Feeding experiments were performed in the laboratory in September 2021 at the Research and Technology Centre West Coast (FTZ) in Büsum, Kiel University, Germany. Experimental copepods were caught in an artificially built lagoon in Büsum (54°08’01.78”N 8°50’32.11”O) with access to the Wadden Sea but without tidal influences.

The pelagic copepod Acartia tonsa (0.5–1.5 mm length) can be found year-round in coastal and estuarine environments at high biomasses (Brylinski, 1981). Acartia sp. use mechanoreception for feeding (DeMott and Watson, 1991; Gonçalves and Kiørboe, 2015) and are sensitive to low-frequency vibrations (≤ 1000 Hz; Yen et al., 1992), which makes it, in addition to its trophic role in ecosystems, the optimal model species for the present noise study. As prey, we used small (< 20 µm) highly motile phytoplankton instead of micro-zooplankton, for instance, ciliates, to exclude potential effects of noise on the escape behavior of small prey animals. The chosen phytoplankton prey, Tetraselmis chuii, is a genus of green algae within the order Chlorodendrales, characterized by a flagellated cell body. Species of this genus are found in both marine and freshwater ecosystems around the world including the German Wadden Sea and are widely used in copepod feeding experiments (Thor et al., 2002; Scholz and Liebezeit, 2012). T. chuii were cultivated and provided by BlueBioTech GmbH, Büsum.

A. tonsa were caught with four light traps (see Kühn et al., 2022) deployed for 1–2 h during dark hours (21:00–23:00). The light traps were positioned at 20–50 m distance to the shore towards the center of the lagoon and positioned on the bottom (1.4 m depth). At this location, there is no direct input of underwater noise through boat traffic and pilot sampling showed a permanent high catching success of A. tonsa individuals. Animals caught in the traps were carefully poured into two cooling boxes where they were maintained (<2 h) until selected in the laboratory for experimental acclimatization. A new population sample of A. tonsa was caught for every experimental day. Additionally, the ambient sound of the lagoon was recorded with a SoundTrap-HF [sampling rate: 96 kHz; calibration Information; End-to-End: 176.3 dB (High); RTI Level @ 1kHz 135.4 dB re 1 μPa; Ocean Instruments, Auckland, New Zealand].

The caught copepods were brought to the laboratory and anesthetized with 15 g L⁻¹ (sea water) magnesium chloride (adapted from Isinibilir et al., 2020). The animals stopped moving after 1–10 min (high individual variations, pers. obs. SK and FK) and were sorted into groups of roughly same-sized, copepodite, and adult A. tonsa stages, irrespective of sex, under a stereomicroscope (Stemi 508, Carl Zeiss Microscopy GmbH). Nauplii stages were excluded from the experiments. The number of A. tonsa in experimental grazing plastic vials (110 ml Kautex wide neck PET containers) varied slightly with prey density (Table 1) and availability (number of copepods caught in light traps also depended on weather conditions and fluctuations in local natural density). In addition, the number of copepods as grazers was determined to ensure a reduction of optimally 40–50% in phytoplankton cell concentration under ambient sound conditions conservatively enabling the detection of measurable effects, if any, that could go both directions, i.e. decreased or increased feeding rates. Seven to 19 A. tonsa individuals per experimental unit were put together (Table 1), in Eppendorf tubes filled with purified (filtered, UV-treated, and ozonized) North Sea water (provided by IMTE, Büsum). The copepods were checked after 30 min and only those that displayed normal swimming behavior were selected for the experimental runs. This was tested by looking for an increased jumping behavior when triggered with white light while being in the Eppendorf tubes to reduce handling stress. Dead or non-normal swimming copepods (< 5%, pers. obs.) were sorted out. Copepods selected for the experiment were then put in Eppendorf tubes without lids (13 ml) closed with a mesh net (5 µm) and acclimatized for 12 h in the experimental aquarium in the dark without food. The aquarium (100 * 50 * 30 cm) was filled with 130 L purified North Sea water (provided by IMTE) that was filtered continuously. Water temperature, oxygen, and salinity were kept constant at 18 ± 0.2 °C, 9.5 mg L⁻¹ ± 0.1, and 36 mS cm⁻¹ ± 0.1, respectively. An overview of sample preparation, experimental design, and work flow is depicted in Figure 1. The experiments took place in a shed located around 100 m away from the main building of the FTZ to exclude low-frequency building sound during the experiments. The experimental room was additionally equipped with sound absorption material (molded pulp egg-texture cartons) on the walls. Further, similar to a setup used by Amorim et al. (2013), we reduced the influence of ground vibrations by placing the aquarium onto a box (120 * 80 * 16 cm) that was filled with a combination of fine and coarse sand up to the top. A marble slate (120 * 80 * 3 cm) was placed onto that box with 10 equally distributed circular rubber studs (6 cm in diameter and 3 cm in height) on which the aquarium was placed. The whole setup was standing on a wagon with rubber wheels.

