Aquariums as Research Platforms: Characterizing Fish Sounds in Controlled Settings with Preliminary Insights from the Blackbar Soldierfish Myripristis jacobus

Abstract

This study highlights the potential of aquariums as research platforms for bioacoustic research. Aquariums provide access to a wide variety of fish species, offering unique opportunities to characterize their acoustic features in controlled settings. In particular, we present a preliminary description of the acoustic characteristics of Myripristis jacobus, a soniferous species in the Holocentridae family, within a controlled environment at a zoological facility in the Canary Islands, Spain. Using two HydroMoth 1.0 hydrophones, we recorded vocalizations of the blackbar soldierfish in a glass tank, revealing a pulsed sound type with a peak frequency around 355 Hz (DS 64), offering a more precise characterization than previously available. The vocalizations exhibit two distinct patterns: short sequences with long pulse intervals and fast pulse trains with short inter-pulse intervals. Despite some limitations, this experimental setup highlights the efficacy of cost-effective methodologies in public aquariums for initial bioacoustic research. These findings contribute to the early stages of acoustic characterization of coastal fishes in the western central Atlantic, emphasizing the value of passive acoustic monitoring for ecological assessments and conservation efforts. Moreover, this study opens new avenues for considering the acoustic environment as a crucial factor in the welfare of captive fish, an aspect that has largely been overlooked in aquarium management.

1. Introduction

More than 1000 documented fish species emit active sounds across a variety of behavioral contexts, such as feeding, courtship, aggregation, fright, aggression, defense, and social interactions [1,2,3,4]. Passive acoustic monitoring (PAM) is a popular method that uses hydrophones to capture underwater sound, allowing for the detection of fish vocalizations and the monitoring of marine biodiversity [5,6]. The acoustic signals linked to specific behaviors, such as spawning, offer valuable insights for the management of wild populations [7] and play a crucial role in monitoring reproduction in captivity. This includes determining sexual status or maturity, assessing spawning readiness, and distinguishing between males and females, among other factors [8]. Furthermore, in the context of tank density, these sounds can serve as indicators for evaluating fish aggressiveness or overall well-being [9].

While bioacoustics research is widely applied across various aquatic environments, public aquariums offer a unique opportunity for such studies. The opportunity to conduct experiments with a wide variety of fish species in controlled conditions allows for the clear identification of sounds produced by specific species, as the controlled aquarium environment eliminates interference from other species and external noise, enabling precise attribution of sounds to the species under study. However, this does not apply to the broadband sounds produced by marine crustaceans, as tank acoustics substantially distort these sounds [10]. Despite some limitations in measurement precision, as noted by [11], these challenges can be mitigated in low-frequency sounds (i.e., below the minimum resonant frequency of the experimental tank) through careful experimental design and methodology [10]. Tank-based experiments afford researchers the opportunity to capture and analyze organisms’ sound production and their responses to auditory stimuli in a controlled and replicable environment [12]. Beyond their utility in bioacoustic studies, aquariums offer an ideal backdrop for investigating the behavior of aquatic life, including fish and other marine organisms. Access to these organisms in their natural habitats can be challenging, but the controlled environment of aquariums facilitates detailed observations [13,14,15]. Hydroacoustic methods employed within aquariums also hold promise for applications beyond research, particularly in aquaculture [16], demonstrating the use of hydroacoustics in monitoring and managing fish stocks in aquaculture farms, providing accurate estimates of fish density and biomass.

Despite these advantages, there are notable challenges associated with conducting bioacoustic studies in public aquariums. The importance of understanding the acoustics of small tanks for valid auditory research and the proper interpretation of results has been highlighted [17]. Overcoming these challenges requires attention to methodological aspects [11]. Another obstacle lies in the high cost of bioacoustic equipment, which can hinder the development of studies in public aquariums. However, recent innovations in the form of low-cost underwater recording devices present a potential solution [18,19], e.g., the use of open-source modified terrestrial recorders, waterproofed to convert them into hydrophones. These affordable alternatives have demonstrated promising results in recording underwater sounds and identifying different species. While the quality and performance of recordings from low-cost devices may not match high-fidelity equipment, they offer a viable option for long-term data collection. This accessibility enables the calculation of ecoacoustic indices and facilitates the monitoring of ecosystem health [20]. In essence, the integration of low-cost devices into bioacoustic studies in public aquariums opens avenues for more extensive and cost-effective research, contributing to a deeper understanding of aquatic ecosystems.

