Turbidity effects on prey consumption and survival of larval European smelt ( Osmerus eperlanus )

The anadromous European smelt ( Osmerus eperlanus ) plays a key role in food webs of many riverine ecosystems in Europe. However, population sizes in several German rivers (e.g. Elbe or Weser rivers) have diminished drastically over the past decade. Turbidity has been considered one of the stressors affecting the successful recruitment of European smelt, as their early life stages may be particularly sensitive to changes in the abiotic environment. In this study, we investigated whether prey consumption and survival of European smelt larvae would be negatively affected by an acute exposure to elevated turbidity. We reared the larvae in the laboratory and exposed them in four separate trials (18 to 26 days post hatch, 9.5 ± 0.8 mm standard length, mean ± SD) to six turbidity levels (0–500 NTU, nephelometric turbidity units). We found that prey uptake increased at low turbidity levels and decreased at high turbidity levels, with an optimum between 100 and 200 NTU. Survival started to decrease at turbidity levels above 300 NTU. In addition, we conducted a systematic literature analysis in which we found that prey consumption of larval and juvenile fishes had been tested across a wide range of turbidity levels, mostly using pelagic (e.g. planktonic) prey items, with more studies focusing on perciform fishes and juvenile rather than larval life stages. Our empirical findings contribute to establishing thresholds for optimal larval European smelt performance under increased turbidity and provide valuable information for developing mechanistic models that assess potential consequences for European smelt recruitment dynamics.


Introduction
Habitats of estuarine and coastal fishes are characterized by tidal, diel and seasonal fluctuations in environmental parameters, e.g.salinity, pH, temperature and turbidity (Cochran 2014;Whitfield 2021).Turbidity, i.e. the scattering and absorption of light in water, is caused by organic matter, plankton or mineral compounds (Wood and Armitage 1997).
In the lower reaches of major rivers, turbulence, driven by tides, density effects and waves, can resuspend sediments and other matter, creating a zone called the estuarine turbidity maximum (Burchard and Baumert 1998;Dyer et al. 2004).The position of this zone can be modulated by tidal variation and vary depending on river discharge, resulting in seasonal differences (Hesse et al. 2019).Turbidity can be affected by land through sediment input via precipitation and run-off from agricultural and residential areas as well as human activities, such as logging or mining (Chow-Fraser 1999;Lemly 1982;Mallin et al. 2009).Shipping traffic and maintenance of waterways add to the anthropogenic sources of turbidity, as heavy and regular intervention in the riverbed occurs (e.g.instream and suction dredging) (Smith and Friedrichs 2011;Wilber and Clarke 2001;Winterwerp et al. 2002).All these actions were reported to result in increases in turbidity (Miró et al. 2022;Pledger et al. 2021).
Anadromous fishes, migrating into rivers for spawning, often have to pass highly turbid estuaries.The anadromous European smelt [Osmerus eperlanus (Linnaeus 1758)] migrates in spring from its coastal habitats, where it reaches sexual maturity, into estuaries and rivers for upstream spawning.While landlocked European smelt populations exist in lakes of coastal areas around the North, Baltic, White and Barents Seas (Kottelat and Freyhof 2007), the anadromous form inhabits coastal and estuarine waters in western Europe and is distributed from the Garonne estuary in the south to the Baltic and White Sea (McAllister 1984).European smelt can dominate the fish fauna in European rivers, such as the Elbe River in Germany (Eick and Thiel 2014), and play a key role in riverine ecosystems, providing food resources for other fishes and birds (Dänhardt et al. 2011;Hennig et al. 2016).Yet, biomass estimates for many European smelt populations are lacking because limited monitoring data are available for the species (Wilson and Veneranta, 2019).European smelt populations are sensitive to environmental disturbances, and various stocks have declined in biomass in recent years (Arula et al. 2017;Keller et al. 2020;Keskinen et al. 2012;Sendek and Bogdanov 2019).An amalgam of drivers is suspected behind these declines: an unregulated, data-deficient fishery, obstacles to spawning migration (e.g.dams and weirs) and adverse changes in environmental factors on the spawning grounds reducing offspring survival (e.g.warmer, more oxygen-deficient and more turbid waters).
