Busy Urban Soundscape Underwater: Acoustic Indicators vs. Hydrophone Data

Abstract

Urban noise pollution extends into aquatic environments, influencing underwater ecosystems. This study examines the effectiveness of acoustic indicators in characterizing urban underwater soundscapes using hydrophone recordings. Three indices, the Acoustic Complexity Index (ACI), Acoustic Diversity Index (ADI), and Normalized Difference Soundscape Index (NDSI), were analyzed to assess their ability to distinguish anthropogenic and natural acoustic sources. The results indicate that the ACI tracks urban noise fluctuations, particularly from vehicles and trams, while the ADI primarily reflects transient environmental interferences. The NDSI, while designed to differentiate biophony from anthropogenic noise, proves unreliable in urban underwater settings, often misclassifying noise sources. These findings highlight the limitations of traditional acoustic indices in urban aquatic environments and emphasize the need for refined methods to improve hydrophone data interpretation. Thus, this study aims to understand the acoustic indicators’ interactions with underwater urban noise, which is crucial for enhancing environmental monitoring and noise mitigation strategies.

1. Introduction

Urban environments are characterized by high levels of anthropogenic noise, with road traffic, public transportation, and construction activities contributing to a complex soundscape. While extensive research has been conducted on the effects of urban noise on terrestrial environments, the impact of these soundscapes on aquatic ecosystems remains largely underexplored [1,2]. Sound is a crucial sensory modality for many aquatic organisms, influencing their communication, navigation, foraging, and predator avoidance [3]. The increasing levels of anthropogenic noise pollution in urban freshwater may significantly alter these behaviors, with potential ecological consequences.

Previous studies have highlighted the impact of noise pollution on marine ecosystems, showing that ship traffic and industrial activities contribute to changes in marine species’ communication and stress responses [4]. Research on freshwater systems, however, has been more limited, despite evidence that noise from bridges, water pumps, and urban runoff systems can introduce significant acoustic disturbances [5]. Studies conducted in riverine and canal environments have suggested that high noise levels can mask biologically important signals, leading to behavioral modifications in fish and amphibians [6,7]. These findings align with research on marine mammal communication, where anthropogenic noise has been shown to disrupt social interactions and foraging behavior [8].

Urban freshwater systems differ from open marine environments in several ways, particularly due to their spatial confinement and proximity to densely populated areas. Unlike in marine ecosystems, where noise pollution often originates from shipping lanes and offshore activities, urban freshwater bodies are exposed to a complex blend of local traffic, infrastructure noise, and recreational activities. These soundscapes are shaped by the reflection and refraction of sound waves off built structures, leading to unique acoustic environments that warrant further study [9]. Furthermore, the ecological impacts of urban noise pollution extend beyond individual organisms to entire aquatic communities. Studies on fish populations in noise-exposed habitats indicate potential long-term effects on species distribution and reproductive success [10]. Amphibians, which rely heavily on acoustic signaling for mating, may also be particularly vulnerable to urban noise interference [11]. Given these potential ecological consequences, it is essential to integrate bioacoustic monitoring techniques into urban freshwater research, allowing for a comprehensive understanding of how anthropogenic noise influences aquatic life [12,13].

Previous research has clearly demonstrated that anthropogenic noise in freshwater ecosystems significantly affects aquatic organisms’ behavior and reproductive success. For instance, recent studies have indicated that elevated noise levels can lead to altered calling behavior and mating success in anuran amphibians [11]. Similarly, anthropogenic noise has been found to interfere with critical biological activities in freshwater fish, such as reproduction, predator avoidance, and effective communication [7,14]. However, predicting specific ecological impacts in urban freshwater environments remains challenging due to their distinct acoustic and ecological characteristics compared to natural freshwater settings, where most prior studies have been conducted. In urban canals like our study area, acoustic environments feature high complexity and persistent anthropogenic noise, differing markedly from quieter, natural freshwater habitats typically studied in previous research. Consequently, predictions about animal behavioral responses based solely on the available literature might have limited accuracy.

Despite rising awareness of urban noise pollution, research on its impact on underwater environments is still in its early stages [14,15,16]. Most studies on anthropogenic noise in aquatic settings have focused on marine environments, with limited attention given to urban freshwater ecosystems [17]. Additionally, traditional ecological monitoring techniques often overlook acoustic assessments, leading to an incomplete understanding of the influence of urbanization on aquatic habitats [18]. The increasing availability of passive acoustic monitoring (PAM) technologies, such as hydrophones, has enabled more detailed investigations into underwater soundscapes [19]. Recent advancements in acoustic analysis techniques, including the use of machine learning algorithms to classify biological and anthropogenic sounds, have opened new possibilities for long-term monitoring and environmental assessment [20,21,22]. Such methodologies can provide critical insights into how noise pollution varies temporally and spatially, aiding in the development of mitigation strategies. While acoustic indicators such as the Acoustic Complexity Index (ACI), Acoustic Diversity Index (ADI), and Normalized Difference Soundscape Index (NDSI) have been widely applied in terrestrial soundscape analysis, their applicability and reliability in urban underwater settings remain uncertain. Our study’s scientific novelty lies in its methodology because it aims to overcome some of the unexplored and preconceived notions about how to differentiate between anthropogenic and natural sound sources. This study adds to the ecology of urban soundscapes by pointing out these drawbacks and highlights the necessity of more advanced techniques for evaluating acoustics in intricate, underwater habitats. This study seeks to bridge these gaps by evaluating the effectiveness of these indices in characterizing urban aquatic soundscapes and assessing the extent of anthropogenic noise influence.