Building on these technological advances, passive acoustic monitoring (PAM) proves particularly advantageous in situations characterized by limited or unavailable sampling data. PAM excels in providing long-term datasets in remote environments, especially when the presence or absence of organisms can be discerned through their sound production [21]. The assessment of fish biodiversity patterns holds paramount importance in aquatic science and conservation. For effective utilization, fish diversity assessments benefit significantly from employing integrated complementary approaches. Passive acoustics, gaining increased attention, emerges as a non-invasive, long-term monitoring tool that leverages biological sounds, produced incidentally or intentionally, as natural tags for the identification and estimation of animal diversity [22].

The Holocentridae family, which includes eight genera and 90 species [23,24] of squirrelfish and soldierfish, is soniferous [9]. These reef fishes are capable of producing sounds using a pair of sonic muscles and their air bladder [25,26]. The calls of Holocentridae species consist of harmonics, with a dominant frequency between 80 and 130 Hz, and are composed of trains of 4 to 11 pulses [25]. The sound-producing mechanism involves the back and forth movements of articulated ribs, which are connected to fast-contracting sonic muscles that insert on the swimbladder [27]. The ability to produce sounds is present in Holocentridae species from different genera and regions, as well as at different developmental stages [28].

Myripristis jacobus was described as a soniferous species [29] in the early 1970s, but its acoustic characteristics were not published in detail and just described as intra-specific agonistic rumbles (frequency, 100–500 Hz; duration, 200–300 ms) [29]. M. jacobus is found in the western Atlantic from the northeastern Gulf of Mexico to Tortugas, Florida, and south to Rio de Janeiro. It has also been recorded from Cape Verde and the Canary Islands off the coast of West Africa [30].

The primary objective of this study is to demonstrate the potential of aquariums as research platforms that allow for the controlled, cost-effective, and species-specific characterization of fish sounds. By providing environments where external noise and interference from other species are minimized, aquariums offer a unique opportunity to gather preliminary acoustic data, particularly for species like M. jacobus that have been under-studied in this context. This study addresses two key research gaps: the absence of a detailed acoustic profile for M. jacobus and the potential of public aquariums to serve as effective platforms for bioacoustic research. By using a simple, low-cost experimental setup, we aim to demonstrate how aquariums can contribute to the preliminary characterization of fish sounds, even in species for which bioacoustic information is limited. This research contributes to the development of biodiversity and structural indices for marine ecosystems by providing a foundation for identifying fish species in field recordings through their acoustic signatures. By characterizing the sounds produced by species like M. jacobus in aquariums, future studies can utilize these data to identify species in natural habitats, ultimately allowing the calculation of biodiversity indices and the assessment of ecosystem structures based on acoustic data from field recordings.

Furthermore, this study utilizes M. jacobus as a model to evaluate the effectiveness of low-cost methodologies within a public aquarium setting. While providing more precise data on a species with previously limited acoustic information, this approach demonstrates the potential of aquariums as research platforms. By leveraging these tools, we can explore and document the bioacoustic diversity of aquatic organisms more comprehensively, opening avenues for further studies in both ecological research and conservation efforts.

2. Materials and Methods

2.1. Data Collection

Soldierfish (Myripristis jacobus) vocalizations were recorded in controlled conditions within a zoological facility located in the Canary Islands, Spain. The recording setup utilized a 0.64 m³ glass tank filled with seawater (salinity 31.1 ppt), maintaining a temperature range of 19.7 °C to 20.5 °C. The experimental tank was located in an auxiliary quarantine area, isolated from the main aquarium facilities. This location minimized external noise and vibrations that could interfere with the recordings. The only fish present in the area were the two M. jacobus individuals in the experimental tank, as the rest of the tanks in the section were occupied by corals. Staff presence in the area was limited to brief periods for coral monitoring, ensuring that the fish were largely undisturbed and isolated from other animals and human activity during the 24 h recording period. The tank was equipped with two independent HydroMoth 1.0 hydrophones (Open Acoustics, Oxford, UK), each featuring an analog MEMS microphone capable of recording at sampling rates up to 384 kHz. Recorded data were stored in uncompressed WAV format on microSD cards.