Fishes can be affected by elevated turbidity in various ways, and, depending on the dose and exposure duration, show modifications at cellular, tissue and organism levels.At the cellular level, early life stages of delta smelt (Hypomesus transpacificus, McAllister 1963), acutely exposed to elevated turbidity, showed altered transcription patterns of stress-related genes that were required for protecting cells against reactive oxygen species (glutathione-S-transferase, gst) and maintenance of energy metabolism (glucose transporter 2, glut2) and regulating general stress (heat shock protein 70 kD, hsp70) (Hasenbein et al. 2016(Hasenbein et al. , 2013)).Regarding tissues, gills are the most exposed structures, and elevated turbidity and suspended sediments have been found to deform gills (i.e.shorten and fuse gill lamellae or decrease their total number) and reduce oxygen diffusion distances through epithelial hyperplasia (Cumming and Herbert 2016;Hess et al. 2017;Lowe et al. 2015).At the organism level, exposure to suspended sediments seems to result in species-specific responses, as some fishes were found to be negatively affected in their oxygen uptake rates (e.g.juvenile cinnamon anemonefish Amphiprion melanopus, Bleeker 1852) (Hess et al. 2017), whereas these effects were not observed in other species (juvenile silver seabream Pagrus auratus, Forster 1801, orange clownfish Amphiprion percula, Lacepède 1802, and spiny chromis Acanthochromis polyacanthus, Bleeker 1855) (Cumming and Herbert 2016;Hess et al. 2017).However, overall performance factors such as growth in juvenile silver seabream and swimming capability in Chinook salmon (Oncorhynchus tshawytscha, Walbaum 1792) have been found to decrease under elevated and prolonged turbidity levels (Cumming and Herbert 2016;Lehman et al. 2017).Although a number of narrative reviews on the effects of turbidity on fishes exist (e.g.Kjelland et al. 2015), few systematic reviews and metaanalyses have investigated organism-level effects of turbidity on fishes (Chapman et al. 2014;Rodrigues et al. 2023).To date, for example, there is no overview of the extent and methods used in studies investigating the effects of turbidity on prey consumption in larval and juvenile fishes.
Turbidity can negatively affect visually oriented aquatic predators by reducing prey capture rates and thus hamper successful foraging (Chapman et al. 2014;Ortega et al. 2020).In fishes, predator-prey relationships have been shown to be affected by turbidity in parameters, such as reactive distances (Hansen et al. 2013;Mazur and Beauchamp 2003;Meager et al. 2005;Minello and Benfield 2018), encounter and attack rates (Grecay and Targett 1996;Pekcan-Hekim et al. 2013;Utne-Palm 2004), search time/ activity (Meager et al. 2005;Meager and Batty 2007;Vollset and Bailey 2011) and prey selection (type, size and contrast) (Jönsson et al. 2011;Minello and Benfield 2018;Reid et al. 1999;Snow et al. 2018).Yet, these effects can be speciesspecific, and in a meta-analysis investigating the effects of turbidity on mobility in fishes, some species reduced their activity in turbid waters while others did not change their activity or increased it (Rodrigues et al. 2023).In addition, piscivore fishes seem to be more strongly affected by elevated turbidity than planktivore fishes (De Robertis et al. 2003).Other behavioural changes, such as reduced risktaking and increased antipredator responses, were observed across different species and life stages when fishes were exposed to elevated turbidity (Leahy et al. 2011;Lehtiniemi et al. 2005;Wing et al. 2021).Early life stages of fishes might be particularly sensitive to turbidity (Partridge and Michael 2010;Phan et al. 2020;Wyatt et al. 2010).Given that fish larvae need to grow fast, have few energy reserves and switch from endogenous (i.e.yolk) to exogenous feeding (i.e.consumption of planktonic prey), any change in environmental conditions affecting their feeding success could have major consequences for their growth performance and survival.However, some research suggests that early life stages of fish can also benefit from nutrient-rich, turbid river flows, leading to better recruitment (Carreon-Martinez et al. 2015).Anadromous European smelt experience high turbidity levels on their nursing grounds in the lower reaches of large rivers.With the recent decline in their population sizes, it is important to gain a mechanistic understanding of how larval European smelt survival and prey consumption are potentially affected by turbidity.
In this study, we experimentally investigated the effects of an acute exposure to turbidity on prey consumption and survival of European smelt larvae using turbidity levels observed at monitoring stations within the tidal Elbe and Weser rivers (Germany).We hypothesized that turbidity would affect fitness of European smelt larvae by reducing prey consumption and survival.To better contextualize our experimental findings, we provide additional information from a systematic literature search on how effects of turbidity on prey consumption of fish larvae and juveniles have been previously investigated.