After acclimatization, we obtained the functional response curves by quantifying the ingestion and clearance rates through vial incubations. For this, T. chuii was filtered through a 50 μm plankton mesh net to remove cell aggregates. The experimental prey densities (1000, 3000, 5000, 8000, and 10000 cells ml⁻¹) were prepared by diluting the start prey cell density with purified seawater and amending it with 40 μl 110 ml⁻¹ growth medium (provided by BlueBioTech GmbH) to avoid variations in phytoplankton growth between vials. For each phytoplankton cell concentration (prey density level) prepared, we also set aside and preserved vials for later confirmation of estimated initial cell concentrations while preparing the different prey density levels by adding 40 µl of Lugol solution (15 g KI + 500 ml dest H₂O + 10g I₂) (Almeda et al., 2018). Information on variations in the initial cell concentrations for the different densities can be found in Table S1 in the Supplementary Material. Grazer vials, each equipped with a group of copepods, were either exposed to ambient aquarium sound or to playback of harbor traffic noise. The days with runs alternated between ambient and noise treatments to ensure experimental copepods would have been exposed to similar environmental Wadden Sea conditions prior to being tested. Further, all five levels of different prey densities were tested simultaneously within a run. On each experimental day, one baseline control vial per prey density level was also included in the run to check for baseline phytoplankton mortality or growth in the absence of grazers during incubation. One to two experimental grazer vials on each experimental day were included in the incubation run to obtain the feeding rates of the copepods for all respective different prey density levels either exposed to ambient aquarium sound or to playback harbor traffic noise (Figure 1, Table 2). We consider each vial community as an experimental unit. Copepods are known to feed with abnormally high rates in the beginning of a feeding experiment because of starvation or handling stress (Mullin, 1963 and emphasized in Frost, 1972). Therefore, each run of baseline control vials and experimental grazer vials was incubated for 24 h. Additionally, we conducted the experiments in the dark to ensure constant nocturnal feeding conditions throughout the incubation (see Stearns, 1986). In all experimental grazer vials, copepods were allowed to graze down the suspensions without phytoplankton prey cell replacement.

We built a slowly rotating underwater plankton wheel for continuous phytoplankton mixing in the experimental vials. Pictures of the plankton wheel can be found in Figure S1 in the Supplementary Material. All experimental control and grazer vials were prepared with a window of 3.5 cm diameter fitted with a 5-µm mesh net to ensure exchange with the surrounding aquarium water. The plankton wheel was made of a stainless-steel circular wire frame (35 cm length, 8.5 cm width, and 21 cm width with added vials) onto which a matrix of three by five stainless-steel baskets were mounted. This was done for easy mounting and removing of experimental vials onto and from the plankton wheel. To each of the five columns, one prey density level had been assigned while within each column, the baskets in rows (three rows per column) were randomly assigned to control and experimental grazer vials within respective prey densities. The wheel was placed in the middle of the aquarium’s water column (4.5 cm of water below and above the vials) and was rotated by a synchronous AC 12 V Motor (CHANCS Motor, TYC-50; 1 rpm) connected to the plankton wheel via gear wheels and closed timing belts. An overview of the experimental setup and workflow is summarized in Figure 1 and Table 2.