The HydroMoth’s signal-to-noise ratio has been estimated as 18.5 ± 1.45 dB (mean ± SE), and it was selected based on comparative tests, revealing a spectral representation closely resembling that of high-specification recorders [18]. Although no formal calibration was performed between the recording devices, the dimensions of the aquarium and the close proximity of the hydrophones to the fish ensured high-quality audio capture. The positioning of the devices minimized directional bias, allowing for consistency in the recorded sounds across both hydrophones. Sensitivity tests demonstrated consistent performance across different HydroMoth units, with a marginally lower sensitivity at lower frequencies (0.02–12 kHz) compared to other underwater recorders such as the SoundTrap 300 HF (Ocean Instruments, Auckladnd, New Zealand) [18].

To mitigate directional bias, and to reduce the attenuation of the low frequencies [10], the HydroMoths were strategically positioned in the middle of the tank, facing opposite sides (see Figure 1). Both HydroMoths recorded simultaneously, and for analysis, only the sound events with the best signal-to-noise ratio, as captured by either device, were used when the same events were detected by both. Each hydrophone was mounted on a PVC tube structure with a plastic base, elevating the devices 20 cm above the tank’s bottom. Additionally, a Garmin VIRB 360° camera (Garmin, Lenexa, KS, USA) was mounted atop the PVC tubes, accompanied by supplemental lighting to enhance visibility during the initial phase of the experiment. The tank was situated in the indoor technical area of the public aquarium at Loro Parque, with lights operational from 7:00 to 18:00, and switched off thereafter.

Recording parameters were standardized, configuring the HydroMoths to record for 15 min every hour at a 192 kHz sampling rate with the amplifier set at medium gain. Prior to introducing the soldierfish into the tank in October 2023, a 24 h background noise recording was conducted. Subsequently, two individuals of Myripristis jacobus from the zoological collection of Loro Parque (length = 14 cm) were placed in the tank for a 24 h period, during which the HydroMoths maintained the same initial configuration. The two M. jacobus individuals used in the experiment consisted of one male and one female. Both fish were in good health, as confirmed by the aquarium’s veterinary team, and were captive-bred from a reputable fish provider. A 360° video camera simultaneously recorded the first hour of the experiment to identify any additional sound sources, such as fish displacing or hitting enrichment devices in the tank. However, the camera’s definition did not allow for the detection of sound production by the fish.

2.2. Data Analysis

Spectrograms were generated using the Audacity software (version 2.4.2), available at http://audacityteam.org (accessed on 1 October 2024). A Hamming window with 1024 samples and 50% overlap was applied during the spectrogram creation process. Sound events within the spectrogram were then meticulously identified and manually extracted for further analysis. The criterion for defining the extent of an sound event was a minimum separation of 1 s between sound pulses; pulses separated by less than 1 s were considered part of the same sound event.

For a comprehensive analysis, all identified sound events were examined using a custom script in Python version 3.13 [31]. This script enabled the calculation of various parameters, such as frequency range, power spectrum, sound event duration, number of peaks, peak characteristics, and inter-peak intervals, providing a detailed characterization of the acoustic features associated with each sound event. Additionally, since all detected sound events fell within the 0 to 1500 Hz range, a 2 KHz low-pass filter was applied to minimize interference from potential reverberation effects in the tank [11].

Then, two linear regressions were performed using an R studio (v. 2023.06.0+421) script in order to statistically test the categorization of the sound events. First, to consider any significant differences of the peak frequency (Hz) depending on the number sound pulses of each recorded sound event. Second, to examine any significant differences of the average distance between peaks or average peak interval (ms) depending on the number of peaks of the recorded sound events. Possible significant differences in other acoustic features such as the minimum or maximum frequencies were not considered due to the similarity of the values recorded for all sound events.