Artificial spawning, egg incubation and animal husbandry
Adult European smelt were caught during their spawning migration in February 2022 by a commercial fisherman using eel pots in the River Weser at the Bremen Weser Weir (53.060°N, 8.864°E, Bremen, Germany).Fish were transported alive to the experimental facilities at the Thünen Institute of Fisheries Ecology (Bremerhaven, Germany) for artificial reproduction.Three females and three males were used with mean (± SD) standard lengths (SL) of 20.5 ± 0.4 cm and 18.8 ± 0.6 cm, respectively.Fish were killed by percussive stunning and strip-spawned afterwards.Eggs were fertilised using the dry fertilization technique (Walker et al. 2010).Then, egg adhesiveness was removed (1.2 g l −1 tannic acid solution for 10 min) and eggs were disinfected (2 ml l −1 hydrogen peroxide, 35%, for 15 min, both chemicals Sigma-Aldrich Chemie GmbH, Taufkirchen, Germany), as described in Reiser et al. (2023) following methods described in McCarthy et al. (2020).A total of 22.8 g eggs was incubated in a MacDonald egg incubator (8 l volume) (Reiser et al. 2023) with a constant flow of degassed well water at 1.7 l min −1 and mean (± SD) temperature of 12.1 (± 0.1) °C, dissolved oxygen (DO) 10.7 (± 0.2) mg ml −1 and pH 8.19 (± 0.02) (Multi 3510 DS, Xylem Analytics, Weilheim, Germany).Fertilisation success was determined after 24 h and reached 98.8%.Eggs were disinfected every 3 days to minimize fungal growth (Reiser et al. 2023), and embryos hatched after 12 days with a hatching success of 75.1% and a size at hatch of 5.7 ± 0.2 mm standard length (SL, from the tip of the mouth to the end of the notochord).Larvae hatched into a rectangular 57.5-l glass aquarium, shaded with black tarp on all sides with an opaque plastic board covering half of the surface area.Larvae were kept at mean (± SD) temperature of 12.1 (± 0.1) °C, 10.9 (± 0.1) mg ml −1 DO and pH 8.13 (± 0.05) at a density of about 525 larvae l −1 .Starting from 5 days post hatch (dph), fish were fed decapsulated artemia cysts and freshly hatched Artemia nauplii at 50 nauplii larva −1 day −1 .Prey was dosed into the aquarium between 8:00 and 19:00 at hourly intervals (Reef Doser EVO 4, AB Aqua Medic GmbH, Bissendorf, Germany) and a light regime of 12:12 L:D (8:00-20:00).For further details on European smelt rearing procedures, please see Reiser et al. (2023).Turbidity trials were carried out 2 and 3 weeks after hatch.Animal rearing and all experimental procedures were conducted in accordance with European directive 2010/63/EU on the protection of animals used for scientific purposes.

Experimental design and sampling
European smelt larvae were acutely exposed to turbidity in four separate trials at 18, 19, 25 and 26 dph at mean (± SD, n = 136, 95, 96, and 96) SL of 9.31 (± 0.65) mm, 9.43 (± 0.62) mm, 9.62 (± 0.86) mm and 9.63 (± 0.98) mm, respectively.Turbidity levels ranged from 0 to 500 NTU (nephelometric turbidity unit), with 0 NTU serving as a control and five turbidity levels (100 to 500 NTU).This range covers turbidity levels observed in the lower reaches of the Elbe and Weser estuaries (Germany).In fact, turbidity in the Elbe River (Fig. 1) has been observed since 2010 to regularly peak above 400 FNU (formazin nephelometric unit; 1:1 conversion rate with NTU, as both FNU and NTU measure scattered light at 90° from an incident infrared and white light beam, respectively) (see Fig. S1 for the turbidity trends of the Weser River, Germany).In our experiments, turbidity was created using bentonite (Diaclean GmbH, Dortmund, Germany), a mineral clay that has been regularly used for experimentally simulating turbidity levels (Gardner 1981;Minello and Benfield 2018).For obtaining stable bentonite concentrations, pre-weighted amounts of bentonite were vigorously mixed with 10 l of degassed well water.The suspension was kept for 21 h before the supernatant was used for the acute exposure trials (see supplementary information, Fig. S2).In pre-trials, the suspension provided stable concentrations for several hours with a mean (± SD) decrease of −2.7 (± 1.5) % after 4 h.
In each of the four trials, the experimental design consisted of 4 replicate beakers per turbidity level and the control, resulting in 24 individual beakers.Beakers were randomly allocated to a treatment using a raffle, which was repeated for every trial.Beakers were loaded with 2 l of water from the respective turbidity level and the control, and water parameters were measured, including temperature (°C), pH, DO and turbidity (NTU).Each beaker was gently aerated.Beakers were placed in dark grey polypropylene pipe sections for cover, and half of its surface was shaded with semi-transparent black plastic foil to mimic conditions in the rearing aquarium and to ensure incident light came from the surface.The amount of photosynthetically active radiation (PAR) in the individual beakers ranged between 4 and 14 µmol m −2 s −1 and was negatively correlated with turbidity [PAR = −0.01(± 0.0) Turbidity + 13.42 (± 0.87), see Fig. S3] (LI-1500 Light Sensor Logger, Li-cor Biosciences GmbH, Bad Homburg, Germany).Prey consisted of decapsulated artemia cysts and freshly hatched nauplii and was added at a concentration of 5 individuals ml −1 .Previous studies successfully reared European smelt larvae on this type and concentration of prey (McCarthy et al. 2020), and the larvae in this experiment experienced the same food for 12 to 19 days before experimentation commenced.Prey was added to the experimental beakers and thoroughly stirred with a paddle to distribute artemia homogeneously before the fish larvae were introduced.