The described workflow for the feeding experiments was conducted under ambient aquarium sound conditions and under the exposure of playback harbor traffic noise (Figure 1). Harbor traffic noise was recorded at the Büsum port (54°07’20.63”N 8°51’32.67”E) on June 8, 2021. The underwater sound recordings were done with an AS-1 hydrophone (Nauta Scientific, Milano; (cross-calibrated at the FTZ) sensitivity: −211 dB re 1V/µPa) that were connected with a P48 hydrophone preamplifier (26 dB gain) to a Zoom H5 Handy recorder (gain=3). The whole setup was calibrated at 1 kHz (sensitivity of −188 dB re 1V/µPa). The harbor noise recording reflects a typical composition of harbor traffic with consecutive passings of a small shrimp trawler, a foot passenger ferry, a former larger trawler now used by an operator for guided nature trips, and a sailboat, with a total duration of 20 ' 11". The first boat passed after 1’36” for 38 s. The second boat passed at 8’40” (49 s). The third one passed from 14’34” to minute 15’14” (40 s) of the recording and Boat 4 passed at 17’42” (60 s) (Figure 2A). All boats on the recording (sampling rate: 48 kHz) were passing at a distance of 10–15 m from the hydrophone. The hydrophone was positioned ~10 cm off the harbor wall at 1 m depth. The original recording was played back in the experimental aquarium with two UW30 underwater speakers (Electro-Voice; frequency response 0.1–10 kHz) connected to a self-designed mobile waterproof audio-suitcase amplifier (MAK2x30/4, Bela P. Event-Studiotechnik) that was connected to a laptop. The original sound file was played via Audacity (Version 2.3.0). For comparison of actual harbor noise and its representation in the laboratory, we played back the harbor traffic noise file using the section capturing the first boat passing and recorded it in an experimental vial mounted onto the plankton wheel at a fixed position with a SoundTrap-HF (calibration Information see above; sampling rate: 96 kHz). Finally, the distance of the plankton wheel to the two UW30 speakers and the gain that was added to the original recording for the exposure were both decided upon the playback sound’s similarity towards the original harbor recording. We analyzed the sound recordings in SpectraPlus-SC (V 5.3.0. 12A, Pioneer Hill Software LLC, Sequim, WA, USA) for frequency and power spectral density (PSD) characteristics visualized in spectra and spectrograms [FFT size: 16384 (for sampling rates of 48 kHz) and 32768 (for sampling rates of 96 kHz); Hanning window, 0.5 overlap]. For comparison, the root mean square (rms) power levels were calculated in SpectraPlus-SC from the PSD spectral data of the different sound recordings (Figure 2B). Based on analysis of original recordings and recordings of playbacks in the laboratory, we decided on the following sound exposure setup. One UW30 underwater speaker was placed on both sides along the plankton wheel with a distance of 15 cm to the experimental vials maintaining the noise exposure during the plankton wheel rotation on the right and on the left side. This recording of harbor noise was then played back in an infinite loop from the start to the end of each experimental exposure day via Audacity (Version 2.3.0).

The spectrogram of the measured original harbor noise recording (Figure 2A) and the spectra of the measured exposures in the aquarium versus harbor noise and ambient sound measured in the field (Figure 2B) are presented. To compare original recordings with its representation as experimental stimuli, Figure 2B shows the spectrum of the harbor noise at the time of the first boat passing (38 s) from the original harbor recordings (173 dB re 1 µPa² Hz⁻¹) in comparison to the recordings of the playbacks made in the vial in the aquarium (174 dB re 1 µPa² Hz⁻¹), the ambient aquarium sound treatment (155 dB re 1 µPa² Hz⁻¹), and the ambient sound in the artificial lagoon (144 dB re 1 µPa² Hz⁻¹). Note that ambient lagoon sound levels as recorded at the place of copepod origin are lower than ambient sound conditions in the laboratory that included potential noise from the water filter and the plankton wheel motor noise (Figure 2B). For boat noise in the original and in the playback recording, most energy was found between 100 and 2000 Hz. We are fully aware of the fact that the playback is not reflecting true harbor noise with shipping traffic found in the field. In an aquarium setup, the experimental sound exposures will divert from the real frequency-sound level distribution due to aquarium wall reflection, aquarium vibration, and the size of the aquarium that does not allow full soundwave cycles for low frequencies. Further limitations of the acoustic setup are pointed out in the discussion.

