Variation in hydropeaking‐induced stranding of Barbus barbus L. and Chondrostoma nasus L. larvae: Assessing the impact of daytime and down‐ramping rates

Unnatural changes in river flow patterns resulting from peak‐operating hydropower plants adversely impact freshwater ecosystems. In particular, the rapid dewatering of shoreline habitats during artificial flow down‐ramping puts early fish life stages at a high risk of becoming stranded if they fail to follow receding water levels in time. While extensive research has been conducted on the effects of hydropeaking on salmonid species, there is limited knowledge on the diverse cyprinid family, particularly on vulnerable early life stages. Hence, this study aims to compare the larval stranding of two cyprinid species, the common barbel (Barbus barbus L.) and common nase (Chondrostoma nasus L.), in response to bank dewatering. We conducted larvae experiments in near‐natural mesocosms, simulating single flow down‐ramping events with varying down‐ramping rates (0.3–1.8 cm·min−1) during the day and at night to quantify stranding rates, also including water temperature and fish development. Our results reveal distinct diurnal patterns for both species, with higher stranding rates during the night than during the day in all experimental scenarios. The data also show higher stranding rates at faster down‐ramping, with interaction effects between down‐ramping rates and time of day. The stranding rates between the two species are similar across most of the scenarios. Scenarios with colder water temperatures show that nase larvae tend to strand more frequently than with warmer temperatures. In conclusion, the study results contribute to the ongoing discourse on hydropeaking mitigation by providing new perspectives on flow‐reduction effects on early cyprinid life stages. Mitigation measures should prioritize the periods during early larval development and factor in prevailing water temperatures. Lowering down‐ramping rates, especially during nighttime, will help minimizing negative impacts on aquatic ecosystems, particularly when combining flow rules and habitat restoration measures.

Although stranding risk linked to these factors is highly species specific (Moreira et al., 2019;Nagrodski et al., 2012;Young et al., 2011), hydropeaking research has so far targeted mainly salmonids.Comparatively little information is available on cyprinid species (Moreira et al., 2019), particularly larval stages (but see Führer et al., 2022;Hayes, Auer, et al., 2023).Hydropeaking, however, not only occurs in salmonid streams but also in river reaches inhabited by cyprinids (De Vocht & Baras, 2003;Schmutz et al., 2015).Subsequently, more process-based studies on the effects of hydropeaking on cyprinids are expedient and valuable for advancing environmental objectives, such as those of the European Water Framework Directive (2000/60/EC), and to define effective mitigation measures.
Recent studies underline that cyprinids like barbel or nase exhibit species-specific responses to artificial flow down-ramping (e.g., Hayes, Schülting, et al., 2022;Pander et al., 2022).Moreover, initial stranding experiments conducted with nase larvae indicate a significant difference in stranding between earlier (III-IV) and later (V) larval stages (sensu Peňáz, 1974), with the highest effects at faster down-ramping during the night (Führer et al., 2022).However, complementary studies are necessary to increase the understanding of even earlier postemergence larvae, for example, larval stage II-III (Führer et al., 2022;Hayes et al., 2019).
To fill remaining knowledge gaps on species-and lifestage-specific effects of hydropeaking on cyprinid fish, we quantified stranding of barbel and nase larvae at different developmental stages in response to hydropeaking, considering time of day.We tested the single and combined effects of varying down-ramping rates and day versus night conditions, using near-natural mesocosms with suitable larval habitats.For barbel, we tested specimens of larval stage III (sensu Nowosad et al., 2021).For nase, we extended the stranding experiments conducted with larval stage III-V (Führer et al., 2022) by testing stage II-III.Total length and water temperature were additionally considered as influencing variables in the experiments (Korman & Campana, 2009;Pepin, 1991;Robinson & Childs, 2001;Seikai et al., 1986).We hypothesized that (i) larval stranding of both species increases with higher down-ramping rates, particularly during the night.For nase, (ii) we expected higher stranding rates at the earlier larval stage II-III compared to stage III-IV.In addition, (iii) we expected barbel to be less susceptible to stranding than nase due to their later emergence and associated warmer water temperatures.