The primary objective of this study is to analyze the acoustic environment of an urban waterway using hydrophone recordings and established acoustic indices. Specifically, the research aims to address the following:

• Assess the applicability of the ACI, ADI, and NDSI in quantifying soundscape characteristics in urban freshwater environments by comparing index performance and determining their sensitivity to both natural and anthropogenic sound sources.
• Examine the extent to which anthropogenic noise, including traffic and construction, influences underwater acoustic index results by identifying and evaluating dominant noise contributors.
• Identify patterns and variations in urban underwater soundscapes and evaluate their potential for determining whether the existing indices can reliably characterize long-term soundscape dynamics for environmental assessment and noise mitigation strategies.

To achieve these objectives, hydrophone recordings were conducted in a freshwater canal in a highly urbanized area. The recordings were analyzed using acoustic indices to quantify soundscape diversity, complexity, and anthropogenic disturbance. The findings of this study will provide valuable insights into the role of urbanization in shaping underwater acoustic environments and contribute to the growing field of bioacoustic monitoring in freshwater ecosystems and its ecological implications.

This paper is structured as follows: Section 2 describes the study site, data collection procedures, and acoustic analysis methods. Section 3 presents the results of the hydrophone recordings, including spectrogram analysis and acoustic index evaluations. Section 4 discusses the implications of the findings, comparing them with the existing literature and highlighting the potential ecological impacts. Finally, Section 5 concludes this study with recommendations for future research and potential applications of acoustic monitoring in urban freshwater environments.

2. Data and Methods

2.1. Case Study Area

This study investigates the influence of urban noise on the aquatic soundscape, selecting a site near a busy crossroads to examine human-induced sound pollution in a densely populated city center. The chosen location was an urban canal in Wroclaw, Poland, specifically along ks. Piotra Skargi and Podwale Street, which crosses underneath a heavily trafficked junction. This area was selected due to its high exposure to anthropogenic noise sources, including vehicular and tram traffic, as well as ongoing construction activities. The complexity of noise sources in this environment provides a representative example of urban aquatic soundscapes affected by human activity. The water body at the study site was characterized by a depth sufficient to allow for accurate acoustic measurements without significant interference from natural hydrodynamic factors, such as strong currents or waterfalls. The selected section of the canal minimized disturbances from surface noise caused by wind and waves while also reducing bottom noise from sediment movement. These conditions ensured that the recorded acoustic environment primarily reflected urban noise contributions rather than natural water dynamics. The hydrophone deployment was planned strategically to capture variations in the urban underwater soundscape. A calibrated omnidirectional hydrophone was positioned at specific locations to monitor the sound levels and assess their spatial distribution. To obtain representative measurements, the hydrophones were placed at varying distances from major noise sources, including roads, tram lines, and construction sites. This approach ensured that the recorded data reflected the interplay between urban noise and the aquatic environment while minimizing localized distortions caused by direct sound reflections from nearby structures. The chosen location included a bridge over the city moat, which provided an accessible and stable survey platform (Figure 1 and Appendix A). This bridge is located near one of the busiest intersections in the city, bordered by ks. Piotra Skargi Street, Podwale Street, and Hugona Kołątaja Street. As a major traffic artery surrounding the city center, this intersection experiences high levels of vehicular, tram, and pedestrian movement, making it an ideal setting for studying the impact of urban noise on an aquatic ecosystem. The proximity of multiple sound sources, combined with a controlled hydrophone deployment strategy, allowed for a comprehensive analysis of how urban noise propagates underwater and shapes the acoustic characteristics of the canal environment.

The decision to conduct recordings at a single position was driven by several practical and methodological considerations. The selected location represents a highly urbanized setting that integrates multiple significant anthropogenic noise sources, including intense vehicular traffic, frequent tram passages, pedestrian activity, and ongoing construction work. Due to logistical constraints associated with deploying hydrophone equipment safely and securely in urban public spaces, we prioritized one strategically positioned measurement point offering stable deployment conditions. Furthermore, this specific site near a busy crossroads allowed for capturing a representative sample of the acoustic environment, encompassing the cumulative effects of various noise sources common to densely populated urban waterways.

2.2. Urban Noise Environment and Acoustic Conditions

The urban canal at ks. Piotra Skargi and Podwale Street serves as a key transition area for residents and visitors, featuring vehicular, pedestrian, and bicycle paths. The area experiences distinct patterns of human activity, with peak visitation hours on weekdays starting from 5:00 a.m. and a late afternoon peak between 4:00 p.m. and 6:00 p.m. on Saturdays (Figure 2). These activity trends are significant as they contribute to the overall soundscape and highlight temporal variations in anthropogenic noise exposure.

