Passive acoustic monitoring has been widely employed to study the occurrence and behavior of vocalizing animals in terrestrial and aquatic habitats. To study the entire community of vocalizing animals, researchers have developed some tools or several acoustics indices to investigate the composition and changes in soundscapes. Here we used spectrograms, long-term spectrograms, several acoustic indices, and species identification to fulfill the acoustic-based biodiversity assessment.

A spectrogram is a visual representation of the spectrum and shows frequencies of a signal varying with time. You can see various sound sources with different shapes or trajectories on the spectrogram, and some bio sounds do have frequency-modulated characteristics. Besides, we used a long-term spectrogram (LTS) to facilitate the analysis for monthly scale recordings, and it clearly shows the periodicity such as diurnal, nocturnal, or daily cycles. If you want to know more about LTS, here are some references :

Scientific paper: Computing biodiversity change via a soundscape monitoring network
Exploring acoustic diversity via soundscape_IR: Python script for making LTS

Acoustic indices have been widely applied in many studies because they take little time and lower cost when processing. Besides, different acoustic indices are designed for evaluating different aspects of the composition of sound sources in soundscapes. Here, we calculated 8 acoustic indices in R by using seewave and soundecology packages :

Shannon entropy (SH), Temporal entropy (TH), and Acoustic Entropy Index (H) represent the spectral, temporal, or total entropy of a time wave. Acoustic Complexity Index (ACI) means the absolute difference between two adjacent values of intensity in a single frequency bin. The acoustic Diversity Index (ADI) and Acoustic Evenness Index (AEI) represent the Shannon index and Gini index applied to 1kHz bins. Bioacoustic Index (BI) represents the total energy of bio sounds. Normalized Difference Soundscape Index (NDSI) is the ratio of human-generated (anthrophony) to biological (biophony) acoustic components.

Acoustic Entropy (H) from Sueur et al. 2008
Acoustic Complexity Index (ACI) from Pieretti et al. 2011
Acoustic Diversity Index (ADI) & Acoustic Evenness Index (AEI) from Villanueva-Rivera et al. 2011
Bioacoustic Index (BI) from Boelman et al. 2007
Normalized Difference Soundscape Index (NDSI) from Kasten et al. 2012

Species identification provides information about the occurrence of species in a recording. We use Sound Identification and Labeling Intelligence for Creatures (SILIC) and BirdNET two machine learning algorithms to detect the avian sounds in soundscapes. Although SILIC only included the database of species in Taiwan, it has the ability to identify frog chorus or even bat echolocation as well. If you are interested in species identification, the following links can bring you to know more about these two powerful tools

Scientific paper:
SILIC: A cross database framework for automatically extracting robust biodiversity information from soundscape recordings based on object detection and a tiny training dataset
BirdNET: A deep learning solution for avian diversity monitoring

Introductions and instructions:
SILIC
BirdNET

To facilitate the sharing and reuse of audio data while safeguarding personal privacy, we take steps to mask any unintended voices captured in recordings. We use WebRTC detect human voice, with the sensitivity currently set to level 3 and a detection threshold of 0.9.

If you appreciated our work and want to know more about Soundscape Monitoring Network, welcome to visit About Us.