| MATERIAL AND METHODS
We conducted fish stranding experiments in 2021 and 2022 using early life stages of nase and barbel at the 'Hydromorphological and Temperature Experimental Channels' (HyTEC) in Lunz am See, Lower Austria (47 51 0 22.5 00 N, 15 02 0 12.0 00 E; https://hydropeaking.boku.ac.at).The HyTEC facility consists of two parallel gravel bed channels (length: 40 m; width: 6 m) with water supply extracted at two different water depths from a nearby lake, a control unit to program, monitor, and record individual discharge rates and water temperatures, and fish rearing tanks (Figure 1a) (Führer et al., 2022).
The experiments comprised of three experimental sets: (1) nase at larval stage III-IV, (2) nase at larval stage II-III, and (3) barbel at larval stage III (Table 1).The first set consists of data published by Führer et al. (2022) and includes experiments with nase larvae stage III-IV conducted in 2021.For set 1 and 2, nase larval stages were grouped, that is, III-IV in 2021 and II-III in 2022, due to similar morphological characteristics between stages (e.g., fin-fold differentiation and fin development) and challenges associated with classifying and timing experiments (for further details on study organisms, see Section 2.1.).

| Study organisms
Nase and barbel larvae were obtained from wild fish and reared in the facility's circular tanks (Führer et al., 2022).The fish tanks were partially covered for shelter, and small stones were added to increase flow heterogeneity.Larvae were fed with live zooplankton obtained from the lake water and with Artemia sp.brine shrimp to meet their natural feeding requirements (Piria et al., 2005;Reckendorfer et al., 2001;Schiemer et al., 2002).Feeding was done several times a day at random intervals to minimize habituation effects.
To assess the larval length throughout the experimental period, we used ImageJ image-analysis software (Schneider et al., 2012) to measure the total length (TL ± 0.1 mm; Table 1; Table 2) of randomly selected larvae (30 ind.per measurement).In addition, we classified the larval developmental stages of barbel and nase according to Nowosad et al. (2021) and Peňáz (1974), respectively (Table 1; Table 2).

| Experimental setup
The stranding experiments were conducted in customized 2.25 Â 2 m mesocosms (Figure 1b).They were embedded in the experimental T A B L E 1 Overview of the three experimental sets conducted with early life stages of nase and barbel, indicating the year and period of the experiments, larval stage, fish length, and water temperature, with the number of independent replicates (n); sorted by experimental period and year.channels with a lateral slope of 2% (i.e., the bank slope) and a longitudinal slope of 0.5% (Führer et al., 2022).The mesocosms consisted of a flat river bar area (i.e., the ramping zone), primarily composed of sand and fine gravel (d max < 10 mm; d 50 = 2.2 mm).A small channel (width: 25 cm, depth: 7.5 cm) served as the mesocosm's deepest section and allowed for efficient clearing of larvae at low flow conditions (hereafter 'low flow channel').Fine-mesh net frames with a 0.75-mm mesh size ensured that fish remained in the mesocosm during the experiments.
The sediments remained the same for all mesocosms and experi- 2) and a micropropeller (Höntzsch FT25GFE-MN20/100/P6).During down-ramping, the riverbank was dewatered over a width of 1.75 m, corresponding to the potential stranding zone (Figure 1c).
Within each experimental trial, the water temperature remained constant (i.e., no thermopeaking) but varied among experiments due to natural lake water temperature variations (Figure 2; Table 2).Water temperatures were recorded for fish rearing tanks and each experimental channel by the facilities' control unit.In addition, we used HOBO Pendant Temperature/Light 64K data loggers (UA-002-64).
The water temperatures at the facility were consistent with seasonal temperatures in rivers inhabited by cyprinids (Baras, 1995;Butz, 1985;Führer et al., 2017), ranging from 8.4 C to 19.4 C across all trials (Figure 2; Table 2).