To ensure all larvae had empty guts before the start of the trials, larvae were caught from the rearing aquarium the night before the trials (14 h without access to prey).This period was required, as a pre-trial showed slow gut evacuation rates (see supplementary information, Fig. S3).For the overnight gut evacuation, larvae were gently transferred with sieves into 24 transparent plastic containers (400 ml) at a mean (± SD) density of 41 (± 11) larvae.All containers were partially shaded with a semi-transparent black plastic foil, and transfer mortality was assessed the next morning.At the day of the trial, 20 larvae were individually pipetted from the plastic containers into 50-ml beakers (n = 24) containing the respective treatment's turbidity level.Larvae spent on average 16 ± 9 min (mean ± SD) in these 50-ml beakers before they were gently transferred into the 2-l experimental beakers and the trials commenced.
Trials were stopped after 4 h, as pre-trials with shorter trial durations (i.e.1-2 h) showed larvae had insufficient numbers of prey in the guts.Previous studies reported a 4-h timeframe to be sufficient for assessing prey consumption in fish larvae (Illing et al. 2015;Partridge and Michael 2010).Beakers were gently poured through sieves to remove the remaining artemia and retain the larvae.Larvae were kept in individual sieves, submerged in petri dishes, and visually assessed for their total number, the number of surviving larvae, and number of larvae with prey in their guts.These numbers were used to calculate fractions per beaker.Gut fullness could not be assessed because of limitations associated with larvae defecating prey during sampling.Additionally, photos were taken of each of the 24 sieves for digital verification of the counts.Four larvae from each beaker were quickly killed by asphyxiation in carbon dioxide-enriched water, photographed (Leica M165FC, Leica, Wetzlar, Germany, and Nikon SMZ25, Nikon Europe, Amsterdam, The Netherlands) and individually frozen for subsequent dry weight measurements.All other remaining larvae were kept alive and returned to a separate rearing aquarium.Length measurements were made by two independent observers using ImageJ (v.1.53k) and by analysing the images in FishSizer (v.3) (Rasmussen et al. 2022).Length measurements differed by 0.19 ± 0.39% across observers and by 1.23 ± 1.09% and 1.20 ± 1.01% from automated FishSizer measurements, respectively.Dry mass measurements were obtained by lyophilizing the frozen larvae for 24 h (Lyovac GT2, SRK-Systemtechnik GmbH, Riedstadt, Germany) and weighing the larvae on an ultra-micro balance (Sartorius MCA2.7S-2S00-MCubis II, Sartorius, Göttingen, Germany).

Systematic literature search
We used a systematic literature search to find studies investigating the effects of turbidity on prey consumption in larval and juvenile life stages of fishes.We followed the PRISMA 2020 recommendations for systematic reviews (Page et al. 2021) and modified search terms from a study reviewing the effects of turbidity on prey capture in aquatic predators (Ortega et al. 2020).We ran the search terms and Boolean operators [("predation rate*" OR "consumpt* rate*" OR "ingest* rate*" OR "forag* success") AND ("turbidit*" OR "transparen*" OR "sediment*") AND ("fish*" OR "shark*"OR "ray") AND ("larv*" OR "juv*")] in Web of Science (WOS) and Scopus to specifically search for studies on fishes and their early life stages.The literature search was conducted on 20 April 2022 and used the WOS's topic section and WOS Core Collection (266 studies retrieved) and the Scopus search within article title, abstract and keywords (52 studies found).Eleven additional studies were identified from other sources, i.e. reference lists.After removing 44 duplicate entries, 276 records were screened for suitability using titles (100 exclusions), abstracts (119 exclusions) and full texts (25 exclusions) (see supporting information Fig. S4 for PRISMA work flow).Criteria for inclusion were that studies were original research (no reviews) and experimental work (no modelling) and that larval or juvenile fishes were examined for their prey consumption in turbid water.
Search and study screening resulted in 41 full-text articles (see supplementary information for their references), from which further study characteristics were retrieved, e.g. the taxonomic groups of investigated fishes, the tested life stages, turbidity ranges and levels, as well as the type and amount of provided prey items.Turbidity levels from two studies using Secchi depths (Jönsson et al. 2011;Snow et al. 2018) were converted to NTU using the equation turbidity (NTU) = 9.4109 secchi depth (m) −1.243 ( B a u g h - man et al. 2015) (similar to the power equation in Carter et al. 2010).Kaolin concentrations from another two studies (Ohata et al. 2011a(Ohata et al. , 2011b) ) were converted to NTU based on the equation kaolin mgl −1 = 0.744 turbidity (NTU) , as per the technical report of the instrument used in these studies.Study characteristics were summarized using descriptive statistics, and indexing keywords, i.e. words or phrases that frequently appear in the titles of an article's references but do not appear in the title of the article itself, were analysed using the 'bibliometrix' package (Aria and Cuccurullo 2017) in R (R Development Core Team 2023).