The noise map of ks. Piotra Skargi and Podwale Street in Wroclaw reveal substantial spatial differences in noise intensity across the study area, with prominent clusters of high noise levels concentrated around major intersections (Figure 3). The most intense noise emissions, exceeding 79 dB, are concentrated in areas of dense vehicular flow, including cars, buses, trams, and bicycles. Such high-intensity zones are characteristic of urban crossroads, underscoring the significance of this study in understanding urban noise pollution and its aquatic implications. Surrounding the high-noise clusters, moderate noise levels ranging from 65 to 71.9 dB (orange) and 59 to 64.9 dB (light orange) are observed. These areas exhibit slightly lower traffic densities or benefit from partial sound barriers, such as adjacent buildings. Nevertheless, even these moderate-noise zones contribute significantly to the overall acoustic environment of the canal. The persistence of noise in these areas suggests continuous exposure of the water body to urban noise, which could have ecological consequences for aquatic species, particularly those reliant on acoustic communication for navigation, mating, and predation avoidance. The canal bisects the research area, with high-noise regions flanking both sides. This spatial arrangement results in consistent exposure of the aquatic environment to noise levels surpassing 65 dB—an intensity sufficient to interfere with the natural underwater soundscape. Prolonged exposure to such anthropogenic noise could disrupt key biological processes in aquatic organisms, including altered communication strategies, and behavioral changes.

The hydrophone was deployed at a depth of approximately 50 cm below the water surface. This positioning minimized interference from surface noise sources, including wind and wave action, while avoiding distortions caused by sediment movement at the bottom.

2.3. Hydrophone Recordings

The data collection period spanned from 11 March 2024 to 19 April 2024. This period was chosen as a representative period for urban underwater soundscapes during typical human activity times. This period includes stable levels of anthropogenic noise activity as well as seasonal stability, as it avoids severe weather conditions such as winter ice cover that may introduce another layer of noise variability. March–April also include seasonal urban activity patterns such as commuting, public transportation, and work weekdays, making it a suitable temporal lens for assessing urban noise in aquatic environments. In addition to the routine activities observed in the study area, an ongoing facade renovation of a nearby building was noted, introducing an additional source of intermittent noise. For this purpose, 30 recordings were made. Acoustic recordings were captured using an H2d Hydrophone from Aquarian Audio & Scientific, connected to a TASCAM DR-07X stereo digital audio recorder via a 3.5 mm input (Figure 4) [23]. Each recording session lasted between 15 and 30 min, ensuring sufficient data collection to capture variability in the soundscape. The recorded audio files underwent post-processing to extract relevant acoustic indicators, such as the Acoustic Complexity Index (ACI), which are presented in later sections. Audacity software (Version 3.7.1) was utilized for this purpose, employing the Notch Filter tool and the Noise Reduction feature. The noise profile was adjusted individually for each recording to optimize data clarity while preserving the integrity of the acoustic signals.

The detailed specifications of each instrument used in the acoustic measurements are summarized in

Table 1. Specifications of instruments used for acoustic measurements.

Instrument	Manufacturer	Model	Specifications
Hydrophone	Aquarian Audio & Scientific	H2d	- Frequency range: 20 Hz–20 kHz (±4 dB) - Sensitivity: −180 dB re: 1 V/µPa (±4 dB) - Polar response: Omnidirectional - Operating depth: Up to 80 m - Cable length: 3 m (shielded)
Digital audio recorder	Tascam	DR-07X	- Channels: Stereo (2-channel) - Recording format: WAV (16/24-bit, 44.1–96 kHz) - Frequency response: 20 Hz–20 kHz +1/−3 dB (Fs 44.1 kHz) - Input: 3.5 mm stereo mini-jack (plug-in power) - Storage: microSD/microSDHC/microSDXC (up to 128 GB) - Power source: Two AA batteries or USB power supply

, based on manufacturer datasheets.

The audio recordings were processed using Audacity (version 3.4.2), an open-source audio editor. This software was employed for preliminary data cleaning, noise filtering (via Notch Filter), and noise reduction (via Noise Reduction feature). Customized noise profiles were generated individually for each recording to optimize clarity while preserving acoustic integrity.

A quantitative acoustic analysis, including the calculation of the acoustic indices (ACI, ADI, and NDSI) and spectrogram analysis, was conducted using R (version 4.3.2), an open-source statistical computing software. The R packages employed in this study included the following:

• seewave (version 2.2.3): for spectral analysis, spectrogram visualization, and acoustic feature extraction.
• soundecology (version 1.3.3): to calculate ecological acoustic indices (ACI, ADI, and NDSI).

These software platforms and libraries enabled detailed, replicable acoustic analyses, providing a robust technical foundation for interpreting the recorded soundscape data. No machine learning-based libraries were applied in this stage of the analysis.

