Graphical Representation of Landscape Heterogeneity Identification through Unsupervised Acoustic Analysis

María J. Guerrero, Camilo Sánchez‐Giraldo, César A. Uribe, Victor D. Martínez, Claudia Isaza

Abstract

Changes in land use and climate change threaten global biodiversity and ecosystems, calling for the urgent development of effective conservation strategies. Recognizing landscape heterogeneity, which refers to the variation in natural features within an area, is crucial for these strategies. While remote sensing images quantify landscape heterogeneity, they might fail to detect ecological patterns in moderately disturbed areas, particularly at minor spatial scales. This is partly because satellite imagery may not effectively capture undergrowth conditions due to its resolution constraints. In contrast, soundscape analysis, which studies environmental acoustic signals, emerges as a novel tool for understanding ecological patterns, providing reliable information on habitat conditions and landscape heterogeneity in complex environments across diverse scales and serving as a complement to remote sensing methods. We propose an unsupervised approach using passive acoustic monitoring data and network inference methods to analyze acoustic heterogeneity patterns based on biophony composition. This method uses sonotypes, unique acoustic entities characterized by their specific time-frequency spaces, to establish the acoustic structure of a site through sonotype occurrences, focusing on general biophony rather than specific species and providing information on the acoustic footprint of a site. From a sonotype composition matrix, we use the Graphical Lasso method, a sparse Gaussian graphical model, to identify acoustic similarities across sites, map ecological complexity relationships through the nodes (sites) and edges (similarities), and transform acoustic data into a graphical representation of ecological interactions and landscape acoustic diversity. We implemented the proposed method across 17 sites within an oil palm plantation in Santander, Colombia. The resulting inferred graphs visualize the acoustic similarities among sites, reflecting the biophony achieved by characterizing the landscape through its acoustic structures. Correlating our findings with ecological metrics like the Bray-Curtis dissimilarity index and satellite imagery indices reveals significant insights into landscape heterogeneity. This unsupervised approach offers a new perspective on understanding ecological and biological interactions and advances soundscape analysis. The soundscape decomposition into sonotypes underscores the method’s advantage, offering the possibility to associate sonotypes with species and identify their contribution to the similarity proposed by the graph.

Introduction

To protect biodiversity, conservationists must first understand landscape heterogeneity—the variation in natural features across a given area. Traditionally, researchers have relied on satellite imagery (remote sensing) for this purpose. However, satellites face significant limitations in moderately disturbed areas or under dense forest canopies, where vegetation may appear uniform from above despite vast ecological differences on the ground.

Furthermore, traditional acoustic monitoring often requires experts to manually label thousands of hours of audio to identify specific species. This process is slow, expensive, and limits the scope of study to known species. This paper introduces an unsupervised framework to analyze the entire "soundscape" automatically, revealing underlying ecological structures without the need for pre-labeled data.

Methodology

The proposed pipeline utilizes Passive Acoustic Monitoring (PAM) and network inference to map landscape diversity through four primary stages:

  1. Sonotype Identification: Audio signals are clustered into "sonotypes"—unique acoustic entities defined by their time-frequency signatures—using unsupervised learning.
  2. Acoustic Fingerprinting: An "acoustic structure" matrix is generated for each site based on the frequency of sonotype occurrences.
  3. Network Inference via Graphical Lasso (GLasso): A sparse Gaussian graphical model estimates relationships between geographical sites. Unlike simple correlation, GLasso identifies "partial correlations," filtering out noise to find the most significant direct similarities.
  4. Total Variation Distance (TVD): This metric identifies which specific sonotypes drive the connections (or lack thereof) between sites in the resulting network.
The proposed unsupervised pipeline for mapping landscape heterogeneity using acoustic data, from acquisition to graphical model interpretation.