Statistical analyses
All statistical analyses were conducted within the R statistical and graphical environment (v.4.1.1)(R Development Core Team 2023).Prey consumption and survival were analysed with Bayesian ordered beta regression models to account for the data being distributed between zero and one using the packages 'ordbetareg' (v.0.2.1) (Kubinec 2022) and 'brms' (v.2.16.3) (Bürkner et al. 2023).All candidate models were run with three chains, 3000 iterations, a warmup phase of 1000 and default, non-informative priors.Model selection was based on leave-one-out cross-validation techniques supplied by the 'loo' package (v.2.5.1) (Vehtari et al. 2023).
Prey consumption and survival data were best explained by multiple linear regression models containing the continuous explanatory variable 'turbidity' with a spline smoother and a categorical 'trial' covariate representing the four different experiments.The length-weight relationship was analysed using a Bayesian log-log linear regression model with a Gaussian family.In this model we included the categorical variable 'trial' as a random slope and the categorical random intercept variable 'artemia', which indicated the presence of prey in the gut.

Prey consumption
The proportion of European smelt larvae with prey in the gut followed a right-skewed, hump-shaped curve across the different levels of turbidity with an optimum between 100 and 200 NTU at 0.49 [0.13, 0.87] and 0.60 [0.13, 1.00], respectively (Fig. 2A) (all values provided as medians and 95% lower and upper quantile intervals [QI] [QI low , QI high ]).At the extremes, i.e. at 0 NTU and 500 NTU, the proportion of larvae with prey in the gut was 0.38 [0.00, 0.75] and 0.24 [0.00, 0.65], respectively.Thus, there was a positive effect of turbidity on prey consumption at 100 and 200 NTU and a negative effect at higher turbidity levels (≥ 300 NTU) (Fig. 2B).More specifically, a 1-unit increase in turbidity increased  or decreased (300-500 NTU) the proportion of larvae with prey by up to 2%.These results were not different across the four trials (Fig. 2C).Overall, the selected model was able to explain 38 [25, 48] % of the observed variability (Bayes_R2).The post hoc analysis revealed considerable differences between 0 NTU and 200-500 NTU and between 100 and 400 NTU (Fig. 2D).

Survival
Survival of European smelt larvae was high at turbidity levels of 0 to 200 NTU, ranging from 0.78 [0.40, 1.00] at 0 NTU to 0.83 [0.46, 1.00] at 200 NTU (all values as medians, 95% QI [QI low , QI high ]).At a turbidity of 300 NTU, survival started to decrease (0.80 [0.41,1.00])and was as low as 0.60 [0.23, 1.00] and 0.40 [0.00, 0.87] at 400 and 500 NTU, respectively (Fig. 3A).The estimated marginal mean decreased about 1% for a 1-unit change in turbidity, starting at 300 NTU (Fig. 3B).Although there was some evidence that the 1 week older larvae (trials 3 and 4) had slightly lower survival rates, overall survival rates did not differ across the four acute exposure trials (Fig. 3C).The selected regression model explained 51 [38,59] % of the observed variability (Bayes_R2).Pairwise comparisons between survival estimates at different turbidity levels revealed some evidence for higher turbidity levels (i.e.> 300 NTU) decreasing survival, with strong evidence (credible intervals not intersecting with zero) at the contrasts of 0 vs 400 and 100 vs 400 (Fig. 3D).The overnight mortality of larvae selected for the experiments (defecation of ingested feed prior to trials was required) increased over the duration of the trials from (mean ± SD, each trial n = 24 beakers) 2.3 ± 3.4 and 3.1 ± 2.7% (trials 1 and 2) in the first week to 12.8 ± 4.6 and 18.5 ± 6.5% in the second week (trials 2 and 3).

Length-Weight
The relationship between standard length (L) and dry mass (W) (Fig. 4A) differed across some of the experimental trials (i.e.3-1 and 4-1) (Fig. 4B), resulting in the following equations: W 1 = −0.79± 0.33, L 1 2.71±0.15, andW 4 = −1.79± 0.44, L 4 3.13±0.19, respectively (mean ± SE of the posterior distribution provided).Therefore, larvae gained around 15 to 23 µg dry mass per millimetre standard length.Larvae in trials three and four had gained lower weight before the experimentation compared to larvae in trials one and two according to their length-weight relationships (Fig. 4B).The presence of Artemia nauplii in the gut had a positive effect and increased dry mass on average by 1 µg (Fig. 4B).Overall, the regression model explained 82

Systematic literature search
The systematic literature analysis revealed that the effects of turbidity on prey consumption in larval and juvenile fishes had been investigated in 41 research articles, published between 1981 and 2020.The qualitative analysis of keywords or phrases consisted of 14 terms that occurred at least five times in the titles of the article's references (available since 1991) (see Fig. 5A).The terms represented the initially used search terms and indicated no strong temporal trends but highlighted important additional factors (e.g.light and light intensity, growth and body size) and showed the relevance of specific species (e.g.largemouth bass, Micropterus salmoides, Lacepède 1802, and bluegill Lepomis macrochirus, Rafinesque 1810).
