Exploring fish community dynamics in photic and mesophotic coral reefs using passive acoustic monitoring

Mesophotic coral ecosystems (MCEs) have garnered growing interest in recent years, as they may offer deep refuges for species threatened by severe climatic impacts in shallow-water reefs. However, whether MCEs truly serve this function remains debated, largely because visual surveys in deeper waters provide only brief snapshots for fish communities. This study utilized passive acoustic monitoring (PAM) to investigate soniferous fish and their responses to different environmental conditions. Two autonomous sound recorders were deployed at 20 and 40 m depths around Lyudao, Taiwan. Using machine learning techniques, we detected a substantial amount of fish sounds and discovered pronounced differences in sound composition between the two sites. Over 2 years of monitoring, fish sound diversity at the 40-m site remained stable, while the 20-m site showed declines during cold seasons and following typhoon events. Although a limited database is currently unavailable for local species identification, family level identification shows that acoustic data well reflect taxonomic divergence across habitats. Future application of PAM in coral reefs will generate high-resolution, long-term data on fish community dynamics and enable deeper insights into the refuge potential of MCEs and their resilience to climatic impacts.