2.4. Acoustic Indicators

To assess the acoustic environment of the study area, three key acoustic indices were employed: the Acoustic Complexity Index (ACI), the Acoustic Diversity Index (ADI), and the Normalized Difference Soundscape Index (NDSI). These indicators provide a quantitative means of evaluating the impact of anthropogenic noise on urban aquatic ecosystems.

The Acoustic Complexity Index (ACI) is based on the observation that many biotic sounds, such as bird songs, exhibit a high degree of variability in intensity [24,25]. In contrast, human-generated noises, such as vehicle traffic and aircraft overflights, tend to be more consistent in their spectral patterns. By capturing the variations in intensity within a given frequency range, the ACI serves as a useful metric for distinguishing between natural and anthropogenic sound sources. This index has been widely applied in terrestrial bioacoustics and is increasingly being explored for underwater applications. The ACI is defined as follows:

$$AC{I}_j=\sum _{j=1}^{J-1}\left(\left|a_{j+1}-a_j\right|\sum _{j=1}^J{a}_j\right)$$

where

• a_j: the amplitude of the acoustic signal at the time step j;
• a_j+1: the amplitude of the acoustic signal at the subsequent time step (j + 1);
• J: the total number of amplitude values within the analyzed segment.

The numerator captures the absolute difference in amplitude between consecutive time steps, highlighting temporal variations in the acoustic signal. The denominator normalizes this sum by the total energy within the segment, ensuring that the ACI remains comparable across different recordings. A higher ACI value suggests greater variability in intensity over time, which is typically associated with biological sound sources, while lower values may indicate a dominance of continuous anthropogenic noise.

The Acoustic Diversity Index (ADI) measures the distribution of acoustic energy across different frequency bins within a spectrogram. The spectrogram is divided into discrete frequency bands, typically ten bins of 1000 Hz each, and the proportion of signals above a predefined threshold (−50 dBFS) is calculated for each bin. The resulting values are then analyzed using the Shannon index, a widely used metric for quantifying diversity in ecological studies. To prevent logarithmic errors, a small positive constant (ξ = 10⁻⁷) is added to the calculation [26,27]. The ADI is defined as follows:

$$ADI=-\sum _{i=1}^I{p}_i\mathrm{l}\mathrm{n}\left(p_i+\mathsf{\xi }\right)$$

where

• p_i: the proportion of acoustic signals exceeding a threshold (−50 dBFS) within the frequency band I;
• I: the total number of frequency bands (10 frequency bins, each 1000 Hz wide);
• ξ: a small positive constant (10⁻⁷) to prevent logarithmic errors when pi approaches zero.

Higher ADI values indicate a more evenly distributed soundscape, whereas lower values suggest dominance by a limited range of frequencies, which could be associated with anthropogenic noise.

The Normalized Difference Soundscape Index (NDSI) quantifies the relative contribution of antrophony (human-generated noise) and biophony (biologically produced sounds) within a given soundscape. This index is computed by taking the ratio of biophonic to antrophonic acoustic components, providing an estimate of the level of anthropogenic disturbance [25,27]. The NDSI is expressed as follows:

$$NDSI={\displaystyle \frac{(b-a)}{(b+a)}}$$

where

• b: the total acoustic energy attributed to biological sources (biophony);
• a: the total acoustic energy attributed to anthropogenic (human-generated) sources (anthropophony).

The NDSI value ranges from −1 to 1, with higher values indicating a greater proportion of biological sounds relative to anthropogenic noise, while lower values suggest a dominance of human-generated noise. In urban environments, the NDSI may be less reliable due to overlapping spectral characteristics of different noise sources, requiring supplementary methods for a more accurate interpretation of urban underwater soundscapes.

3. Results

3.1. Spectrogram Analysis

A spectrogram analysis was employed to visualize the temporal and spectral characteristics of the underwater acoustic environment. The spectrograms presented in Figure 5 illustrate the distribution of acoustic energy across different frequencies over time, highlighting key noise sources and their relative contributions to the soundscape. The vehicular noise is a dominant feature of the underwater soundscape. High-intensity noise signatures corresponding to cars and trams appear prominently, with broad spectral bands extending from 100 Hz to 5000 Hz. The most pronounced peaks occur between 500 Hz and 2000 Hz. Trams introduce distinct spectral bands between 500 Hz and 1500 Hz, marked by continuous tonal components indicative of wheel–rail interactions and electrical systems. Additionally, sirens and horns generate periodic high-energy bursts within the 1000–4000 Hz range, disrupting the underwater acoustic environment. Biophonic contributions are sparse in comparison. Bird vocalizations, detected at frequencies above 2000 Hz, appear intermittently but are often overshadowed by antrophonic sounds. Other biologically derived signals, such as faint human-related activities including coughing, are observed within the 300–1000 Hz range but do not constitute significant acoustic components. Temporal analysis indicates that noise intensity fluctuates throughout the recordings, with peak urban activity periods aligning with the most intense anthropogenic noise contributions. Persistent low-frequency background noise, below 300 Hz, is present across all the recordings and likely originates from structural vibrations and distant traffic sources. The prevalence of vehicular noise highlights the extent of urban acoustic intrusion into the aquatic environment, as continuous exposure to frequencies overlapping with biologically relevant communication ranges raises concerns about potential ecological disruptions.