In most experiments, juvenile (81%) rather than larval life stages (19%) were investigated, with most species belonging to the orders of Perciformes (37%), Salmoniformes (13%) and Osmeriformes (13%) (Fig. 5B).We grouped the type of prey based on their occurrence in the water column (Fig. 5C) and found that prey consumed during experimental trials consisted mostly of pelagic (or limnetic) species (e.g.zooplankton, fish larvae) (72%) followed by benthopelagic species (e.g.mysids and shrimps) (18%) and benthic species (e.g.chironomids, oligochaetes) (9%).In a single experiment, fruit flies (Drosophila melanogaster, Meigen 1830) were provided as prey on the water surface.In the bulk of the experiments, single levels (or concentrations) of prey were used (82%), with 18% percent of the experiments testing two or more prey levels.Most studies used clay minerals (e.g.bentonite 38%, kaolinite 8% or calcarenite 3%), natural sediments (31%) or algae (20%) to create turbidity.Other materials, such as diatomite, volcanic ash or cyanobacteria, were not included in the analysis as the concentrations could not be converted to NTU (Fig. 5D).In most experiments (74%), up to five different levels (or concentrations) of turbidity were tested, with only a few trials consisting of six, seven or eight turbidity levels (26%).

Discussion
In this study, we found that acute exposure to high turbidity negatively affected prey consumption of European smelt larvae.An optimum was found between 100 and 200 NTU, proving that moderate turbidity has a positive effect on prey consumption.Survival started to decrease at turbidity levels > 300 NTU.A systematic literature analysis showed that prey consumption of early life stages of fishes had been tested across a wide range of turbidity levels, mostly using pelagic (e.g.planktonic) prey items, with more studies focussing on perciform fishes and juvenile rather than larval life stages.
Tidal river estuaries naturally possess zones of increased turbidity but anthropogenic factors, such as dredging activities for maintenance of waterways and altered river runoff associated with climate change, can exacerbate turbidity levels.For example, in the lower stretches of the Elbe River (Germany), turbidity has increased noticeably within the past decade (see Fig. 1).This has been partially related to a below-average freshwater discharge since 2013 and a significant increase in the maintenance dredging quantities of fine-grained sediments that doubled from an average of 4.5 million m 3 year −1 (2008 to 2013) to 8.5 million m 3 year −1 (2014 and 2019) (Weilbeer et al. 2021).Shallow embayments within the Elbe River are considered important nursery areas for fishes, such as the European smelt (Sepúlveda et al. 1993); yet, these areas have been found to be particularly exposed to industrialization (e.g.partial filling for land reclamation), elevated turbidity and increased sedimentation, reducing the amount of favourable habitat for early life stages of European smelt (Scholle and Schuchardt 2020).In our experimental setup, we aimed at mimicking the range of turbidity levels measured in the Elbe and Weser rivers.Although average turbidity is rarely > 200 NTU (see Fig. 1), the observed peaks may suffice to negatively affect early life stages of European smelt directly by affecting larval performance traits, such as prey consumption.
The impact of turbidity on fish larvae is quite ambivalent (Arevalo et al. 2023).Turbidity has been found to provide refuge for fish larvae and reduce predation mortality by visual predators, sometimes resulting in reduced stress levels, higher growth rates and elevated ichthyoplankton abundances under turbid conditions (Bestgen et al. 2006;Carreon-Martinez et al. 2014;Lehtiniemi et al. 2005;Pangle et al. 2012;Pasparakis et al. 2023;Sirois and Dodson 2000;Suzuki and Pompeu 2016).However, several studies found that turbidity negatively affected prey consumption (e.g.feeding rates) of planktivorous fishes (Breitburg 1988;Figueiredo et al. 2015;Gardner 1981;Lowe et al. 2015;Wellington et al. 2010).Many studies observed, similar to Fig. 4 Length-weight relationship of European smelt (Osmerus eperlanus) larvae (18-26 days post hatch, n = 422).A Overall relationship of larvae with (y, red) or without (n, grey) Artemia nauplii in the gut is shown across all four experimental trials after an acute exposure to different turbidity levels (0 to 500 nephelometric turbidity units, NTU) (raw data shape coded).From the model's posterior distribution, the median is shown as the black regression line and the blue areas show the 0.5, 0.8 and 0.95 quantile-based credible intervals, respectively.B The model's post hoc contrasts are visualised with results from the posterior distribution presented (medians and 0.8 and 0.95 highest posterior density intervals).Highest posterior density intervals not intersecting with zero provide evidence for differences in the length-weight relationship across trials (comparisons 1-4) and between larvae with and without Artemia nauplii in the gut (y-n) our findings, hump-shaped relationships between prey consumption and turbidity levels in larval and juvenile fishes, such as Pacific herring (Clupea pallasii, Valenciennes 1847) (Boehlert and Morgan 1985), yellow perch (Perca flavescens, Mitchill 1814) (Manning et al. 2014;Pangle et al. 2012), black and white crappie (Pomoxis nigromaculatus, Lesueur 1829, and P. annularis, Rafinesque 1818), bluegill sunfish (Lepomis macrochirus) (Pangle et al. 2012) and delta smelt (Hypomesus transpacificus) (Hasenbein et al. 2016).In these studies, prey consumption increased at lower and intermediate turbidity levels and decreased when turbidity was high, probably because of an initial increase in contrast of the planktonic prey at lower turbidity (Utne-Palm 2002) and impaired vision and elevated (oxidative and osmotic) stress levels at higher turbidity (Hasenbein et al. 2016).Interestingly, feeding rates of adult European smelt (Osmerus eperlanus), feeding on phantom midges (Chaoborus flavicans, Meigen 1830), also peaked at intermediate turbidities (Horppila et al. 2004).Overall, many planktivorous fishes show a unimodal trend in how turbidity affects prey consumption in their larval and juvenile stages across a wide range of aquatic ecosystems.
