside flow passed the remaining gravel shoreline or a steep erosion bank, which formed the transition to the adjacent floodplain forest.Median values of current velocity were over-critical (greater than maximum sustainable water velocity) in 2012, at all measured distances from the shore (Fig. 3).
Flow conditions were under-critical at the IR and over-critical at the OR points, irrespective of the year (Table 1).Between years, though, differences in water depth and current velocity were higher at the OR than at the IR points.

General
In both study years, a total of 97 drift samples were taken, of which 76 contained a sum of 5255 marked nase larvae and 4168 plastic particles, and the remainder were empty.The overall RR of larvae and particles and those of single groups were all higher in 2012 than in 2011 (Table 2).However, this result is mainly attributable to one net (2012, SS1, offshore, exposed simultaneously with release), which captured 1349 marked fish (i.e., 28% of the overall RR for both years) and 1325 particles (i.e., 18% of the overall RR for both years), all deriving from the OR.By contrast, the mean RR per standardized volume (1000 m 3 ) of filtered water (i.e., RR_vol) of most groups and all larvae combined were higher in 2011 than in 2012 (Table 2).This difference was significant for total larvae (n = 76, Mann-Whitey U = 1369, P < 0.05).No interannual difference in the overall RR_vol of particles was detected (n = 76, U = 1230, P = 0.251).

Testing the hypotheses
When the RR of particles and fish larvae are compared for each year for the total catch, for the two developmental stages, and the two release points, a linear relationship between the RR should be observed if both the larvae and the passive drifting particles disperse in the same manner, whereas a disproportionally high or low RR of larvae should indicate a distinct deviation from a passive dispersal mode (Fig. 4).In all combinations, catches with distinctly higher and lower RR of fish larvae were observed; comparatively few catches revealed a direct proportionality.Overall, there seemed to be a relatively even spread of RR to the left and right of the diagonal in 2011, whereas there was a proportionally greater RR of larvae than of passive particles in 2012.This between-year trend seemed consistent for both larval stages and position of nets, except in the case of L4 larvae collected offshore, where in 2012 there were approximately similar number of samples with RR to the left and right of the diagonal.
In 2011, the situation at SS1 differed from all others and was characterized by a high percentage of catches with higher RR of larvae (Fig. 5).At the other three stations, located further downstream of the release point, the percentage of catches with a distinctly lower RR of larvae compared with the passive drifting particles was found.A low percentage of samples with similar RR of fish and particles was observed only at SS2 and SS3.In 2012, when the discharge was distinctly higher (see Fig. 3), the differences between the stations diminished; most samples at each sampling station had higher RR of larvae.Furthermore, lower percentages of catches with lower RR of larvae were observed,   whereas samples with similar RR of fish and passive particles were found throughout the sampling area at each station.
For almost all combinations of year, larval stage, release, and net position, the largest differences between RR of larvae and particles were observed at SS1, which was located 65 m downstream of the release points (Fig. 6).The differences were less downstream of SS1, but the patterns were inconsistent.The large differences of RR at SS1 were less pronounced for both stages of the inshore-released larvae.Also, no consistent pattern was apparent for differences in RR for inshore, midshore, and offshore drift nets.
The GLM results indicated that year was a significant explainer of DIFF, whereas stage and release were not (Table 3).The factor station was not significant in the test of the model effects, but the pairwise comparison of marginal means indicated significant differences in DIFF between SS1 and SS4.All interactions between factors (full model) were tested and were found to be nonsignificant.
Regarding the year, differences between fish larvae and particles were smaller in 2012 (Fig. 7).With respect to the sampling stations, the median of DIFF was highest farthest upstream (SS1; Fig. 7).A decreasing trend of the median of DIFF with increasing distance from release (i.e., from SS1 to SS4) was observed.