| Experimental design
We performed 20 experimental scenarios, based on the combination of (i) species, (ii) set, (iii) daytime, and (iv) vertical downramping rate (Table 1; Table 2).Scenarios with nase comprised two experimental years, which differ by larval stage, fish length, and water temperature (see Table 2; Figure 2).Each larva was tested only once (i.e., no reuse of fish).After finishing the experiment, the larvae were kept in individual tanks and stocked into rearing ponds at the end of each experimental scenario prior to release.Channel location was randomly assigned before each experiment to avoid channel effects.
F I G U R E 2 Overview of single down-ramping trials (n = 163) on 2% sloped banks with early life stages of nase and barbel, grouped by experimental set (Table 1), separated by day (triangle) and night (circle).Panel (a) shows the water temperatures of the single experiments, while panel (b) summarizes the water temperature distribution per set (the boxplot's lines and whiskers represent the median values and interquartile ranges, respectively).
The experimental design consisted of the following steps.For each mesocosm, (i) 100 larvae were taken from the rearing tanks and (ii) gently stocked at high flows (80 LÁs À1 ) in the shoreline zone of the mesocosm (Figure 1b,c).Fish were stocked by tilting the small buckets towards the flow and waiting until all individuals had independently left the bucket, avoiding a flight response of larvae.The water temperature between the rearing tanks and the mesocosms was identical.
Stocking followed (iii) an acclimation period of 15 min at unchanged flows to prevent stocking-related responses during the following experimental phases.After acclimation, (iv) flow down-ramping ensued by automatically reducing the discharge to 10 LÁs À1 .After down-ramping, (v) larvae were quantified according to their location: stranded fish on the substrate, specimens swimming in the low flow channel, and those displaced into the nets (Führer et al., 2022).Larvae found lying on the substrate close to the downstream net (i.e., the 'near-net zone'; see Figure 1b) were not counted as stranded but were summed up with displaced larvae, as we observed that larvae previously displaced into the net dropped onto the dewatered substrate later.In this regard, preliminary video-based analysis revealed behavioural differences between our target species, leading us to define the longitudinal extent of the near-net zone at ≤1 cm for nase and ≤2 cm for barbel.The larger barbel near-net-zone was necessary to ensure comparability of stranding rates between the two species, as barbel larvae tended to actively escape from the net by vigorously flapping their caudal fin, which caused them to fall further away from the net onto the substrate, or they attempted to move on the dewatered substrate away from the net.Observations also indicated that larval displacement into the net occurred mostly before flow downramping.Respectively, those individuals were not available for possible stranding (Führer et al., 2022;Heggenes & Traaen, 1988) and were excluded from stranding calculations in a later step (see Equation 2).
After the first clearing, (vi) the mesocosms were reflushed, and the sediments and nets were re-examined for any remaining larvae.All larvae found during this second clearing run were counted as nonstranded individuals, except those observed flushing out from the substrate, which were counted stranded.This comprehensive assessment, including a third clearing run before each trial, yielded a mean recapture rate of 98.9% ± 2 SD.
The number of replicates varied by scenario (Table 2), for example, due to restricted time at night due to longer daylight phase and limited experimental period due to larval growth.In addition, we had to exclude single replicates from the dataset due to external factors, including heavy rain showers or unintended interferences during trials.
The frequency of stranded larvae (Str calc ) was determined for each trial as follows: Str calc ind: whereby the 100 is the count of stocked larvae, N the larvae retrieved from the nets, and N 0 representing those found on the substrate in the near-net zone.C denotes the number of individuals cleared from the low flow channel.According to Equation ( 1), missing individuals were assumed to be stranded but hidden in the substrate and were, therefore, included for stranding quantifications (Auer et al., 2017;Führer et al., 2022).
The stranding rate (Str rate ) was calculated as follows:

| Statistical analyses
Our study aims to assess differences in stranding of nase and barbel larvae (i.e., the response variables) at varying down-ramping rates (0.3-1.8 cmÁmin À1 ) and times of day (day and night), which are considered as explanatory variables.
We used odds and odds ratios (ORs) to determine variation in stranding between pairs of experimental scenarios.The OR is an appropriate statistical method for the present data; for example, it is insensitive to the marginal distribution of the response variable (Arminger et al., 1995;Hedderich & Sachs, 2016;Karlson & Jann, 2023;Morris & Gardner, 1988).We calculated the odd of stranded (Str calc ) and non-stranded (C) larvae for each experimental scenario, composed of the variables species Â set Â daytime Â downramping rate (see Table 2).ORs were then calculated for pairs of experimental scenarios (i.e., the odd of scenario A vs. the odd of scenario B).ORs can indicate the direction of the association between the scenarios as well as the strength of the association (Sandercock, 1989).Following, we standardized the ORs from À1 to +1 by calculating Yule's Q (Yule, 1912).With Yule's Q, negative values result from a lower ratio of stranded to nonstranded larvae in Scenario A, while positive values result from a lower ratio of stranded to nonstranded larvae in Scenario B. We assumed the equivalence of scenario pairs for Yule's Q between À0.08 and +0.07, a common assumption in medical sciences, though it may differ across various disciplines.To indicate statistical significance, we calculated the 95% confidence interval (CI) of Yule's Q by using the logit method (Woolf, 1955).If the CI of a scenario pair deviates from the equivalence assumption, the pair of experimental scenarios can be regarded as different (Sandercock, 1989).To also express significance by the widely used p-values, we used the Cochran-Mantel-Haenszel test with a significance level (α) of 0.05 (CMH; Cochran, 1954;Mantel & Haenszel, 1959) and calculated the power coefficient Phi.
The analysis approach described above only assesses difference between pairs of experimental scenarios.In a follow-up step, we aim to identify the hierarchy of explanatory variables and which effect sizes (main and interaction effects) explain observed differences in stranding (irrespective of species).We used the RPART decision tree routine in R© (CART algorithm with Gini index; see Therneau & Atkinson, 2022), examining the dichotomized stranding rate (median as response variable due to the unbalanced data distribution).In addition to the variables tested above (species, down-ramping rate, and daytime), larval stage, total length, and water temperature were also considered as explanatory variables.The variables are categorial (species, daytime), ordinal (down-ramping rate, larval stage), and metric (total length, water temperature).
We used the R© package 'rstatix' (Kassambara, 2023) to test for differences in water temperature between the three experimental sets (naturally resulting due to different experimental years and times of larval development; see Figure 2), using the nonparametric Kruskal-Wallis test (accounting for the Dunn-Bonferroni correction for multiple pairwise testing), as data normality and homoscedasticity were not met.
When aggregating all trials (n = 163), the stranding rate ranged from 0.00 to 0.89 (Figure 3), with a median of 0.13.The stranding rates vary by species, set, daytime, and down-ramping rate (Figure 3).There was no evidence of a difference in stranding between the two experimental channels and mesocosms, respectively (Q: 0.03; CI: À0.01 to 0.08; p = 0.168).Figure 4 shows the test results between all pairwise comparisons of experimental scenarios (Table 2).These results are described in detail in the following sections.

| Species-specific effects, including larval stage
In the following sections, the results of nase (sets 1 and 2) and barbel (set 3) experiments are compared.Water temperature differed significantly between all three sets ( p < 0.001) (Table 1; Figure 2).
At the down-ramping rate of 0.7 cmÁmin À1 , daytime stranding was comparable for both species, with a median of 0.06 for nase (n = 8) and for barbel (n = 11) (Figure 3a).
We also found significant differences between nase in set 2 and babel at nighttime down-ramping rates of 0. (Figure 3a).