To further quantify the characteristics of the underwater soundscape and evaluate the degree of antrophonic influence, acoustic indicators were analyzed. The Acoustic Complexity Index (ACI), Acoustic Diversity Index (ADI), and Normalized Difference Soundscape Index (NDSI) provide quantitative insights into the distribution and dominance of noise sources. The ACI, which measures the variability in sound intensity, reveals fluctuations that align with peaks in anthropogenic noise, particularly during high-traffic periods. The ADI, which assesses the spread of acoustic components across frequency bands, highlights increased diversity during instances of transient noise but does not indicate a strong biological presence. Meanwhile, the NDSI, designed to differentiate biophony from anthropophony, predominantly yields values indicative of overwhelming anthropogenic dominance, with minimal evidence of biologically driven acoustic events. These indices collectively reinforce the spectrogram analysis, demonstrating the substantial imprint of urban noise on the aquatic acoustic environment and its potential implications for ecosystem dynamics.

3.2. Acoustics Complexity Index (ACI)

The Acoustic Complexity Index (ACI) serves as a key metric in evaluating the temporal variability in the recorded soundscape. The ACI values for the analyzed recordings range between 114.86 and 280.27. This index lacks clearly defined value thresholds that would allow for unambiguous interpretation. However, it is designed to increase in the presence of biophony detected. To interpret the index values, a mapping process was conducted, involving the identification of segments with distinct value increases and verification of their content through auditory analysis of the source recordings.

The ACI values in Figure 6 fluctuate significantly across the recording period, ranging approximately between 100 and 240. This variability suggests dynamic acoustic conditions influenced by both anthropogenic and natural sources. The figure illustrates several notable peaks in the ACI values, correlating with specific events. One such peak, reaching 195.9, occurs when an object makes contact with the hydrophone, producing a brief but distinct increase in acoustic complexity. Other peaks, recorded at 216.2, 198.2, and 190.8, coincide with the presence of passing cars, reinforcing the observation that vehicular noise significantly contributes to underwater acoustic variability. While all the recordings exhibit fluctuations, recording 8 demonstrates a distinct pattern with more pronounced peaks, particularly around anthropogenic noise events. The other recordings show similar trends but with varied intensities, further confirming that urban noise intrusions dominate the underwater acoustic soundscape. A spectral analysis of ACI trends shows that fluctuations in complexity primarily occur within the 500–2000 Hz range, corresponding to the dominant frequencies of vehicular and tram noise. The presence of repetitive tonal structures, characteristic of engine noise and rail contact, enhances short-term amplitude variations, leading to increased ACI values. A comparison of the spectral analysis in relation to the obtained ACI trends shows that fluctuations in complexity primarily occur within the 500–2000 Hz range, corresponding to the dominant frequencies of vehicular and tram noise. The presence of repetitive tonal structures, characteristic of engine noise and rail contact, enhances short-term amplitude variations, leading to increased ACI values. This suggests that the primary drivers of acoustic complexity in urban aquatic environments are of anthropogenic origin rather than biological sources.

A further analysis of recordings 11–20 reinforces these observations (Figure 7). The ACI values remain highly variable, with notable peaks aligning with external interferences, tram activity, and vehicular noise. The highest ACI value recorded in this dataset, 244.8, corresponds to an external interference event, while another peak at 244.8 is associated with tram movement. These spikes demonstrate that sudden, high-energy sound sources continue to dominate the underwater soundscape. Additionally, noise from passing cars, with ACI values reaching 211.2 and 199.5, confirms that vehicular activity remains a consistent contributor to acoustic complexity.

Finally, the analysis of recordings 21–30 further confirms these patterns, showing strong variability in the ACI values across the dataset. Several notable peaks, reaching values of 234.0, 231.7, and 223.4, are directly linked to tram activity, emphasizing the continuous influence of public transportation on the aquatic soundscape (Figure 8). Additionally, two major peaks, recorded at 249.9 and 239.9, correspond to external interferences, suggesting occasional disturbances unrelated to consistent traffic noise. The dominance of trams in shaping the acoustic complexity remains evident, with their prolonged presence and broad spectral signatures contributing to sustained fluctuations in the ACI.

Comparison across all three datasets highlights a consistent trend: antrophonic sources dominate the acoustic environment, with trams and cars producing the most pronounced ACI fluctuations. The identification of external interferences in the latest dataset suggests additional noise sources beyond transportation, which may warrant further investigation.

