Abstract:The analysis of audio recordings through acoustic indices has been proposed as an efficient way to measure and monitor biodiversity, given the assumption that higher levels of biodiversity produce more rapidly-changing and complex sound. However, in previous work in south China, we have found only moderate correlations between the acoustic indices and bird species richness, when analyzing recordings that were made at the same time as conducting point counts of birds. Here, we extended that work in three study regions in Guangxi Province, making observations both inside relatively undisturbed forest reserves and in the surrounding agricultural lands. We found that of 42 correlations between bird species richness and the acoustic indices (seven commonly used acoustic indices [ACI, ADI, AEI, AR, BIO, H, NDSI] each calculated in six habitat/region combinations) only seven had even a moderate relationship (i.e., |r| ≥ 0.20). To understand this paucity of relationships, we listened to a subsample of the recordings, scoring them for various kinds of biophony (incidence of bird and insect sounds), anthropophony (incidence of observer-produced sounds, sounds made by other humans, and sounds made by machines) and geophony (primarily wind, occasionally water). These analyses found that insects were positively related to the acoustic indices, especially in forests. Moreover, anthropophony and geophony, were, in general, negatively correlated to the indices. The incidence of bird vocalizations also had consistent and strong correlations to the residuals of the models relating the acoustic indices to bird species richness, demonstrating that the indices are clearly sensitive to the amount of bird sounds, but that is not necessarily strongly correlated to bird diversity. Multivariate models confirmed that the amounts of different sound categories were more influential on the acoustic indices than bird species richness. Our results demonstrate how competing noises, in both relatively undisturbed environments and heavily modified ones, can confound relationships between the acoustic indices and biodiversity. We discuss possible remedial steps for using acoustic indices in noisy soundscapes, including tailoring the selection of indices and frequency bands for the objectives of the project, or, alternatively, using a big-data approach that combines multiple indices.

Correlation Clustering of Bird Sounds

Introducing the Software CASE (Cluster and Analyze Sound Events) by Comparing Different Clustering Methods and Audio Transformation Techniques Using Animal Vocalizations

Improving Bird Classification with Unsupervised Sound Separation

Birds Sound Classification Based on Machine Learning Algorithms

Machine learning and statistical classification of birdsong link vocal acoustic features with phylogeny

Cross-modal correlation learning for clustering on image-audio dataset.

Active Learning for Bird Sounds Classification

Bird Sound Classification Based on ECOC-SVM

Machine Learning-based Classification of Birds through Birdsong

Acoustic bird species classification under low SNR and small-scale dataset conditions

Deep embedded clustering of coral reef bioacoustics

Bird Sounds Classification by Large Scale Acoustic Features and Extreme Learning Machine.

Song-based Classification techniques for Endangered Bird Conservation

Investigation of Different CNN-Based Models for Improved Bird Sound Classification

The Incidence of Bird Sounds, and Other Categories of Non-Focal Sounds, Confound the Relationships Between Acoustic Indices and Bird Species Richness in Southern China

Weakly-Supervised Classification and Detection of Bird Sounds in the Wild. A BirdCLEF 2021 Solution

Sliding-window Based Scale-Frequency Map for Bird Sound Classification Using 2D- and 3D-CNN

Acoustic component detection for automatic species recognition in environmental monitoring

SSL-Net: A Synergistic Spectral and Learning-based Network for Efficient Bird Sound Classification

Estimating the Repertoire Size in Birds using Unsupervised Clustering techniques

Classification of animal sounds in a hyperdiverse rainforest using Convolutional Neural Networks