Clustering, categorizing, and mapping of shallow coastal water soundscapes

Clea Parcerisas, Dick Botteldooren, Paul Devos, Elisabeth Debusschere2022 · DOI

Abstract

For many inhabitants of shallow coastal waters, the underwater soundscape is a crucial source of information for survival, especially where visibility is poor. However, these environments are increasingly saturated with anthropogenic noise. Passive Acoustic Monitoring (PAM) offers a cost-effective, non-invasive way to monitor these habitats, yet standardized, automated methods for soundscape characterization are lacking. This study proposes a methodology for clustering underwater soundscapes and linking them to environmental parameters using explainable artificial intelligence (XAI). Applied to the Belgian Part of the North Sea (BPNS), the model identifies background sound categories and maps their spatio-temporal distribution, providing a holistic view of the marine acoustic environment.

Introduction

In marine ecosystems, sound serves as the primary medium for communication and navigation. Shallow coastal waters are particularly complex acoustic environments, influenced by both biological activity and heavy anthropogenic pressure, primarily from shipping. Traditional acoustic analysis often focuses on specific events or species; however, "soundscape analysis" seeks to understand the entire acoustic scene. The challenge lies in processing vast amounts of data into meaningful, interpretable categories that can be mapped across space and time.

Methodology

The researchers developed an automated pipeline consisting of four main stages:

  1. Feature Extraction: Acoustic data was converted into "hybrid millidecade bands," a frequency-dependent format optimized for long-term spectral comparison.
  2. Unsupervised Clustering: The team employed UMAP (Uniform Manifold Approximation and Projection) for dimensionality reduction and DBSCAN (Density-Based Spatial Clustering of Applications with Noise) to identify natural groupings in the acoustic data without prior labels.
  3. Predictive Modeling: A Random Forest (RF) model was trained to predict these acoustic categories using 12 environmental variables, including bathymetry, seabed habitat, and vessel density.
  4. Explainable AI (XAI): SHAP (SHapley Additive exPlanations) values were used to interpret the RF model, identifying which environmental factors most strongly influenced specific soundscape classifications.

Data Integration

The study utilized acoustic data from the LifeWatch Broadband Acoustic Network, covering 14 deployments across 6 stations in the BPNS. This was integrated with environmental data from EMODnet (bathymetry, vessel density, seabed maps) and ERDDAP (salinity, temperature, wave height).

Vessel density map of the Belgian Part of the North Sea, showing the distribution of shipping activity used as a primary environmental driver in the classification model.

Key Insights and Results

The methodology successfully identified four distinct soundscape categories with a high classification accuracy of 91.68%.

  • Environmental Drivers: The analysis revealed that seabed habitat was the most significant variable in explaining acoustic differences between locations, suggesting the physical environment defines the baseline "background" sound.
  • The Shipping Signature: One specific category (Category 0) was strongly associated with high shipping density and high salinity. SHAP analysis confirmed that these areas exhibited significantly higher low-frequency noise (63 Hz and 125 Hz), consistent with large vessel signatures.
  • Spatio-Temporal Mapping: By applying the predictive model, the researchers generated maps showing how soundscape categories shift across the BPNS over different months, demonstrating the model's utility for large-scale monitoring.

Conclusion and Future Work

This study demonstrates a robust, automated approach to marine soundscape monitoring. While the current 1-minute temporal resolution is effective for identifying ambient trends, it may overlook short-term biological events. Future research will focus on increasing temporal resolution to capture specific "bio-sounds" and addressing data gaps regarding smaller vessels not tracked by AIS. This holistic mapping approach provides a vital tool for assessing ecosystem health and the impact of anthropogenic noise in sensitive coastal regions.

Summary updated

Abstract

For many of its inhabitants, the underwater soundscape is a rich source of information that may be crucial for their survival. Moreover, in shallow coastal waters where visibility is poor, the importance of sound is emphasized. Yet coastal waters are also rich in anthropogenic sounds which may disturb the ecosystem. Passive Acoustics Monitoring (PAM) is a flexible, non-invasive, and cost-effective solution to acquire information at habitat or community level. Studying the acoustic scene of a habitat in a global, holistic way is known as soundscape analysis. However, there are currently no standardized methods to characterize and understand marine soundscapes in an automated way. Here we propose a methodology for clustering underwater soundscapes and linking the obtained categories to environmental parameters in space and time. This is done using explainable artificial intelligence. The methodology is applied to a PAM dataset collected in the Belgian Part of the North Sea. The obtained categories focus on background sound, which includes all combinations of sounds that occur under certain conditions at specific places. With this information, the marine acoustic scene and its change over space and time can be mapped for the whole area of interest.
BibTeX
@misc{parcerisas2022clustering,
  author = {Clea Parcerisas and Dick Botteldooren and Paul Devos and Elisabeth Debusschere},
  title = {{Clustering, categorizing, and mapping of shallow coastal water soundscapes}},
  year = {2022},
  doi = {10.61782/fa.2023.1070},
}
Figure