| Main and interaction effects
Following, the decision tree analysis examines which variables constitute main and interactive effects in explaining the variability of larvae stranding rate.At the first level of the decision tree model, daytime was identified as main effect (Figure 5).For night trials, the median (median = 0.07; n = 66).Nighttime trials are further split by downramping rate, with higher down-ramping rates (≥1.1 cmÁmin À1 ) showing a three times higher stranding rate (median = 0.39; n = 38) compared to slower down-ramping rates (≤0.7 cmÁmin À1 ) with a median of 0.13 (n = 32).For higher down-ramping rates, the decision tree splits by total length at the third stage, with smaller larvae (<13.9 mm) showing a median stranding of 0.49 (n = 16), which is the highest value in the model and almost double that of larvae with a length ≥13.9 mm (median = 0.25; n = 22).The decision tree model did not select species and larval stage.
Comparative works on hydropeaking effects on stranding of earliest life stages are expedient.To fill this knowledge gap, this study assessed the stranding rate of nase (sets 1 and 2) and barbel (set 3) larvae regarding the single and combined effects of varying flow down-ramping rates and times of day.
The experiments show that the stranding susceptibility of barbel and nase larvae is similar, which was contrary to our expectations.
Our results confirm the hypothesis that larval susceptibility to stranding is higher at night than during the day for both species and that the stranding risk is positively linked to down-ramping rates.However, day and nighttime experiments indicate that water temperature may also affect the stranding of cyprinid larvae at higher down-ramping rates, as stranding rates of nase were higher at colder (set 1) than under warmer conditions (set 2), although nase larvae were less developed in the experimental set with warmer temperatures.

| Time of day effects
Stranding was evident for both species and in all experimental scenarios.As expected, nocturnal stranding was significantly more pronounced than stranding during the daytime.This pattern was evident for all three experimental sets and nearly all tested down-ramping rates, except for barbel trials at the lowest down-ramping rate of 0.3 cmÁmin À1 , where no difference regarding the time of day could be detected.The highest stranding rate of about 0.9 in single trials was observed during night trials at the fastest ramping rate (1.8 cmÁmin À1 ).
Overall, nighttime stranding was about 2.5 times higher than during the day.The largest difference between median day and night stranding was detected for barbel at a down-ramping rate of 1.5 cmÁmin À1 , exhibiting nighttime stranding rates about 5.5 times higher than at day.
The decision tree model suggests that time of day effects are central in explaining stranding rate variability (Figure 5).Observing fish during the experiments revealed behavioural differences that can explain these findings.During nighttime trials, we noted that larvae of both species remained mostly stationary near the stocking point at first and dispersed later and to a lesser extent, tending to stay closer to the shoreline in shallow areas.We believe that this is likely due to the reduced visual cues caused by the lack of light at night.Contrastingly, during the day, some larvae remained near the stocking point, while others dispersed in small groups in the mesocosm, likely searching for habitats characterized by deeper areas with higher flow velocities (see also Führer et al., 2022).Hence, larvae dwelling at shallow shoreline habitats had to react more quickly to the receding water level during flow down-ramping and had to perform longer lateral shifts to avoid stranding than larvae staying in deeper areas.We believe this diurnal behaviour is likely leading to higher stranding at night, especially at high down-ramping rates, when larvae have less time to shift.Studies with nase larvae (Führer et al., 2022;Hayes, Auer, et al., 2023) and young-of-the-year brown trout (Salmo trutta L.) and grayling (Thymallus thymallus L.) also found higher stranding at night than during the day (Auer et al., 2014(Auer et al., , 2017)).However, some authors found no or only little diurnal differences for salmonids (Beck and Associates, 1989;Bradford, 1997), while others reported higher stranding rates during the day (Bradford, 1997;Bradford et al., 1995;Halleraker et al., 2003;Monk, 1989;Saltveit et al., 2001;Woodin, 1984).Therefore, stranding risk will likely be affected by diverse biotic and abiotic factors (Bradford, 1997;Larrieu et al., 2021;Moreira et al., 2019;Nagrodski et al., 2012;Young et al., 2011), namely fish species, developmental stage, and size, and/or river morphology, habitat diversity, hydropower operation, season, and water temperature, among others.Some of these aspects are discussed further below.

