Acoustic event detection for spatiotemporal patterns of bird activity in protected areas

The spatiotemporal variation in bird vocalization activity reflects systematic patterns related to circadian rhythms and human activity intensity. Understanding these patterns is crucial for interpreting bird diversity in protected areas increasingly impacted by human activities. Passive acoustic monitoring (PAM) enables long-term, noninvasive tracking of these patterns, yet the sheer volume of complex data poses significant analytical challenges. Acoustic event detection (AED) models provide a critical solution by extracting acoustic events from continuous recordings. Using PAM data collected from August to December 2023 in the Huanglong Protected Area, Sichuan, China, this study applied locally trained bird species recognition models and AED models to quantify spatiotemporal variation in bird acoustic activity across multiple temporal scales and disturbance intensities. Results identified key species within the protected area and revealed distinct daily and monthly activity patterns, with bird activity peaking in mid-morning, remaining low at night, and declining over successive months. Bird activity also decreased along gradients of increasing human disturbance. These patterns were further supported by false-color spectrograms with consistent time-frequency structures. Overall, this study demonstrates the value of AED-based PAM for characterizing bird community dynamics and supporting biodiversity monitoring and tourism management in protected areas.