Multi-Index Ecoacoustics Analysis for Terrestrial Soundscapes: A New Semi-Automated Approach Using Time-Series Motif Discovery and Random Forest Classification

High rates of biodiversity loss caused by human-induced changes in the environment require new methods for large scale fauna monitoring and data analysis. While ecoacoustic is increasingly being used shows promise, analysis interpretation big produced remains a challenge. Computer-generated acoustic indices potentially provide biologically meaningful summary sound, however, temporal autocorrelation, difficulties statistical multi-index lack consistency or transferability different terrestrial environments have hindered application those contexts. To address these issues we investigate use time-series motif discovery random forest classification multi-indices through two case studies. We semi-automated workflow combining (acoustic complexity, entropy, events per second) to categorize sounds unfiltered recordings according main source sound present (birds, insects, geophony). Our approach showed more than 70% accuracy label assignment both datasets. The categories assigned were broad, but believe this great improvement on traditional single index environmental as can now give ecological meaning way that does not expert knowledge manual validation only necessary small subset data. Furthermore, which largely ignored researchers, has been effectively eliminated technique applied here first time expect our will greatly assist researchers future it allow datasets be rapidly processed labeled, enabling screening undesired sounds, such wind, target biophony (insects birds) bioacoustics research.