| Down-ramping rate effects
The decision tree model (Figure 5) indicated that, after the time of day, the down-ramping rate is key in determining cyprinid larval stranding during the night, showing that larval stranding of both species increases with higher down-ramping rates.During the day, however, only nase experiments of set 1 (larval stage III-IV, colder water temperature) followed this pattern (Figure 3).In contrast, daytime experiments with nase in set 2 (larval stage II-III, warmer water temperature) and with barbel (set 3 with larval stage III) exhibited no relationship between stranding and down-ramping rate.Compared to stranding of barbel and nase in set 2 at a down-ramping rate of 1.5 cmÁmin À1 , nase trials in set 1 showed even higher stranding at a lower down-ramping rate (1.1 cmÁmin À1 ).
The effects of down-ramping rate on stranding have also been reported differently in the literature.Multiple studies show that stranding lowers when ramping rates are reduced (Auer et al., 2014;Bauersfeld, 1978;Bradford et al., 1995;Halleraker et al., 2003).This pattern was likewise shown for nase at larval stage V (Führer et al., 2022).Also, Austrian-wide field data studies by Schmutz et al. (2015) and Hayes et al. (2021) link higher down-ramping rates (>0.5 cmÁmin À1 and >0.2-0.4 cmÁmin À1 ) to a poor or bad fish ecological status and reduced grayling biomass, respectively.A case study of the hydropeaked Bregenzerach River (Austria), inhabited by cyprinid species, showed that fish biomass did not recover after implementing mitigation measures (increased base flow and decreased peak flow), which can be explained by the unaltered ramping rate (Parasiewicz et al., 1998), possibly still exceeding ecological thresholds.In contrast, Woodin (1984), Beck and Associates (1989), and Monk (1989) found no or only a weak relationship between fish stranding and down-ramping rate for salmon species in the Skagit River (USA) and in experimental studies.
In general, fish become less susceptible to stranding as they increase in size (Moreira et al., 2019;Young et al., 2011).The downramping rates' effects can also be related to species and water temperature, especially during early larval development.

| Species-specific effects, including larval stage
We observed different behavioural patterns between species.During barbel daytime trials, some larvae remained in place despite the receding water levels and shifted just before the area was completely dewatered.Other barbel shifted laterally between shallow and deep areas during down-ramping.We assume that they could sense the water depths quite accurately, as these individuals usually did not become stranded.We observed this specific behaviour also during nase daytime trials.In comparison, nase larvae tended to shift earlier to deeper areas during the receding flow.
A comparison of the two experimental sets with nase at a downramping rate of 1.5 cmÁmin À1 revealed that larval stranding in set 1 (larval stage III-IV) was significantly higher than in set 2 (larval stage II-III), and this pattern was irrespective of the time of day (Figure 3).
This difference in stranding rates was unexpected, as nase larvae had similar lengths at the start of both sets (Table 1; Table 2).In addition, the larval developmental stages only slightly differed between both sets, with even earlier stages tested in set 2 (larval stage II-III) than in set 1 (larval stage III-IV).
A reasonable explanation for the divergent results of the two sets with nase is the water temperature, which differed between the sets due to natural variations in lake water temperature (Figure 2).Like Scruton et al. (2005), we observed that fish tend to be less active, exhibit lower mobility, and are more sedentary at colder water temperatures.Although our study did not focus on water temperaturerelated changes in stranding, we assume that this is a key factor of the observed differences in stranding for two reasons.First, the lower water temperatures in set 1 (mean = 9.4 C ± 0.7 SD; Table 1) may have lowered fish activity (Volkoff & Rønnestad, 2020), reduced swimming performance (Heggenes & Traaen, 1988;Kaufmann & Wieser, 1992;von Herbing, 2002;Wardle, 1980;Wieser & Kaufmann, 1998), potentially resulting in a more sedentary behaviour (Scruton et al., 2005), and increasing the stranding susceptibility (Halleraker et al., 2003;Saltveit et al., 2001).This temperature-related effect was particularly evident during the day (Figure 5).Second, the higher water temperatures in set 2 (mean = 11.1 C ± 0.6 SD; Table 1) induced faster growth during the experimental period (without changing the feeding routine between both sets), with larvae ending up, on average, 4.5 mm longer than in set 1 (Table 1; Table 2), even though fish of set 2 remained in an earlier larval development stage than set 1. We hypothesize that the fish swimming performance improved due to the increase in total length (Flore et al., 2001), which may also entail a preference for deeper habitats, resulting in reduced susceptibility to stranding.
For barbel, we assume that colder water temperatures may also increase stranding risk, particularly concerning critical temperatures or thermopeaking (Auer et al., 2023;Olson & Metzgar, 1988).However, we only studied barbel stranding in a limited temperature range (Table 1; Figure 2).Therefore, further research is needed to verify this hypothesis.
The frequency of stranded larvae was calculated according to the location where individuals were found after down-ramping, and missing larvae (on average approx.1% per trial), likely concealed in the substrate, were regarded as stranded (see Section 2.3).Fish found in the near-net-zone (Figure 1b) were excluded from stranding calculations (see Equation 1), although we also observed real stranding within this zone (Führer et al., 2022).In this regard, we pursued a conservative approach.Generally, we achieved very high recapture rates and assume our results underly only marginal uncertainties.
In our experiments, we simulated single peak events to quantify larvae stranding.In hydropeaked rivers, however, artificial flow fluctuations can occur several times a day (Greimel et al., 2016), potentially causing long-term effects by repeated bank dewatering (Auer et al., in preparation;Halleraker et al., 2003;Hayes, Bruno, et al., 2023).
In consequence, the cumulative effects of repeated flow reductions (Bauersfeld, 1978;Young et al., 2011) that are likely affecting fish populations (Hayes et al., 2021) have to be taken into account when considering management options.