3.3. Acoustic Diversity Index (ADI)

According to its original assumptions, the Acoustic Diversity Index (ADI) is used to determine the acoustic diversity of the analyzed segment. An increase in the index value indicates a greater presence of sounds across different frequency ranges. For analysis purposes, the recording segments were transformed from the time domain to the frequency domain and subsequently divided into ten equal intervals, each 4500 Hz wide. The interpretation of the ADI results followed the same methodology as the ACI analysis, utilizing value mapping onto specific recording fragments and a spectral sound analysis. The results indicate that increases in the ADI values are primarily associated with external disturbances, such as the microphone striking underwater objects and the effect of wind on the water surface. The first dataset, covering recordings 1–10, confirms these trends, with notable peaks in the ADI values linked to external interferences and sporadic environmental factors (Figure 9). Key spikes in the ADI values occur at 0.956, 1.051, and 1.482, all attributed to sudden external interactions with the hydrophone. A lower but distinct peak of 0.741 corresponds to water noise, indicating that non-anthropogenic disturbances also influence ADI fluctuations. Additionally, an isolated ADI peak at 1.190 is recorded when a biological entity makes contact with the hydrophone, further emphasizing that the ADI responds more sensitively to environmental interference rather than structured urban noise sources.

The second dataset, covering recordings 11–20, further supports these findings (Figure 10). A distinct peak at 1.234 is linked to the combined effect of vehicular noise and wind, illustrating how environmental and anthropogenic factors jointly influence the ADI. The most pronounced peak, reaching 1.831, occurs due to the interaction between car noise and an external interference event, demonstrating that momentary disruptions significantly increase acoustic diversity. These patterns reinforce the observation that the ADI is primarily influenced by irregular and transient disturbances rather than continuous structured noise, such as tram or vehicular traffic alone.

The final dataset, covering recordings 21–30, follows similar trends observed in the previous analyses (Figure 11). Peaks in the ADI values, such as 0.822 and 0.820, are linked to external interferences, while a distinct peak at 1.195 results from a combination of water noise and wind effects. Additionally, a notable peak at 1.183 occurs due to simultaneous car noise and wind, further reinforcing the role of environmental factors in influencing ADI fluctuations. The highest recorded ADI value, 1.444, is associated with water noise, suggesting that natural aquatic processes occasionally contribute significantly to acoustic diversity.

These results confirm that the ADI is primarily responsive to sporadic, transient disturbances rather than continuous urban noise sources. The influence of external interferences, wind, and water noise remains prominent across all datasets, highlighting the need for refined analytical methods to better distinguish between natural and anthropogenic contributors to acoustic diversity.

3.4. Normalized Difference Soundscape Index (NDSI)

The Normalized Difference Soundscape Index (NDSI) ranges from −1 to +1, where values close to −1 indicate a dominance of anthropophony, while values near +1 suggest its complete absence. The characteristic value mapping revealed that in the analyzed urban environment, the NDSI is not a reliable indicator. In many cases, sound samples with NDSI values oscillating around +1 contained clearly audible street noise, including passing cars and trams, which undermines the effectiveness of this index in assessing the underwater soundscape of urban areas.

An analysis of recordings 1–10 reveals inconsistencies in the expected relationship between the NDSI values and the actual acoustic environment (Figure 12). A peak at 0.860 corresponds to water noise combined with external interference, while another at 0.698 is attributed to a similar mix of environmental sounds. A lower NDSI peak at 0.444 still includes street noise from passing cars, contradicting the assumption that higher NDSI values indicate biophony. At the same time, negative values such as −0.916 and −0.956 align with strong anthropogenic signals, including vehicle noise and human voice, reinforcing the general trend that low NDSI values reliably indicate the presence of urban noise. However, the presence of street noise even in positive NDSI ranges challenges its applicability for distinguishing biophonic activity in urban aquatic environments.

An analysis of recordings 11–20 further supports this limitation (Figure 13). A peak at 0.793 occurs in the presence of both tram and car noise, demonstrating that NDSI values do not always align with expected biophonic dominance. Other peaks, such as 0.498 and 0.448, are linked to external interferences combined with car noise, reinforcing the observation that positive NDSI values do not necessarily indicate the absence of anthropogenic activity. Similarly, a peak at 0.326 corresponds to wind and siren noise, further questioning the index’s ability to differentiate natural from human-made sounds. Negative NDSI values, such as −0.977 and −0.934, align with strong anthropogenic sources, confirming that low NDSI values reliably indicate human noise dominance. However, the inconsistency in higher NDSI values suggests that the index struggles to accurately reflect urban underwater soundscapes.

An analysis of recordings 21–30 reinforces this pattern (Figure 14). A peak at 0.519 occurs in the presence of both motorcycle and water noise, demonstrating that moderate NDSI values can still be linked to mechanical disturbances. Another peak at 0.805 is attributed to wind, while values of 0.658 and 0.633 correspond to a mix of water noise and vehicular sounds. These cases highlight the ongoing issue of NDSI misclassification in urban environments, where environmental and anthropogenic sources overlap in frequency domains. Conversely, negative NDSI values such as −0.940 and −0.858 reliably indicate the presence of urban noise, with both linked to a combination of water noise and passing cars.

The results across all the datasets indicate that while negative NDSI values are reliable indicators of human-made sound dominance, positive values do not necessarily correlate with biophonic presence. This further suggests that the NDSI alone is insufficient for classifying urban underwater soundscapes and may require additional spectral analysis or refined computational approaches to improve accuracy.

