Potential impacts from simulated vessel noise and sonar on commercially important invertebrates

Husbandry

Field-collected animals were housed outdoors in ambient water and light conditions at the University of Rhode Island’s (URI) aquarium facility in Narragansett, Rhode Island. Individual plastic mesh bins were used to separate and prevent aggressive interactions and allow tracking of individual animals. Juvenile H. americanus (cephalothorax length ≈45–70 millimeters (mm)) were collected by the Rhode Island Department of Environmental Management Narragansett Bay ventless trap program. Juvenile C. sapidus (carapace width ≈50–80mm) were hand-collected in Point Judith Pond, Rhode Island. The animals were immediately transported to the URI flow-through facility, measured, and acclimated to the facility for at least 14 days. Crabs and lobsters were fed a rotating diet of frozen butterfish (Peprilus triacanthus), scup (Stenotomus chrysops), and squid (Loligo sp.), 3 days per week to satiation, with uneaten food removed.

Acoustic exposure and tank characterization

Animals were exposed to either low-frequency acoustic energy (simulated vessel noise, “Boat Noise”), mid-frequency acoustic energy (simulated sonar), or ambient silence (control). Our method of acoustic exposure relied upon a laboratory tank environment (Fig. 1). While in situ exposure would be optimal, a laboratory exposure allows more controlled conditions and easier recording of experimental trials. However, the sound field can be difficult to control in a tank setup. Previous exposure experiments (Celi et al., 2013; Wale, Simpson & Radford, 2013a; Wale, Simpson & Radford, 2013b; Filiciotto et al., 2014; Celi et al., 2015; Walsh, Arnott & Kunc, 2017), while suitably replicating the animal’s natural environment, may have inadvertently induced an uneven sound field. A sound source positioned at the side of a tank may produce unknown artifacts due to asymmetric reflection-based interference patterns. To overcome an asymmetric sound field, the vertical cylinder exposure setup for this study was based upon a modified standing wave tube design similar to that employed in previous studies (Fay & Popper, 1974; Hawkins & MacLennan, 1976; Wysocki et al., 2007; Dossot et al., 2017), and we characterized the sound exposure field to optimize the animals placement (Fig. 2).

For this study, the sound projector was placed at the bottom center of a double-walled (5 cm) fiberglass cylindrical tank measuring 1.8 m in diameter (Fig. 1A). The overall water depth was 1.5 m. The tank size allowed for four replicate exposure pens to be placed over the sound source (Fig. 1B). The pens were made from an acoustically transparent mesh and hung from above using polypropylene line attached to a wall outside the sound field, which decoupled any unintended mechanical vibrations into the pens. The multiple enclosures permitted four simultaneous exposures per experimental iteration. The lighting was strictly controlled and animal behavior was tracked using downward-looking GoPro® cameras.

Because bioacoustic tank experiments may harbor a sound field which may be different from an animal’s natural environment (Rogers et al., 2016), our goal was to create a suitably uniform sound field over the spatial extent of the pen enclosures and also over the frequency band of interest (below 4 kilohertz (kHz)). In addition, particle motion analysis has been identified as a prerequisite for acoustic ecology going forward (Popper et al., 2014; Hawkins, Pembroke & Popper, 2015; Nedelec et al., 2016; Hawkins & Popper, 2017; Hawkins & Popper, 2018; Popper & Hawkins, 2018). To empirically characterize the sound field, acoustic measurements were taken at a 5-centimeter (cm) sampling grid along a radial cross-section of the tank, from the surface down to the top of the sound projector in the center of the tank, and over to the edge of the tank. At each measurement location a series of tones swept between 60 Hz to 4 kHz determined the frequency response of the tank. Acoustic pressure levels were measured using an Underwater Sound Reference Division (USRD) F-42 hydrophone. The acoustic particle acceleration was measured using a prototype vector sensor developed specifically for the U.S. Navy by Wilcoxen Research Inc. Both the F-42 hydrophone and the Wilcoxen vector sensor were calibrated in USRD facilities, over the band of interest, using identical mounting fixtures employed during our tank characterization measurements.

An interpolated sound field map of the tank cross-section was calculated at individual frequencies and for signals used during each treatment (Fig. 2). At low frequencies (below 1 kHz), the sound field was fairly uniform and behaved similar to a standing wave tube apparatus because only one acoustic mode was excited. For example, at 300 Hz the sound pressure level radiated evenly above the source (Fig. 2A) and the maximum particle motion occurred at the pressure-release surface (Fig. 2B). However, at higher frequencies (above 1 kHz), the sound field became more complex because multiple acoustic modes existed within the tank (Figs. 2C and 2D). At frequencies above 1 kHz, a true standing wave tube would be too small to meet the spatial requirements for the animal husbandry aspects of the experiment. Therefore, at these higher frequencies, our goal was to create as uniform a sound field as possible and quantify any acoustic variability.