| Management implications
Our study underlines that time of day and down-ramping rate are major factors impacting cyprinid larval stranding.Reduced downramping rates may give larvae with lower swimming capacities more time to shift to deeper areas, leading to a decreased stranding risk.This is especially relevant for nighttime hydropower operations, as larval barbel and nase are particularly susceptible to nocturnal stranding, as also indicated by previous studies of Führer et al.
(2022) and Hayes, Auer, et al. (2023).Both studies found similar results, even for more developed and larger-sized nase larvae, respectively.Furthermore, water temperature is an additional factor potentially influencing stranding, as nase experiments revealed lower stranding at warmer conditions than colder ones, which subsequently could influence recruitment success (Godinho et al., 2022).
This would be relevant concerning natural year-to-year variations in river water temperature, with colder conditions potentially reducing the abundance of early life stages.However, more studies targeting temperature-related changes in cyprinid stranding, including tests on critical water temperatures and thermopeaking (e.g., Mameri et al., 2023), are required to define effective mitigation measures, such as introducing temperature-dependent restrictions on flow down-ramping.This may be an appropriate approach to reduce larval stranding, particularly for nase, which usually emerge from the riverbed substrate in the spring when water temperatures can still be cooler, for example, due to snowmelt.
Regarding hydropeaking mitigation frameworks, our results support limiting down-ramping rates at certain times of day or year, such as during and after larval emergence, especially at night (Hayes et al., 2019).The earliest fish developmental stages are particularly sensitive to flow fluctuations (Harby & Noack, 2013;Hayes et al., 2019;Hunter, 1992) and therefore critical in determining population vitality (Elliott, 1989).As water temperatures affect growth, development, and larval behaviour (Pepin, 1991;Robinson & Childs, 2001;Seikai et al., 1986), and thus the stranding risk, we call for consideration of water temperature and fish length when determining and implementing flow rules.Based on our results, we recommend referring to the total length of fish when defining periods of restricted hydropower operations (such as the "emergence window" concept; Hayes et al., 2019) instead of the larval stage.Also, fish length can be more easily determined in the field compared to the developmental stage.