The feeding incubation was ended after 24 hours (except one case that ended after 28 h) by gently taking out each experimental vial from the aquarium plankton wheel. Each grazer vial was then visually checked for swimming activity. Here, individuals were noted as “not active” or “active”. A copepod was assigned and noted “not active” when there was no movement or only sinking after a few seconds when triggered with white light. Lugol solution 40 µl (15 g KI + 500 ml dest H₂O + 10 g I₂) was then added to the baseline control and grazer vial to preserve copepods and phytoplankton before taking out the next one. The plankton wheel was set on hold, and for the noise treatment, the playback was stopped only after all vials were taken out, preserved, and stored in a cooling box. In the laboratory, we counted the initial, baseline control and grazer prey cell concentration using a Fuchs-Rosenthal counting chamber (area: 16 mm², depth: 0.2 mm, volume: 3.2 μl) under microscopes (100x; Zeiss Axioscope and Leitz Aristoplan). After counting, copepods were removed from the vial and their prosome length (µm) was measured under a microscope using a calibrated Moticam X3 Plus laboratory camera (Software Motic Images Plus Version 3.0.19.108b) attached to a microscope (50 x; Leitz Aristoplan). Sizes were only noted for intact (not damaged) animals. Clearance rates (volume swept clear from prey cells in ml copepod⁻¹ h⁻¹) and ingestion rates (ingested prey cells h⁻¹ copepod⁻¹) were calculated after Frost (1972).

Descriptive statistics are presented as mean ± standard error if not stated otherwise. The cell concentrations between 1K and 10K cells ml⁻¹ were decided on being treated as a categorical fixed factor with five levels of prey cell concentrations (PC). However, choosing a regression model with cell concentrations as a continuous variable led qualitatively to the same results.

Statistics were performed in R version 3.4.4 (R Core Team, 2022). Figures for data presentation are based on the package ggplot2 (Wickham, 2016).

A linear mixed effect model (function lmer in package lme4; Bates et al., 2015) was fitted on the obtained data based on results per experimental vial community. We decided on using a mixed model with varying intercept per day to account for potential variation in the copepods that were caught daily from a certain population in the field. Hence, we included the ingestion rate of A. tonsa as the response variable and included the fixed effects of underwater noise treatment (A.N., noise vs. ambient), prey cell concentration (PC, five levels), mean body size per vial (MS, centered), number of copepods per vial (D, integer), and the number of animals “not active” after the incubation (S, integer). In addition to these fixed effects and continuous variables, we included day as a random factor to account for differences in general conditions among different experimental runs. We stepwise selected the optimal model with all reasonable combinations of the fixed variables (see Zuur et al., 2009) based on the lowest Akaike’s information criterion (AIC) with the fewest parameters (ΔAIC < 2; Burnham and Anderson, 2002). Note that p-values were computed using Kenward-Roger standard errors and df (package: jtools, Long, 2022).

Clearance rates were fitted to the Rogers random predator model as it accounts for prey depletion over time (Rogers, 1972). For this, we used the gnls function in R (package nlme; Pinheiro et al., 2022), in order to get the coefficient estimates for the ambient sound and harbor traffic noise treatment, and integrated the following equation:

Ne is the number of prey items eaten, N0 is the initial number of prey, a is the capture rate, and h and T are handling time and the incubation time, respectively. In order to receive the estimates for the clearance rates’ coefficients a and h, this equation was divided by the initial number of prey (N0) (similar to Hollings disk equation calculations in Schultz and Kiørboe, 2009). Note that the number of prey items eaten (Ne) is found on both sides of this equation which is solved by applying the Lambert W function (Bolker, 2008).

The experimental results on the copepods’ ingestion rates (y-axis) at increasing prey densities from 1000 to 10000 cells ml⁻¹ (x-axis) when exposed to ambient aquarium sound and playback harbor traffic noise are shown in Figure 3. Mean ingestion rates decreased by 48%, 24%, 64%, 48% and 29% at prey densities of 1000, 3000, 5000, 8000, and 10000 cells ml⁻¹, respectively, when exposed to playback harbor traffic noise compared to ambient sound conditions. In both sound treatments, ingestion rates increased with increasing prey density (Figure 3, Table 3). The optimal AIC-based model structure that describes the ingestion rates pattern is shown in Table 4 in bold (n_{each model} = 49, df = 8, AIC = 681, delta = 0.36). It was found that a combination, not an interaction, of the fixed variables sound treatment (A.N.) and density of prey cells ml⁻¹ (PC) described the obtained data from the feeding experiments best. Mean copepod lengths per vial (MS), the density of copepods per vial (D), and the number of animals “not active” in the end of the experiments (S) were not included in the optimal model. The optimal model (R² marginal = 0.76 and R² conditional = 0.77) shows that increasing prey densities (PC) from 1000, over 3000, 5000, 8000, to 10000 cells ml⁻¹, significantly increased copepod ingestion rates (Table 5, Figure 3). Specifically looking into the sound treatments, the model estimated a significant decrease by 330 ingested cells hour⁻¹ copepod⁻¹ when exposed to harbor noise compared to ambient aquarium sound conditions (LMM, n = 51, df = 3.92, t = −4.26, p < 0.010).