Datasets

The methodology was validated using a simulated dataset of 11 sites and 10 sonotypes to establish ground truth. It was then applied to the Puerto Wilches Dataset, comprising 19,598 audio recordings from 17 sites in an oil palm plantation in Santander, Colombia, covering oil palm, secondary vegetation, and forest fragments.

Results and Discussion

The study demonstrates that biophony (biological sound) serves as a high-resolution proxy for ecological complexity. The Graphical Lasso approach achieved 98% accuracy in identifying site relationships in simulated tests, significantly outperforming the standard Bray-Curtis dissimilarity index (78%).

Comparison of acoustic structures (fingerprints) between two different sites, illustrating how sonotype distributions define a site's unique acoustic footprint.

Key findings include:

  • Complementary Data: The correlation between the acoustic network and satellite-derived spectral indices was remarkably low (0.11). This indicates that acoustic analysis provides unique ecological insights that remote sensing cannot capture.
  • Acoustic Barriers: The model revealed that some forest fragments were acoustically isolated, suggesting that surrounding oil palm plantations act as physical barriers to animal movement—a nuance missed by satellite imagery.
  • Temporal Dynamics: Network density was found to peak during the day, with different species groups driving connectivity at different hours (dawn, day, dusk, night).

Conclusion

This unsupervised approach offers a powerful new perspective for soundscape analysis and conservation. While current implementation requires empirical threshold selection for the GLasso model, future work aims to develop threshold-free models. By linking automated sonotypes back to specific biological species, this method could eventually combine the efficiency of machine learning with the depth of traditional taxonomy to provide a comprehensive view of ecosystem health.

Summary updated

Abstract

Abstract Changes in land use and climate change threaten global biodiversity and ecosystems, calling for the urgent development of effective conservation strategies. Recognizing landscape heterogeneity, which refers to the variation in natural features within an area, is crucial for these strategies. While remote sensing images quantify landscape heterogeneity, they might fail to detect ecological patterns in moderately disturbed areas, particularly at minor spatial scales. This is partly because satellite imagery may not effectively capture undergrowth conditions due to its resolution constraints. In contrast, soundscape analysis, which studies environmental acoustic signals, emerges as a novel tool for understanding ecological patterns, providing reliable information on habitat conditions and landscape heterogeneity in complex environments across diverse scales and serving as a complement to remote sensing methods. We propose an unsupervised approach using passive acoustic monitoring data and network inference methods to analyze acoustic heterogeneity patterns based on biophony composition. This method uses sonotypes, unique acoustic entities characterized by their specific time-frequency spaces, to establish the acoustic structure of a site through sonotype occurrences, focusing on general biophony rather than specific species and providing information on the acoustic footprint of a site. From a sonotype composition matrix, we use the Graphical Lasso method, a sparse Gaussian graphical model, to identify acoustic similarities across sites, map ecological complexity relationships through the nodes (sites) and edges (similarities), and transform acoustic data into a graphical representation of ecological interactions and landscape acoustic diversity. We implemented the proposed method across 17 sites within an oil palm plantation in Santander, Colombia. The resulting inferred graphs visualize the acoustic similarities among sites, reflecting the biophony achieved by characterizing the landscape through its acoustic structures. Correlating our findings with ecological metrics like the Bray-Curtis dissimilarity index and satellite imagery indices reveals significant insights into landscape heterogeneity. This unsupervised approach offers a new perspective on understanding ecological and biological interactions and advances soundscape analysis. The soundscape decomposition into sonotypes underscores the method’s advantage, offering the possibility to associate sonotypes with species and identify their contribution to the similarity proposed by the graph.
BibTeX
@inproceedings{guerrero2025graphical,
  author = {María J. Guerrero and Camilo Sánchez‐Giraldo and César A. Uribe and Victor D. Martínez and Claudia Isaza},
  title = {{Graphical Representation of Landscape Heterogeneity Identification through Unsupervised Acoustic Analysis}},
  booktitle = {Methods},
  year = {2025},
  doi = {10.1111/2041-210x.70041},
}
Figure