| Conclusions
This study aimed to quantify barbel and nase larvae stranding in response to different down-ramping rates and times of day, also considering larval stage and water temperature.The most critical factor for both species was time of day, with nighttime stranding rates two to three times higher than during the day.The effect of the downramping rates was most pronounced at night, with higher ramping rates yielding increased fish stranding.Even though both species showed a similar stranding susceptibility for most comparisons, some scenarios indicate that water temperature-related effects potentially influence stranding, as nase stranded more frequently at colder than under warmer conditions.We recommend that mitigation frameworks limit down-ramping rates during the earliest life stages of barbel and nase, particularly at night and at low water temperatures.
Overview of the HyTEC-facility with the two experimental channels with detailed information about the experimental mesocosm setup, including the water level at low and high flow, with the blue shading indicating the wetted area during the experiments; adapted from Auer et al. (2023).(b) Oblique view of a mesocosm, with additional details on features and dimensions.The red line indicates the intersection axis for the section A-A.(c) Cross-section A-A with the flow velocity distribution in the mesocosms at high flow rate of 80 LÁs À1 (top) and low flow rate of 10 LÁs À1 (bottom); adapted from Führer et al. (2022).
is composed of the four variables, coded as: study species Â experimental setjdaytimejdown-ramping rate.Na = nase, Ba = barbel, D = day, N = night.
ments and were smoothed before each trial to ensure constant conditions.Accordingly, the nets were cleaned to avoid clogging during the experiment.The experiments consisted of a high flow of 80 LÁs À1 , followed by down-ramping until a low flow of 10 LÁs À1 .At high flows, the water depth at the shoreline was 4 cm, and 17 cm in the low flow channel.The flow velocity was <0.075 mÁs À1 close to the shoreline and <0.25 mÁs À1 in the low flow channel (Figure 1c).Only the low flow channel remained wetted at low flow, with a water depth of about 10 cm and flow velocities largely <0.1 mÁs À1 .Flow velocities were determined using a 2-D ADV flow meter (SonTek FlowTracker

F
I G U R E 3 Stranding rates of the 20 experimental scenarios depending on the down-ramping rate (0.3-1.8 cmÁmin À1 ) with the number of replicates (n) for each scenario.The experiments were performed with nase in 2021 (yellow), nase in 2022 (dark grey), and barbel in 2022 (blue), during day (a) and night (b).The boxplots with the associated bold lines and whiskers represent median values and interquartile ranges.White squares indicate mean values.Dots represent individual trials, with colours corresponding to those of the boxplots.combining all three sets at day, the median stranding rate was 0.08 at a down-ramping rate of 0.3 cmÁmin À1 (n = 28), 0.06 at 0.7 cmÁmin À1 (n = 19), 0.16 at 1.1 cmÁmin À1 (n = 6), and 0.09 at 1.5 cmÁmin À1

For
the same down-ramping rates tested with nase in set 2 (larval stage II-III), the difference in nocturnal stranding between the F I G U R E 4 Pairwise comparisons of the 20 experimental scenarios (for coding explanation see Table 2): Yule's Q and confidence interval (CI), complemented with Cochran-Mantel-Haenszel (CMH) test results and the power coefficient (Phi), sorted by Yule's Q.For the CMH test: ***p < 0.001, **p < 0.01, *p < 0.05; n.s.indicates no statistical significance.

stranding was 0 .
21 (n = 70), which is about 2.5 times higher than for day experiments, with a median of 0.08 (n = 93).The second level decision rule further separates the daytime trials by water temperature, with lower temperatures indicating a stranding rate more than double (median = 0.16; n = 27) than higher temperatures F I G U R E 5 Decision tree model for larvae stranding rate, including species, daytime (D = day; N = night), downramping rate (DR rate), larval stage, total length (TL), and water temperature (WT) as explanatory variables.Values in the boxes represent the median of the stranding rate, with stronger fill colours indicating increasing rates.
Overview of the 20 experimental scenarios conducted with larval nase and barbel.