3. Results

Each of the two Hydromoths generated 24 uncompressed WAV files, amounting to a total of 12 h of recordings. The two M. jacobus individuals produced sound patterns, which were recorded and analyzed over the 24 h period. The tank’s isolation from other species and limited external noise interference resulted in high-quality recordings, allowing us to distinguish between simple pulse sequences and more complex pulse trains. Initially, 71 sound events were detected. Of these, 22 instances were excluded due to a low signal-to-noise ratio. The remaining 49 sound events were considered the same type, as the acoustic features of the sound pulses were similar in all of them. In line with the necessity to streamline and standardize the description of fish sounds [4], we present the sound events herein in terms of their frequency, duration, or structure, and denote them as pulses or pulse trains, eschewing the traditional term ‘rumble’.

The predominant frequency distribution of sound pulses indicated that the peak frequency was concentrated around 0.3 kHz (mean = 355 Hz, SD = 64), with a minimum below 50 Hz (mean = 17 Hz, SD = 1), and a maximum around 0.7 kHz (mean = 676 Hz, SD = 97). The majority of sound events (n = 44, Figure 2) were characterized by either single pulses or small series, typically up to four pulses (mean = 1.64, SD = 0.78) (

Table 1. Characteristics of the sound events detected in the experiment.

Sound Class	Number of Pulses	n	Fdom (Hz) *	Fmin (Hz) *	Fmax (Hz) *	Pulse Duration (ms) *	Pulse Interval (ms) *
Pulse (s)	1	16	381 (92)	16 (0.2)	656 (22)	12 (3)
	2	14	358 (39)	16 (1)	658 (19)	20 (7)	328 (139)
	3	4	308 (53)	16 (0.1)	650 (0.4)	32 (13)	337 (164)
	4	4	362 (44)	16 (0.2)	491 (317)	23 (1)	222 (208)
Pulse trains	13–21	5	375 (20)	17 (2)	710 (21)	37 (3)	51 (16)
All	1–28	49	355 (64)	17 (1)	676 (97)	31 (10)	158 (158)

Fdom = dominant frequency, Fmax = maximum frequency, Fmin = minimum frequency. * mean (SD).

). The average duration of individual pulses was 31 ms (SD = 10 ms), with inter-pulse intervals spanning from 106 ms to 632 ms (mean = 158 ms; SD = 158 ms). Notably, sounds comprising a single pulse were the most frequently recorded (n = 23), followed by sequences of two pulses (15), three pulses in 5 instances, and four pulses in 1 instance.

Ten percent of the sound events (n = 5) Figure 3 were characterized by pulse trains with a variable number of pulses (mean = 16.6, SD = 4.1), ranging in duration from 0.34 s to 0.95 s (mean = 0.60 s, SD = 0.25) (Table 1). Inter-pulse distances varied from 38 ms to 67 ms (mean = 37 ms, SD = 3). A meticulous analysis of pulse trains was possible due to the high sampling rate implemented in this experiment.

4. Discussion

The recorded sounds in our experiment displayed a frequency range consistent with the calls documented in the available literature for the species [29]. Despite the absence of detailed calculations in previous studies, the reported frequency ranges align with those observed in our experiment. Notably, the ‘rumble’ sounds described by Bright (1971) [29] in a single instance closely resemble the pulse train sounds identified in our study, despite the lack of prior published references to the most common sound pulses produced by M. jacobus in our work. Pulsed sounds similar to those recorded in Puerto Rico during experiments in the 1960s (https://macaulaylibrary.org/asset/116776 (accessed on 1 October 2024)), where M. jacobus individuals were subjected to varying stress levels, exhibit characteristics consistent with sound pulses observed in our experiment—single pulses or short sequences of well-separated pulses. These sounds were produced under less stressful conditions, such as when a stick was moved in the tank. In contrast, pulse trains (‘rumbles’) were generated during intense manipulation by researchers and in response to some stressful and painful situations. This aligns with previous literature associating ‘rumbles’ with intra-specific agonistic behaviors [29]. Consequently, the reduced number of train pulses in our experiment can be explained by the fish being left undisturbed for 24 h, with potential stressors limited to intra-specific agonistic activities within the tank.