4. Discussion

This study assessed the applicability of acoustic indicators in characterizing urban underwater soundscapes based on hydrophone recordings. The results indicate that while the ACI is useful for tracking fluctuations in anthropogenic noise, the ADI primarily reflects transient environmental interferences, and the NDSI fails to reliably differentiate between biophonic and antrophonic sources in urban aquatic environments. These findings raise important questions about the limitations of traditional acoustic indices in highly modified urban settings and underscore the need for methodological improvements.

The effectiveness of acoustic indicators in urban underwater environments remains a subject of ongoing debate [14,28,29]. The ACI successfully captured variations in noise levels associated with human activities, such as vehicular and tram traffic, aligning with findings from previous studies that have used the ACI to monitor urban noise pollution in terrestrial and aquatic environments [30,31]. However, its ability to detect biophonic activity remains questionable in heavily urbanized settings, as anthropogenic disturbances often produce similar variability in acoustic patterns. While the ADI was initially designed to capture the richness and diversity of soundscapes, our findings suggest that in urban aquatic environments, it is more sensitive to transient environmental interferences, such as wind and water turbulence. This aligns with research by Zhao et al. (2019) [32], who noted that the ADI could be influenced by short-duration non-biological sounds, limiting its effectiveness in distinguishing structured anthropogenic noise from natural acoustic elements. However, the same study highlighted that narrowband anthropogenic noise sources, such as road traffic, can adversely affect ACI performance, leading to misinterpretations of acoustic complexity in urban freshwater habitats. In our urban freshwater environment, traffic and tram noise clearly dominated, frequently creating narrowband spectral signatures that caused substantial fluctuations in the ACI values. This interference aligns with Zhao et al.’s [32] findings, confirming that the presence of persistent anthropogenic narrowband noise can complicate interpreting ACI results. These observations emphasize that urban freshwater soundscapes, characterized by continual narrowband interference from traffic and public transportation, possess unique acoustic profiles compared to marine or less urbanized freshwater ecosystems. Such distinctions must be considered when applying acoustic indices to assess urban aquatic soundscapes, underscoring the need for context-sensitive analysis methods or complementary acoustic metrics.

The NDSI, commonly employed to differentiate anthropophony from biophony, has shown inconsistencies in its application to urban underwater settings. High NDSI values in our study often corresponded to spectral gaps rather than actual biophonic presence. This corroborates findings by Gottesman (2019) [33], who reported similar misclassifications when applying the NDSI in urban environments, suggesting that the index may require significant adaptation for use in aquatic soundscape monitoring. The limitations observed in all three indices suggest that standard acoustic indicators may not be entirely suitable for monitoring urban underwater noise pollution without additional refinements. Studies by Darras et al. (2024) [34] and Minello et al. (2024) [35] highlight the need for a more integrative approach, combining multiple indices or employing advanced signal-processing techniques to enhance classification accuracy. Machine learning and deep learning approaches have also been proposed as potential solutions to improve automated noise differentiation, allowing for the more precise categorization of urban, biological, and environmental sound sources [20,36]. Recent research has proposed methodological enhancements to the NDSI, such as more adaptive frequency band divisions in combination with spectrum segmentation techniques. For instance, Gasc et al. [31] (2015) and Lindseth and Lobel [29] suggest dynamically adjusting frequency bands according to the specific acoustic features of the recorded environment. Such adaptive frequency segmentation can significantly reduce misclassification rates by aligning analysis windows with actual spectral distributions, better capturing unique urban sound signatures. Additionally, emerging approaches integrating machine learning (ML) classification techniques offer substantial promise for improving NDSI accuracy. ML-based methods typically involve supervised or semi-supervised algorithms trained on labeled audio data, distinguishing between biophony, geophony, and anthropophony by learning from distinctive spectral–temporal patterns rather than fixed frequency bands alone [21]. The findings highlight significant challenges in applying standard acoustic indices to urban underwater soundscapes, emphasizing the pervasiveness of anthropogenic noise across multiple frequency bands [37]. Unlike in natural environments where biophony and geophony are dominant, urban soundscapes are overwhelmingly shaped by human activities, which complicates traditional analytical frameworks. The need for tailored analytical methods in urban aquatic environments seems critical for obtaining meaningful ecological insights [38,39]. Long-term hydrophone monitoring is essential for identifying trends in urban underwater noise pollution. A more extensive temporal dataset would allow researchers to distinguish between persistent background noise and intermittent disturbances, such as peak traffic periods or seasonal variations. Additionally, integrating hydrophone data with other environmental monitoring tools, such as remote sensing and water quality assessments, could offer a more holistic understanding of how urbanization influences aquatic ecosystems. For instance, correlating underwater noise levels with biological responses from aquatic organisms, such as changes in fish behavior or communication patterns, would provide valid information on the ecological impacts of noise pollution.

