Biodiversity assessment using passive acoustic recordings from off-reef location—Unsupervised learning to classify fish vocalization.

This article quantitatively characterizes the off-reef acoustic environment near Grande Island in the Zuari estuary, India, during the pre-monsoon period using passive acoustic recordings. It identifies prominent fish choruses from four fish groups—Sciaenidae (drums and croakers), Terapon theraps (tiger perch), planktivorous fish, and an unidentified "type A" fish—alongside snapping shrimp sounds, analyzing their temporal and spectral call parameters through oscillogram segmentation. Biodiversity was assessed using three acoustic indices: Acoustic Evenness Index (AEI), Acoustic Complexity Index (ACI), and root mean square sound pressure level (SPLrms) across full-band (50–20,050 Hz), low-frequency fish band (100–2,000 Hz), and high-frequency shrimp band (2,000–20,000 Hz). The study applied a hybrid unsupervised machine learning approach combining principal component analysis (PCA) and K-means clustering to classify fish sounds, achieving an overall classification accuracy of approximately 90%, demonstrating the method's potential for real-time monitoring of fish populations and biodiversity in marine ecosystems.