Integrative Bioacoustics for Species Monitoring

Abstract

Passive acoustic monitoring (PAM) has emerged as a transformative approach to biodiversity surveillance, enabling continuous, non-invasive recording of soundscapes from which species presence, abundance, and community composition can be inferred across taxonomic groups spanning birds, bats, anurans, insects, marine mammals, and fish. This study integrated three complementary bioacoustic analytical frameworks -- acoustic index analysis, automated species detection via deep learning, and soundscape ecology -- to assess their combined utility for multi-taxon biodiversity monitoring across 84 sites spanning five biome types. A network of 312 AudioMoth v1.2 recorders deployed for continuous 12-month recording periods generated 2.84 million hours of audio data. Deep learning classifiers (BirdNET for birds, AVES for general fauna, custom LSTM for bats) achieved mean species identification accuracy of 91.4% +- 3.2% (birds), 84.7% +- 5.4% (bats), and 78.4% +- 6.8% (anurans) in blind validation tests against expert-verified recordings. Acoustic diversity index (ADI), acoustic complexity index (ACI), and bioacoustic index (BI) collectively explained 74.3% of variance in ground-truth species richness across sites (multiple regression R2 = 0.743). Soundscape composition analysis via non-metric multidimensional scaling (NMDS) of acoustic spectral fingerprints correctly classified biome type with 87.4% accuracy, confirming the potential of acoustic signatures as ecosystem-level biodiversity indicators. Cost analysis demonstrated that PAM-based biodiversity monitoring costs USD 12 +- 4 per site-month compared with USD 184 +- 47 per site-month for equivalent manual point-count surveys, a 93.5% cost reduction. These results establish integrative bioacoustics as a scalable, cost-effective foundation for national biodiversity monitoring frameworks aligned with GBF Target 21.