Figure 4 shows the clearance rates with increasing prey density. For the ambient sound conditions, the estimated capture rate a was 0.013, which significantly decreased by 0.008 under the exposure of harbor traffic noise (GNLS; n = 51, df = 47, t = −3.7, p < 0.001). There was no significant difference in handling time h between ambient (estimate 0.005) and harbor noise conditions (GNLS; n = 51, df = 47, t = −0.12, p = 0.9).

One of the mechanisms at hand would be masking or distraction. It is suggested that noise can mask essential natural sound cues for invertebrate settlement (as found in other meroplanktonic species Pine et al., 2012). Copepods detect prey through visual, chemical and mechanical signals (Fields, 2014), and the magnitude of importance for each detection mechanism is feeding mode- and species-specific (see Jakobsen et al., 2005; Fields, 2014), although mechanoreception may be the most common means of perception in different feeding modes (DeMott and Watson, 1991; Gonçalves and Kiørboe, 2015). Acartia tonsa is able to switch between feeding modes and prey types depending on the prey size (Jonsson and Tiselius, 1990), prey motility (Kiørboe et al., 1996), prey density (Kiørboe et al., 1996), and external disturbances such as turbulences (Saiz and Kiorboe, 1995; Kiørboe et al., 1996; Strickler and Costello, 1996). In ambush-feeding mode, copepods wait motionless in the water column until they perceive a hydromechanical signal generated by a potential prey, then reorientate themselves towards the prey, and attack it by directional jumps (Tiselius and Jonsson, 1990; Saiz and Kiorboe, 1995). On the other hand, feeding on small non-motile prey involves the generation of a feeding current with feeding appendages and thoracopods (Tiselius and Jonsson, 1990; Gonçalves and Kiørboe, 2015). Note that, in our current study, we cannot differentiate which feeding mode had been used. Saiz and Kiorboe (1995) studied the effect of turbulent water on the two feeding modes of A. tonsa and found only decreased clearance rates when exposed to turbulences exceeding natural turbulence levels in the field probably due to impaired prey perception (ambush feeding) and eroded feeding currents (suspension feeding). Even though turbulence and sound are not directly comparable, high levels of underwater noise could impair similar mechanisms as turbulences.

The detection of the potential prey, in general, depends on the strength of its velocity difference to the ambient to elicit a behavioral response of the copepod. In the copepods Labidocera madurae and Acartia fossae a velocity strength of only 20 µm s⁻¹ in the vibration frequency range from 40 to 1000 Hz is sufficient to trigger antennal neuroreceptors to fire (Yen et al., 1992) and a study on the escape response of A. tonsa has shown a threshold signal strength or velocity difference for deformation at 150 µm s⁻¹ and accelerations as low as 130 µm s⁻² in the near field of a siphon flow (Kiørboe et al., 1999). Due to the high sensitivity of copepods to fluid disturbances, the harbor noise exposure used in the present study may have been above detection thresholds, which could have led to masking or distraction, but further measurements of particle motion velocities in an aquarium setup are needed to test this.

Our results are inconsistent with a feeding experiment by Tremblay et al. (2019) in which A. tonsa was exposed to a noise egg, a waterproof device that produces low-frequency sound (de Jong et al., 2017). Nevertheless, they found a physiological response correlated with oxidative stress (Tremblay et al., 2019). Hydromechanical disturbances from different prey types are sensed by mechanoreceptive setae on the first antenna of calanoid copepods (Yen et al., 1992; Gonçalves and Kiørboe, 2015), such as Acartia sp. Solé et al. (2021b) performed an ultrastructural analysis of the setae on the first antenna of the ectoparasitic copepod species Lepeophtheirus salmonis, which uses mechanoreception similar to calanoids, but for host detection. They found that the setae had fused when L. salmonis was exposed to noise for 4 h. Maximum setae fusion occurred when exposed to a combination of 350 Hz and 500 Hz sound (cf. Figure 2). The mechanism was hypothesized to be related to oxidative stress, possibly followed by acoustic trauma. Similar mechanisms may also occur in A. tonsa upon exposure to harbor noise, where setae fusion may lead to impaired prey perception. However, the noise frequency ranges that affect A. tonsa most may be different from those of L. salmonis.