In previous studies, effects of turbidity on prey consumption in larval and juvenile fishes have been measured using a heterogenous array of methods.In the systematic literature search, we observed that differences in methods were mainly related to the (1) tested turbidity range and levels, (2) materials used for experimentally creating turbidity, (3) developmental stage of the tested fishes, (4) acclimation time to elevated turbidity, (5) exposure duration and (6) prey type, size and density during trials.We conducted descriptive and bibliometric analyses but refrained from a meta-analytical approach, including the use of subgroups or random effect models to account for some of the observed heterogeneity because we considered the variability in the used methods and measured responses too high to provide meaningful insights.Most studies found in this search aimed at simulating naturally occurring turbidity.However, turbidity differs largely across various aquatic ecosystems, and fishes experience-and may be adapted to-very different turbidity levels.For example, rivers in British Columbia (Canada) have been considered clear with ≤ 1 NTU (Harrison River) and turbid with 27-108 NTU (Fraser River) (Gregory and Levings 1998), while turbidity levels in the tidal Elbe and Weser rivers (Germany) can be much higher (see Figs. 1 and S1).In addition to the differences in turbidity ranges, several studies investigated turbidity effects on prey consumption only at a few distinct levels (i.e. in control-treatment approaches).
Approximately a quarter of the studies (see Fig. 5D) assessed prey consumption at more than five turbidity levels, aiming at detecting functional responses and characterizing performance curves.These are considered beneficial for assessing threshold levels and critical limitations to organismal performance (e.g.obtaining resources and survival) (Kingsolver et al. 2014).We think that the observed mixed trends (i.e.positive, negative and hump-shaped optimum curves) only partially reflect species-specific differences in prey consumption.The difference in these trends could be an artefact of testing a few narrow and discrete turbidity levels rather than continuous gradients that span a fish's full performance breadth including potential upper and lower limits.However, as described above, a multitude of other factors could contribute to the different trends.
Materials for artificially creating turbidity ideally remain in suspension for maintaining consistent conditions and do not change the physicochemical properties of the aquatic experimental setups.In the systematic literature analysis, we found that a number of organic substances and mineral components were used.These consisted of algae, such as the flagellate Brachiomonas submarina, Bohlin 1898 (Lehtiniemi et al. 2005;Salonen and Engström-Öst 2010), or the green algae Nannochloropsis spec., Hibberd 1981 and Scenedesmus acutus, Meyen 1829 (Hasenbein et al. 2016(Hasenbein et al. , 2013;;Radke and Gaupisch 2005), and several mineral clays (e.g.bentonite, kaolinite).In addition, natural sediments were often collected, further processed (i.e.filtered through 40-63 µm mesh and autoclaved) and used for creating turbidity (Gregory and Northcote 1993;Li et al. 2013;Lowe et al. 2015).Regarding the mineral components, the comparability of experimentally created turbidity and field measurements may be hampered by the homogeneous size and reflectivity of certain substances, e.g.kaolin particles, in laboratory setups (Chacin and Stallings 2016).The most commonly used mineral substance is bentonite, a colloidal clay with the ability to elevate pH levels in aqueous suspensions (Kaufhold et al. 2008).In fact, in our trials we used bentonite and observed an increase in mean (± SD) pH from 8.09 (± 0.03) to 9.22 (± 0.23) at 0 and 500 NTU, respectively (see Fig. S6).This increase in pH may both affect the overall solubility of bentonite at higher turbidity (see the nonlinear relationship between measured turbidity and bentonite mass in this study, Fig. S2B) and physiologically affect fishes through changes in acid-base regulation and an inhibition of sodium uptake and ammonia excretion (Scott et al. 2005).However, algae-based turbidity may also result in gradually increasing levels of pH (and DO) through photosynthetic processes, and a behavioural study showed that several freshwater fishes avoided pH levels > 9.5 (Serafy and Harrell 1993).In contrast, Phan et al. (2020) used kaolin clay for creating turbidity and observed negligible variation in pH with 8.00 and 8.05 at 0 and 700 NTU, respectively.Prospective work on indicator species, such as the European smelt, who are sensitive to changes in water quality (Thomas 1998), can benefit from such a comparative overview regarding designing experimental setups and contextualizing findings.