Discussion
Active and passive mechanisms of fish drift in rivers have been commonly postulated (Robinson et al. 1998;Humphries 2005;Janác et al. 2013) but rarely tested and observed in situ (but see Lechner et al. 2016).To date, studies have tracked drift movements by snorkelling (Kennedy and Vinyard 1997), related larval pathways to numerical particle tracing (Schludermann et al. 2012), and compared temporal dispersal patterns of young fish and passive floats (Lechner et al. 2014b).In summary, all results indicated active drifting behaviour and self-directed dispersal.The present paper highlights the first in situ observation of the spatial distribution of drifting larvae, opposed to an absolutely passive element of dispersal (i.e., particle transport).
Generally, our results show that larval nase disperse unlike passive particles.Below, our results are discussed in detail following the order of the formulated hypotheses.

Discharge
Our hypothesis has been confirmed, as differences in the spatial dispersal patterns of larval nase and plastic particles were smaller during high channel flow in 2012.Theoretically, this observation could be because of passive washout in 2012 or deliberate entering and active dispersal in 2011.
The first possibility, that larvae were washed downstream passively during high flow, could be because a rise in river discharge often involves a decline of current-reduced inshore habitats (Korman et al. 2004), which are preferably used by young nase (Keckeis et al. 1997), and if such refuges are scarce, strong currents during elevated discharge (Fig. 3) may induce extensive washouts of 0+ fish communities (Harvey 1987;Bischoff and Wolter 2001).Once entrained, regaining habitat and shelter is limited by the larvae's weak swimming performance (Wolter and Sukhodolov 2008), and a passive longitudinal displacement is likely.These findings are, however, equivocal, as other studies have detected no influence of elevated river discharge, and the concomitant increase in current speed, on drift abundance (Reichard et al. 2001;Reichard and Jurajda 2004).In addition, Fig. 6.Bubble plots of the differences of the recapture rates between passive particles and fish larvae at different sampling stations (SS) and drift net positions (IN = inshore, MID = midshore, OFF = offshore).Results are plotted within separate sampling years (2011,2012) for all larvae combined (Total) and subgroups defined by developmental stage (L2, L4) and point of release (inshore (IR), offshore (OR)).
higher washout rates in 2012 should be reflected in higher overall mean RR_vol of larvae.In fact, the mean number of drifting fish per 1000 m 3 was significantly higher during low flow in 2011, suggesting an alternative possibility.
The second possibility, that larvae deliberately entered the drift and dispersed actively under mild hydraulic conditions in 2011, could be because low flows, and the associated low values of turbidity, turbulence, and current velocity (Fig. 3; Rakowitz et al. 2014), promote orientation and swimming performance of fish larvae (Lupandin 2005;Liao 2007;Reeves and Galat 2010), ultimately resulting in a significant higher deviation from particle transport.The active drifting of nase larvae, along the investigated shoreline in 2011, was also indicated by their temporal dispersal patterns (Lechner et al. 2014b).It could be argued that the higher mean RR_vol of larvae in 2011 is a result of dilution effects during elevated discharge in 2012.This is contradicted by the finding that the mean RR_vol of plastic particles did not differ between years.
In summary, a significant influence of the river flow on larval fish drift was revealed.Low discharges may motivate young nase to enter the current, perhaps by providing favourable hydraulic conditions for active (and oriented) downstream movements.During high flows, larvae apparently stay inshore and avoid drifting.If entrained, however, they disperse passively.Generally, it has been shown that elevated river discharge at the sampling site is not associated with high washout rates of fish larvae.The revitalized gravel bar seems to be quite insensitive to discharge-driven changes in suitable shoreline habitat availability.This should be beneficial for the young stages of species with similar habitat requirements.