Both low- and mid-frequency exposure signals were played through an audio player for 50 min after a 10-min acclimation period. The 50-min period was divided into 10 min of ambient silence, 30 min of sound exposure, and 10 min of ambient silence. This method allowed for comparison of behavior before, during, and after sound exposure. Control animals were handled identically, but were only exposed to ambient silence.

The low-frequency treatment simulated merchant vessel noise by employing a recording downloaded from the URI Discovery of Sound in the Sea (DOSITS, 2016) audio gallery, which consisted of broadband energy with significant mechanical-borne harmonics (e.g., 60 Hz). The signal was low-pass filtered below 1 kHz (to ensure low-frequency content only) and transmitted from a USRD J-11 acoustic projector. Sound pressure levels were measured between 169–172 dB referenced to 1 microPascal (re 1µPa) over the 60 Hz to 1 kHz band (over the pen volume). Particle acceleration variability was measured to be within 6 dB re 1 µ-meter/s² over the pen floor’s spatial extent as it was suspended in the water column. This exposure was approximately representative of a close pass (<500 m) by a mid-size container vessel in a coastal area.

The mid-frequency treatment consisted of two simulated sonar signals transmitted in repeated fashion. A 1-s 1.67 kHz continuous wave pulse was followed by a 2.5 to 4.0 kHz 1-s linear frequency modulated chirp, with a 1-s pause inserted between each signal. Signals were transmitted from a Lubell Labs® source. Sound pressure levels were measured between 177–182 dB re 1 µPa over the pen volume. While some sonar signals can be well in excess of 200 dB re 1 µPa at 1 m, this exposure is representative of anticipated received levels similar to testing and calibration activities often performed by the Navy. For the 1.67 kHz signal, particle acceleration variability was measured to be within 8 dB 1 µ-meter/s² over the pen floor’s spatial extent. It is worthwhile to note that broadband nature of the chirp signal served to level out the high-frequency specific variability identified during the tank characterization.

Behavioral observations and physiological analysis

Animals in the stress portion of the study were acclimated for 7 days in the flow-through seawater system, then exposed (1 h) to either control (no noise, but in the experimental enclosure), boat noise, or simulated sonar (Fig. 1A). Experimental enclosures were 50 × 50 cm that included a triangular shelter of approximately 312 cm² in one corner (Fig. 1B). The animals were monitored for acute activity changes and shelter use changes through analysis of experiment videos in EthoVision® XT, and for physiological changes at 0, 1, 3, and 7 days post exposure through extraction of hemolymph from the branchial cavity of crabs and lobsters using a 20 gauge needle. Hemolymph was then measured for glucose levels using an Invitrogen™ Amplex Red Glucose test kit, and for HSP using an Invitrogen Qubit kit for total protein and a Sigma-Aldrich ELISA kit for HSP27. As mentioned earlier, HSPs are effective in measuring stress in whole animal homogenates or brain tissue in other work, though it is compared here for use in hemolymph, a non-destructive sampling method. All assays were measured on a Biotek™ ELx800 plate reader.

Competitive interactions

The impact of sound on competitive interactions was assessed using boat noise-exposed and control C. sapidus, introduced to similarly sized invasive C. maenas in the same arena described above, with the addition of a small piece of food being available in the corner opposite the shelter. C. maenas were chosen as the competitor since this species is present in similar numbers in the salt marshes where the C. sapidus were found, and has similar diet preferences, and would therefore be in direct competition with them.

Only the C. sapidus were exposed in this situation since they are broadly distributed in the water column, through both deep channels and the intertidal during high tide while C. maenas are more commonly found in shallow, intertidal areas. As boat noise at potentially damaging levels to crustaceans (e.g., >130 dB re 1 µPa, as reviewed by Edmonds et al., 2016) may not propagate far in water, though not enough data are available to make that conclusion, this could be a stressor affecting competitive interactions. Competition trials were performed at 24 h after exposure conducted in the same way as the previous experiment. Animals were acclimated to experimental conditions for 1 h in an adjacent holding tank, then monitored during a 1 h trial during which video was recorded via GoPro® cameras and behaviors categorized with EthoVision® XT. Since these are portunid (swimming) crabs that are found out on open substrate, behaviors were categorized as dominant (aggressive) behaviors (fighting instigator, food handling, food defense, merals spread, touch/push), submissive (active avoidance) behaviors (burrowing/burrowed, shelter use, rapid escape), or neutral (no clear dominant/submissive tendency) behaviors (climbing, being a victim/defender of attack, sitting in the open). Total seconds each crab performed each behavior was calculated and was then modeled using a Bayesian Linear Mixed Model, using an offset of total video time, to determine statistical differences (see below). Percent time spent was also calculated for all behaviors and was visualized as a donut plot (Wickham, 2016).