European smelt requires good habitat quality for recruitment success, and its spawning sites are vulnerable to physical disturbance associated with increased urbanization, river watershed development and industrialization in many tidal stretches of rivers (Graham et al. 2021).Understanding how changes in turbidity directly affect prey consumption and survival in larval European smelt will therefore assist in developing future experimental approaches as well as mechanistic models and assessing potential consequences for recruitment dynamics.Due to the ambivalent nature of turbidity and as turbidity is typically correlated with other variables, the exact thresholds at which the relationship of turbidity and prey consumption is reversed have to be determined.This is particularly true in the estuarine turbidity maximum zone.Due to the hump-shaped nature of the relationship, an ambivalent relationship of turbidity and recruitment is highly likely as well.Moderate turbidity, with its positive effects on hunting success, may produce a positive correlation of turbidity and larval fish recruitment (Harris and Cyrus 1995).Very high levels of turbidity, which were shown do decrease prey capture success (Engström-Öst and Mattila 2008;Ljunggren and Sandström, 2007;Stuart-Smith et al., 2004), however, may negatively correlate with turbidity.Despite the positive effects of moderate turbidity, very high levels have been shown to reduce physiological condition and increase larval mortality (Griffin et al. 2009;Grimaldo et al. 2020).Given the current decline in European smelt population sizes in many German rivers (e.g.Elbe and Weser rivers) (Scholle and Schuchardt 2020), the important role of European smelt in the food web and simultaneously increasing turbidity levels, our empirical findings contribute to establishing thresholds for optimal larval European smelt performance under increased turbidity to the same extent as for the negative effects.
Still, further and more comprehensive research is required to assess how early life stages of European smelt will be affected in their vital rates and overall survival by elevated turbidity and other factors.Future studies on European smelt could, for example, evaluate effects of turbidity levels on physiological and metabolic traits, investigate the sensitivity across additional life stages, sizes, body conditions and growth after prolonged exposure to turbidity and combine findings from controlled (short-and long-term) laboratory experiments with field investigations (e.g.fish abundances and prey dynamics).Therefore, regular monitoring of different life stages of European smelt will be critical to better assess and understand temporal changes in fish biomass and abundance.This monitoring should, next to information on temperature and salinity, also collect data on turbidity.Lastly, it will be crucial to also look into the potentially interactive effects of turbidity with other relevant abiotic and biotic factors (e.g.dissolved oxygen, pH, temperature, prey quantity and quality, predator presence) for holistically evaluating the drivers affecting performance and survival of early life stages of European smelt.

Fig. 2
Fig. 2 Proportion of European smelt larvae (Osmerus eperlanus)with prey in the gut after acute exposure to different turbidity levels (0-500 nephelometric turbidity units, NTU).A Overall effect of turbidity on prey consumption in larval European smelt is shown using predicted Bayesian posterior distributions with the regression line representing the median and the coloured areas the 0.5, 0.8 and 0.95 quantile-based credible intervals, respectively.Raw data (shape coded) were collected in four separate trials.C Predictions were made at the originally chosen turbidity levels, and juxtaposed medi-

Fig. 3
Fig. 3 Survival of European smelt larvae (Osmerus eperlanus) after an acute exposure to different turbidity levels (0-500 nephelometric turbidity units, NTU).A Overall effect of turbidity on survival in larval European smelt is shown using predicted Bayesian posterior distributions with the regression line representing the median and the coloured areas the 0.5, 0.8 and 0.95 quantile-based credible intervals, respectively.Raw data (shape-coded) were collected in four separate trials.C Predictions were made at the originally chosen turbidity levels, and juxtaposed medians and 0.8 and 0.95 quantile-based credible

Fig. 5
Fig. 5 Characteristics of studies investigating turbidity effects on prey consumption in larval and juvenile fishes.A Temporal trends of indexing keywords from these 41 articles analysed, occurring at least 5 times, i.e. words or phrases that appeared in the titles of an article's references (available since 1991).Symbols represent medians and bars show lower and upper 25th and 75th percentiles, respectively.An interactive network visualisation of all indexing keywords can be found online (Open Scien ce Found ation).B Overview of the investigated groups of fishes, separated by life stages.C Prey types used in the feeding trials, categorized by their position in the water column,