Developmental stage
Our hypothesis was not confirmed, as the overall differences in the spatial dispersal of drifting larval nase and plastic particles were not affected by stage.Surprisingly, earlier developmental stages (L2), with supposed weak swimming and orientation abilities, displayed clear deviations from passive transport.Our results are consistent with associated laboratory experiments by Zens et al. 2018; the authors investigated the drift mode of nase larvae in a racetrack flume and detected, even in high currents, a high percentage of active, oriented L2 fish.Apparently, the underlying capacities are likewise executed in the wild (e.g., a large river).In agreement with our study, dispersal simulations in the Colorado River also demonstrated that low swimming performances of young fish can result in high deviations from passive diffusion models (Korman et al. 2004).
Stage-specific peculiarities in (active) spatial dispersal cannot be derived from DIFF alone; drift patterns where younger individuals concentrate near the bank and older ones in mid-channel (Reichard et al, 2004), or vice versa (Braaten et al. 2010), could remain concealed.Such characteristics are revealed in the unequal distribution of L2 and L4 nase over the sampling area (i.e., in particular drift nets), potentially originating from differences in behaviour, orientation, and swimming.Studies on the upstream and downstream trajectories of young nase in a racetrack flume, for instance, found correlations between developmental stage (L2, L4) and the frequency of occurrence of predefined movement patterns (i.e., active upstream, active downstream, active-passive, passive) in a velocity gradient (Glas et al. 2017).
In summary, downstream drifting of the nase includes a strong active component, already shown during early ontogeny.Assuming this is representative for many other riverine species, "ichthyoplankton" is a misleading term, at least for their larval stages.

Current velocity
Our hypothesis was not confirmed, as the spatial deviations from passive transport were not smaller for those larvae released in high current velocities.Indeed, the maximum individual values of DIFF were recorded for L4 larvae released offshore and collected at the first sampling station.
Our results indicate that drifting nase larvae are active in overcritical flows.Correspondingly, the flume experiments of Zens et al. (2018) showed that small nase, even if they cannot hold position in strong currents, are able to reduce their downstream transport and perform lateral movements.In previous studies on larval nase dispersal in the Danube, individuals from the (overcritical) OR points were found in proximity to these points in inshore nets and habitats (Schludermann et al. 2012;Lechner et al. 2014b), indicating active movement towards the shore.In addition, Warner sucker larvae (Catostomus warnerensis), placed in high currents midstream, avoided downstream displacement by actively moving inshore (Kennedy and Vinyard 1997).
It is quite possible that the activity level of drifting larvae is unexpected high in swift currents; the concomitant increase of swimming speed with current speed has been shown in many studies on specific threshold values in swimming physiology (e.g., critical swimming speed, U crit ; Plaut 2001;Fisher et al. 2005).Potentially, larvae that are abruptly transferred from transport bags into over-critical currents intuitively try to resist it or to evade it by swimming with high power (in lateral directions).