Whether the impact of noise on the ingestion rates of copepods feeding on phytoplankton is due to masking the hydromechanical signals of potential prey or distraction or related to physiological or morphological changes remains open. The magnitude and direction of responses in zooplankton when exposed to underwater noise is further most probably species- and stage-specific and depends on the sound source level and experimental design (Vereide and Kühn, 2023; Tremblay et al., 2019; Solé et al., 2021b). For future studies, we suggest a combination of empirical and modeling approaches investigating how noise impacts feeding in different copepod species, sexes, and stages.

We were limited to measure only the sound pressure part of the harbor traffic exposure even though particle motion is known being detected by invertebrates rather than sound pressure (Nedelec et al., 2016). Nedelec et al. (2015) and Simpson et al. (2016) measured 20–40 dB (µm s⁻²)² Hz⁻¹ higher particle acceleration in laboratory tank experiments compared to in situ recordings while sound pressure levels measured in these tanks were similar to the in situ recordings. We therefore may underestimate the true exposure in terms of particle motion. Further we would like to address that copepods most probably perceive velocity rather than acceleration (Kiørboe et al., 1999) which should be considered when reporting particle motion in sound-related future work on copepods. Playback of harbor traffic noise is partly distorted in small tanks from the original recordings (see Akamatsu et al., 2002; Jones et al., 2019) as seen in Figure 2B. However, at higher frequencies (>1000 Hz), such distortions may be less biologically relevant due to copepod sensitivity to lower frequencies (Yen et al., 1992; Solé et al., 2021b). Further, the continuous rotation of the plankton wheel during incubation imposes some variation onto the vials housing the experimental copepod communities per day, potentially leading to random noise exposures instead of the regular exposure from the looped playback. Differences in noise regularity, from e.g. ship traffic versus operational wind turbine noise, may lead to different behavioral outcomes (Nedelec et al., 2015).

We consider our results to be realistic for near-field shipping noise levels, for example in ports or along shipping lanes, servicing e.g. offshore wind farms or the oil industry and construction work.

Ingestion rates are known to be linearly correlated to egg production in A. tonsa and other calanoid copepods species (Kiørboe et al., 1985). Further, copepod growth is limited by food quantity (Anderson et al., 2021). Food quantity, in this case phytoplankton density, is known to vary throughout the year and being site- and species-specific. In general, phytoplankton densities ranging from low (×10³ l⁻¹) to high abundances (×10⁶ l⁻¹) especially during spring blooms (Lefebvre et al., 2011; Alprol et al., 2021). Our results on ingestion rates at different prey densities are therefore representative for natural phytoplankton abundances. Reduced feeding due to anthropogenic underwater noise at all phytoplankton prey densities, as presented in this study, may thus lead to decreased egg production, limited growth and development, and, in turn would lead to lower abundances of certain copepod species. Thus, the decrease of certain copepod species affects both the interactions to lower levels, e.g. phytoplankton and smaller zooplankton, as well as to higher trophic levels and thus may alter the fate of organic carbon’s transfer through the food chain (Steinberg and Landry, 2017) and its storage (Turner, 2015).

Our study did not consider ontogenetic and developmental aspects and sex differences. It is known that older copepod developmental stages are more mechanoreception-sensitive compared to younger developmental stages (Fields and Yen, 1997; Kiørboe et al., 1999) and hence adult animals may be more vulnerable to underwater noise. The effects of continuous noise on different developmental stages in copepods should be further investigated like done in a previous study on the effects of impulsive underwater noise on A. tonsa (Vereide et al., 2023). Further, differences in feeding efficiencies between female and male copepods are known but this difference is mainly explained by body size (van Someren Gréve et al., 2017), which was included in the statistical analysis of our study our study. Note that if noise affects the detection and response to hydrodynamic signals from prey, it could also alter the perception of fluid signals from potential mates and predators (Fields, 2014), thus affecting community dynamics.

At this point we cannot extrapolate from the feeding response of a single species on a single prey species when exposed to a single stressor to whole community-level dynamics. Our results, however, underline the need to further investigate the consequences of anthropogenic underwater noise on zooplankton.

In conclusion, we found that elevated noise levels similar to those measured the North Sea (Farcas et al., 2020; Kinneging and Tougaard, 2021) impair copepod feeding. However, noise exposure should be further investigated in the field to disentangle the potential effect of real-life shipping noise from playback noise in small tanks.