Statistical analysis and modeling

Physiological and behavioral response data were statistically modeled using linear mixed models (LMM) and Bayesian linear mixed models (Bayesian LMM) in R Studio (R Core Team, 2020). We included animal grouping as a random term for all LMMs in order to account for any potential pseudoreplication effects. To meet all assumptions and to avoid singularity, data from the animals’ HSP and hemolymph glucose response, along with resource competition, were modeled using Bayesian LMM (Chung et al., 2013; Lüdecke et al., 2020). For the resource competition models, in addition to the grouping parameter, C. sapidus–C. maenas pairing were also included as a random term. The LMMs were utilized to test changes in animal activity level (velocity) and shelter use, as these models did not have issues of singularity once random effects were included (Bates et al., 2015). For all models, model fit was determined (Lüdecke et al., 2021), performance was checked (Lüdecke et al., 2021), and output was visualized (Breheny & Burchett, 2017). All models met all model assumptions.

Behavioral observations and physiological analysis

Neither activity level (velocity) nor shelter use in C. sapidus and H. americanus varied significantly by exposure to either the mid-frequency or low-frequency treatments (LMM). However, there was a difference between the overall exploratory activity by species, with lobsters being more active than blue crabs (LMM, Conditional R² = 0.299, Species Fixed Effect Estimate = 11.346, Standard error = 5.001, t-value = 2.269, p = 0.04263).

With respect to physiological indicators of stress, there were significant differences in the stress levels indicated by hemolymph glucose after 7 days in both species (Figs. 3A, 3B, Figs. S1A, S2B), whereas HSP27 did not indicate any differences between treatments (Bayesian LMM). Both species showed significantly elevated glucose levels vs. controls 7 days after exposure to mid-frequency simulated sonar (Fig. S1A, C. sapidus, Bayesian LMM, Conditional R² = 0.429, Mid-Frequency Sonar Fixed Effect = 151.85, Standard error = 48.17, t-value = 3.152, p = 0.001619; Fig. S1B, H. americanus, Bayesian LMM, Conditional R² = 0.564, Mid-Frequency Sonar Fixed Effect = 152.73, Standard error = 37.74, t-value = 4.047, p = 5.1906 × 10⁻⁵) despite showing no immediate behavioral reaction to this treatment, and no elevation of glucose in earlier (24 and 72-h) samples (Figs. 3A, 4B). Other than the significant difference in glucose levels 7 days post exposure between control and mid-frequency simulated sonar exposed animals, no other time periods were significant.

Competitive interactions

Behavioral impacts on the competitive ability of C. sapidus persisted at 24 h post-exposure to simulated boat noise (the low-frequency treatment), and a change in behavior was also noted in its one-on-one competitor, an unexposed green crab, C. maenas, in the allocation of time for different activities during these trials. These changes were indicated by a change in allocation of time for a variety of behaviors coded from the video footage (Fig. 4). The behaviors observed for C. sapidus and C. maenas were related to each other, for instance C. maenas were frequently fighting victims of C. sapidus if they were exhibiting food defense behaviors. When C. sapidus were exposed to boat noise they displayed a decrease in time allotted to food handling, food defense, and instigating fights (Figs. 4A, 4B) and an increase in sitting in the open, rapid escape, and merals spread (an aggressive claw display). Conversely, the C. maenas competing with a boat noise exposed C. sapidus (Fig. 4D), exhibited a decreased its time fighting (all fighting behaviors), food defense, time in the open, and climbing (Figs. 4C, 4D) compared to C. maenas competing with an unexposed C. sapidus. When each behavior was modeled as a Bayesian LMM, the behaviors of pooled dominant and dominant food significantly declined for C. sapidus exposed to boat noise and increased for competing C. maenas (Fig. 5; Pooled Dominant Bayesian LMM–Conditional R² = 0.732; Dominant Food Bayesian LMM–Conditional R² = 0.664). For both Bayesian LMMs, the fixed factors of species (Pooled Dominant–ß = −1299.0 ± 376.7, t-value = -3.449, p-value = 0.000563; Food Dominant–ß = −1112.1 ± 332.9, t-value = −3.340, p-value = 0.00083) and the interaction between species and treatment (Pooled Dominant–ß = 1378.8 ± 575.4, t-value = 2.396, p-value = 0.0165; Food Dominant–ß = 1479.0 ± 508.5, t-value = 2.908, p-value = 0.00364) were significant, while the fixed factor of treatment alone (Pooled Dominant–ß = −1035.1 ± 1197.7, t-value = −0.864, p-value = 0.387; Food Dominant–ß = −1016.9 ± 619.4, t-value = −1.345, p-value = 0.251) was not significant. In the Bayesian LMM for neutral and submissive behaviors, there were no significant differences between treatments, species, and their interaction.