The findings of this study align with those reported for other members of the Holocentridae family, as documented in previous works by Winn (1963), Parmentier et al. (2006), and Parmentier et al. (2011) [25,26,32]. Although the peak frequency identified in this experiment is slightly higher than the predominant frequency reported for other species within the family, it is noteworthy that the existing literature consistently observes a concentration of amplitude towards the lower end of the spectrum. The pulse trains observed in the context of this experiment for M. jacobus are also in concordance with the established characteristics described for the Holocentridae family by Parmentier et al. (2011) [26].

While only one study has delved into the acoustic behavior of a coastal species in the western central Atlantic, namely, Similiparma lurida [33], notable parallels emerge in the utilization of single pulsed sounds or small sequences, and the frequency characteristics appear comparable, with S. lurida exhibiting a mean frequency of 394 Hz and a standard deviation of 98 Hz. Although pulse trains have not been explicitly documented for S. lurida, it is conceivable that they may exist, given the limited information derived from a single experiment. The convergence in frequency characteristics between two taxonomically distinct species poses a potential challenge for future automatic classification systems reliant on passive acoustic monitoring (PAM). This observation underscores the need for targeted research efforts to augment our understanding of the fine-scale temporal patterns inherent in fish acoustic signals, which may encode valuable interspecific information [34]. Consequently, there is a compelling call for further behavioral investigations to ascertain whether these nuanced temporal differences play a crucial role in species recognition, a factor of paramount importance for the refinement of PAM tools.

The experimental framework employed in this study, utilizing small glass tanks, imposes certain limitations on sound characterization, particularly concerning pulse durations and frequency distributions, as highlighted in works by Bamse et al. (2023), Akamatsu et al. (2002) and Jézéquel, et al. (2022) [10,11,35]. Notably, the frequency peaks recorded in our study occurred at frequencies significantly below the minimum theoretical resonant frequency calculated for the tanks, and the high frequencies were removed with a low-pass filter (2 kHz). Moreover, the attenuation distances allowed for a sufficient water volume where the fish were in close proximity to one of the two Hydromoths, ensuring a favorable signal-to-noise ratio. Remarkably, 70% of the detected sounds exhibited a reasonable signal-to-noise ratio and were consequently utilized for sound characterization. The strategic placement of two Hydromoths within the tank, facing opposite sides, not only addressed the asymmetrical sensitivity of the devices but also optimized sensitivity within the attenuation range of the tank. This underscores the need for more detailed studies of the sound production of Myripristis jacobus in low-noise anechoic conditions. In essence, although the experiment aimed to provide a preliminary description of the vocal activity of a single species, it is essential to approach the precision of the measurements with caution, recognizing them as an initial exploration of the sound characteristics of the species. While the use of small tanks introduces potential limitations related to reverberation and resonance, which can affect the propagation and quality of recorded sounds, this was not the primary focus of the current study. These factors have been extensively analyzed by Akamatsu et al. (2002) [11], who provide detailed guidance on mitigating these challenges in small-tank environments. For further information on how these conditions may influence sound production and capture, and recommendations for addressing such limitations, we direct readers to Akamatsu et al. (2002) [11]. Future studies aiming to address these challenges in more detail should consider the refinements and methodologies outlined in that work.

The distribution of sounds during the 24 h experiment was quite homogeneous, with no significant prevalence in sound production. The lighting conditions, alternating between on and off without dimming, may have influenced the sound production rate of the two individuals housed in the tank. It is well-established that the acoustic activity of fishes tends to escalate during dusk and dawn [36,37]. To enhance the yield of vocalizations for future experiments, especially those focusing on the fine-scale temporal patterns in sound pulses, incorporating a lighting setup that simulates a diel cycle with dusk and dawn phases could be a significant improvement.