Recent studies have evaluated the effectiveness of traditional acoustic indices, including the ACI, ADI, and NDSI, in monitoring biodiversity and ecosystem health. A comprehensive meta-analysis by Alcocer et al. [40] found that while these indices have a moderate positive correlation with biodiversity metrics, their performance can be inconsistent across different environments and taxa. This variability suggests that relying solely on these indices may not provide a complete picture of ecological conditions. To enhance the accuracy of the ADI, a frequency-dependent version (FADI) has been proposed. The FADI employs a new threshold scheme to reduce noise impacts, thereby providing a more reliable assessment of acoustic diversity, especially in environments with significant background noise. The traditional NDSI compares energy in specific frequency bands to estimate the balance between anthropogenic and biological sounds. However, its application has shown mixed results. For instance, a study by Bradfer-Lawrence et al. [41] demonstrated that the performance of the NDSI varied depending on factors such as location, season, and traffic levels. This finding indicates that the NDSI’s effectiveness as a standalone metric may be limited without considering contextual environmental factors. Given these insights, it is essential to interpret the results of the ACI, ADI, and NDSI within the specific context of our urban aquatic environment. The inherent variability and potential limitations of these indices underscore the importance of using them alongside other ecological assessments.

Given the limitations of traditional acoustic indices, alternative approaches should be considered. Advanced signal-processing techniques, including spectro-temporal analysis and machine learning classifiers, could improve noise source differentiation and provide more reliable assessments of underwater soundscapes. Recent advancements in artificial intelligence have demonstrated promising applications in bioacoustics, allowing for real-time detection and classification of noise sources, which could be highly beneficial for future urban soundscape research [42]. To enhance the reliability of acoustic monitoring in urban aquatic settings, future research should focus on developing hybrid indices that integrate multiple acoustic metrics, improving classification performance. Incorporating machine learning and deep learning models into acoustic data analysis will refine signal differentiation and automate the detection of noise sources with greater accuracy. Conducting controlled experiments will be essential to validate how acoustic indices respond to specific urban noise sources and determine their limitations in highly modified environments.

Moreover, recent developments in acoustic metamaterials, such as multilayer overlapping resonators and gradient Fabry–Perot structures, have demonstrated remarkable low-frequency broadband sound absorption capabilities (e.g., Wang et al. [43]). Though not tested within our current research scope, integrating these advanced materials into future studies and mitigation strategies could effectively reduce underwater noise pollution, ultimately benefiting aquatic ecosystems exposed to persistent urban acoustic disturbances. Additionally, investigating the ecological consequences of continuous anthropogenic noise exposure on aquatic organisms is crucial for understanding the long-term effects of urban noise pollution. By addressing these challenges, future studies can significantly improve the accuracy of urban underwater soundscape assessments and contribute to more effective environmental management and noise mitigation efforts.

5. Conclusions

This study highlights the challenges of using traditional acoustic indicators to assess urban underwater soundscapes, emphasizing the dominance of anthropogenic noise across multiple frequency bands. The analysis of the ACI, ADI, and NDSI demonstrates their respective limitations, with the ACI proving effective in tracking urban noise fluctuations, ADI responding primarily to environmental interferences, and NDSI failing to reliably differentiate between biophonic and antrophonic sources. These findings underscore the necessity of refining existing methodologies or developing new analytical frameworks tailored to the complexities of urban aquatic environments.

The findings from this study have practical applications for addressing urban noise pollution and protecting underwater habitats. This research demonstrates the inadequacy of established acoustic indices in distinguishing between human-created and natural sound sources and provides justification for improving the existing monitoring systems. Improved systems will allow for enhanced short-term noise monitoring to better inform policymakers on environmental noise for urban waterfronts and transit corridors. Moreover, these findings can be considered within the design of underwater infrastructure by city planners and engineers to maximize supported design elements currently used to limit sound transmission (i.e., buffer zones and sound-dampening material). Lastly, these findings are helpful to environmental agencies to develop more precise measures of ecological health so that conservation efforts could effectively mitigate adverse urban noise pollution.

Study Limitations

Despite its valuable insights, this study is limited by the constraints of the selected acoustic indices, which do not fully capture the complexity of urban underwater soundscapes. The broader implications of this research point to the need for long-term hydrophone monitoring and the integration of acoustic assessments with other environmental variables. The incorporation of machine learning and advanced signal-processing techniques offers a promising pathway to improving the classification and interpretation of urban underwater noise data. As urbanization continues to expand, understanding the effects of noise pollution on aquatic ecosystems is crucial for implementation.

Further Recommendation for Future Research

Finally, this study contributes to the growing body of research on urban soundscape ecology and underscores the urgent need for methodological advancements in underwater acoustic monitoring. Future studies in this area can target refinement of acoustic indices to increase their reliability in urban aquatic systems. We aim to broaden the scope of the study to include a number of urban water bodies that will allow us to determine the generalizability of our results. All of the future work mentioned above aims to expand the understanding of urban underwater soundscapes and contribute to improved noise management. Thus, future studies should build on these findings by developing refined acoustic indices and exploring innovative analytical approaches to enhance our understanding of urban underwater soundscapes and their ecological consequences, effective conservation strategies, and minimizing human impact on underwater habitats.