Distance from release
Our hypothesis was not confirmed, as, contrary to our expectations, the highest median value of DIFF was recorded at SS1, closest to the point of release.
Behavioural experiments with juvenile fish have found that motor activity increases after being placed into a novel situation (Pavlov et al. 2008).This exploratory behaviour, in combination with the above-mentioned swimming reactions at release, might Note: Significant results are presented in bold. 2 = chi squared, df = degrees of freedom, p = error probability, 95% UCL, 95% LCL = value of the upper (U) and lower (L) confidence interval of the mean difference.
have caused the observed patterns in our study.Additionally, it is possible that nase larvae, after the first "shock" of release, voluntarily drifted passively.Experiments with reef fish have shown that larvae, as it becomes uneconomical to try to swim against the current, reduce swimming speed and go with the flow (Hogan and Mora 2005).
Overall, our results suggest that the downstream dispersal of nase larvae in the Danube can be classified as an active-passive process (according to Lechner et al. 2016).There is strong evidence that behavioural and physiological traits impact the spatial drift characteristics.It is, however, questionable whether these findings are transferable to other species.Active drifting and the associated behaviour may be a peculiarity of fluvial specialists, and less so in stagnophilous and indifferent species.Comparable experiments with representatives of the particular guilds are desirable.
With respect to the nase, insights gained by this study on the drift mode and the spatial drift patterns constitute convertible knowledge in species conservation and river management.For example, it can be assumed that (unpredictable) events of elevated discharge, common in regulated rivers (Robinson et al. 1998;Humphries et al. 2002), are disruptive, as they prevent active (and oriented) dispersal of larvae.The detected activity of drifting individuals in over-critical flows implies a high energy consumption (Wieser 1995).The availability of suitable settlement sites and a quick arrival at these locations might therefore be crucial for larval survival.However, such retention sites are especially scarce in regulated rivers (Schiemer et al. 2001).
Furthermore, our results suggest that upcoming mathematical models of nase drift distance and derived statements on habitat connectivity in rivers must incorporate a "behaviour-activity" term into their equations.A next logical step, the correlation of (positive and negative) DIFF values with the abiotic conditions (i.e., current speed, current direction, water depth, turbulence, etc.) upstream of drift nets, would allow for a specification of this term and shed light on the fundamentals of preference and avoidance.Then, more precise predictions on the spatial distribution of drifting larvae in the river cross-section could be made.This, in return, would increase precision on drift-based population estimates and aid the design of "larval-friendly" anthropogenic river structures.The negative impacts of water abstraction systems, power stations, and weirs on drifting fish are manifold (e.g., entrainment, injuries, stranding, stress, death) but might be reduced when built at sites were drift densities are minimal (Pavlov et al. 2008;Ellsworth et al. 2010;Bracken and Lucas 2013).

Fig. 1 .
Fig. 1.Study area during low flow (1010 m 3 •s −1 ) in 2011 and elevated discharge in 2012 (2050 m 3 •s −1 ) with release points and sampling stations.The drift net setup (inshore: IN, midshore: MID, offshore: OFF), likewise applied at each sampling station, is outlined in the upper left corner.Current direction is indicated by arrows.

Fig. 3 .
Fig. 3. (Left) Discharge regimes during sampling periods in 2011 (a) and 2012 (b).Arrows and bold font indicate sampling dates.Long-dashed line is regulated low flow, short-dashed line is mean flow.(Right) Box blots display current speeds at different distances from the shoreline.The dashed line corresponds to the maximum sustainable water velocity (MSWV) of nase larvae (Flore et al. 2001) calculated by MSWV = 4.39 + 0.456•TL, where TL is the mean total length of all recaptured individuals.

Fig. 4 .
Fig. 4. Comparison of recapture rates (RR) of passive drifting particles and fish larvae different sampling stations at the same location in the free-flowing main channel of the River Danube in two consecutive years.The combinations show (a and b) the result of the total catch, (c and d) inshore (IN) released larvae of the second developmental stage (L2), (e and f) offshore (OFF) released larvae of the L2 stage, (g and h) IN larvae of the fourth developmental stage (L4), and (i and j) OFF larvae of the L4 stage.The line in each graph indicates direct proportionality in the case when the passive particles and the larvae disperse in the same manner.

Fig. 5 .
Fig. 5. Percentage of samples with a disproportional high, low, or similar recapture rate (RR) of fish larvae (L) and passive particles (PP) at different sampling stations at two different years with different discharge conditions.

Fig. 7 .
Fig. 7. Box plots (median; 25th and 75th percentile; 5% and 95% outlier) of DIFF (difference between the recapture rate of larvae and particles) for the four tested factors.

Table 1 .
Measured values (mean) of water depth and current velocity at both release points (IR = inshore release, OR = offshore release) as well as for the different drift net stations (1-4) and positions (IN = inshore, OFF = offshore, and MID = midshore) for the separate study years.

Table 3 .
Results of the generalized linear model (GLM) analysing the impact of four factors (year, station, release, stage) and one covariable (filtered volume) on DIFF (difference between the recapture rate of larvae and particles).