Despite encountering challenges and experimental hurdles in the acquisition of recordings for M. jacobus, our findings underscore that a preliminary description of a species’ acoustic profile can be achieved within a simplified experimental setting. However, it is imperative to exercise caution in interpreting results, particularly given the acknowledged limitations in small glass tanks. This necessitates a call for further investigations under controlled conditions. The study delineates significant impediments, encompassing the acoustic characteristics of the tanks and ambient noise and lighting conditions. Addressing these challenges becomes imperative to refine experimental designs, while upholding the fundamental criteria of cost-effectiveness and simplicity, thereby mitigating the identified sources of primary bias.

The study also points out some important elements related to the characterization of fish bioacoustics, such as the influence of lighting experimental conditions on sound production, signaling avenues for potential enhancements in future experiments. These enhancements, facile to implement and cost-effective, underscore the importance of simulating a diel cycle to capture variations in acoustic activity. This could carry welfare implications for public aquarium collections subjected to abrupt light changes between day and night, emphasizing the significance of simulating dusk and dawn periods, in addition to diel cycles or seasonal changes in day length. This prompts exploration into the prospect of utilizing such studies as a potential tool for welfare monitoring.

The outcomes derived from our study hold paramount significance in the progression of bioacoustic research within public aquariums. The clear distinction between pulse sequences and pulse trains observed in this controlled setting demonstrates the value of aquariums in capturing species-specific sounds without interference from external noise or other species. These findings contribute to the broader field of fish bioacoustics by providing insight into how controlled environments can facilitate the study of species that have previously lacked detailed acoustic profiles. This study capitalizes on the utilization of low-cost devices while leveraging existing facilities and life support systems, thus circumventing substantial financial investments. Aquariums offer a unique and invaluable resource for advancing the field of fish bioacoustics due to their access to a diverse range of species and the ability to control experimental conditions in ways that are not possible in the wild. One of the significant advantages that aquariums can provide is the opportunity to conduct long-term experiments with fish of varying sizes. By utilizing different sized individuals within the same species, aquariums can facilitate studies that explore how size influences sound production, which is crucial for understanding species-specific communication and behavior. Additionally, aquariums can systematically group fish of different sizes within controlled tank environments, allowing researchers to discern how these variations affect acoustic characteristics. This ability to manipulate and observe fish under controlled conditions makes aquariums ideal for conducting detailed investigations into the relationships between fish size, sound production, and acoustic communication.

Moreover, the acoustic traits identified in this study offer opportunities for comparative research across other species within the Holocentridae family or other soniferous fish. Such studies can explore the evolutionary and ecological significance of sound production in these species, providing deeper insights into their communication strategies, mating behaviors, and territorial interactions. The data can also guide behavioral studies aimed at linking specific acoustic patterns to particular behaviors, such as aggression, courtship, or responses to environmental stressors.

In a broader context, this research paves the way for the development of PAM-based tools for tracking biodiversity and ecosystem health. Future studies can leverage these initial findings to calculate biodiversity indices and develop ecosystem structure models based on acoustic data collected in the field. This underscores the broader relevance of this study, which not only advances the acoustic characterization of M. jacobus but also sets the stage for more extensive research into the bioacoustics of other marine species.

5. Conclusions

This study demonstrates the potential of public aquariums as valuable platforms for bioacoustic research. By leveraging the controlled environment of an aquarium, we were able to successfully record and characterize the sounds produced by Myripristis jacobus, providing the first detailed acoustic profile for this species. The controlled setting allowed for the isolation of species-specific vocalizations without interference from other species or external noise, underscoring the role aquariums can play in advancing bioacoustic research.

The findings from this study have important practical implications, particularly for the identification of fish species in natural environments through passive acoustic monitoring (PAM). The acoustic data collected here can serve as a reference for field recordings, aiding in the identification of Myripristis jacobus and potentially other species with similar acoustic traits. This approach could facilitate more accurate assessments of fish biodiversity and contribute to the development of ecosystem monitoring tools.

Furthermore, the study highlights the feasibility of using low-cost methodologies in aquariums to collect high-quality bioacoustic data. This makes bioacoustic research more accessible to institutions with limited resources and opens up opportunities for long-term monitoring of species in both captive and natural settings.