Abstract:

Background：Biodiversity loss driven by human activities and climate change has intensified the demand for efficient, scalable, and non-invasive monitoring approaches. Passive acoustic monitoring (PAM), based on the soundscape concept, captures biological, geophysical, and anthropogenic sounds over large spatial and temporal scales, providing a promising framework for biodiversity assessment. Because changes in species composition and habitat conditions are often reflected in soundscape structure, soundscape-based analysis has become a key component of ecoacoustics. 

Progress：Current soundscape-based biodiversity assessments mainly rely on two automated approaches: automatic recognition methods and acoustic index methods. Automatic recognition enables species- or sound-source-level identification using machine learning, but its performance is limited by the scarcity of annotated data, sensitivity to noise and sound overlap, and weak cross-regional generalization. Acoustic index methods offer computational efficiency by summarizing soundscape characteristics into numerical metrics, yet their ecological interpretability and robustness vary across habitats and environmental conditions, often leading to inconsistent or contradictory results. 

Prospects：Future progress requires standardized data acquisition and processing protocols, the development of open soundscape feature databases, and methodological integration of automatic recognition and acoustic index approaches. Combining these methods with multi-source environmental data is expected to enhance robustness, ecological interpretability, and comparability, supporting more reliable soundscape-based biodiversity monitoring and conservation decision-making.

Key words: soundscape assessment, biodiversity, automatic recognition, acoustic indices, passive acoustic monitoring