Assessing Taiwan's Coral Reef Ecosystem: Detecting Long-term Snapping Shrimp Sounds

Snap sounds, which constitute primary components of underwater soundscapes, serve as crucial indicators of ecosystem health. However, reliable detection of these acoustic signatures presents significant challenges, as methods relying solely on amplitude thresholds often result in overestimation of snap rates. This study presents a time-series data processing approach that integrates amplitude thresholds with kurtosis measurements to improve the accuracy of snap sound detection. Analysis of recordings from Taiwan's marine protected areas revealed predominantly low kurtosis values, highlighting the analytical challenges inherent in complex coral reef ecosystems. Through comparative analysis with manually verified snap sounds, kurtosis thresholds provide stricter and more accurate identification of snap sounds. Additionally, compared to smoothed kernel methods, kurtosis thresholds more effectively exclude non-snap signals. This improved detection approach enhances the robustness of quantifying snap rate variations, thereby improving ecosystem assessments and supporting more effective conservation strategies.