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

Abstract

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

Classification and ecological relevance of soundscapes in urban informal settlements

“…To address the challenge of efficiently synthesizing large volumes of audio data, unsupervised and supervised machine learning techniques have been increasingly used to categorize sound recordings for a variety of purposes ranging from classifying single sound source recordings (Grinfeder et al, 2022; Yang & Kang, 2013) to classifying acoustic ‘scenes’ or the environmental context of sound recordings (Barchiesi et al, 2015; Duque‐Montoya & Isaza, 2018; Elise et al, 2019; Waldekar & Saha, 2020), as well as whole soundscapes. Yet there remains a need to show these classification procedures are efficient and transferable—that is, that specific methods and classifications are not highly context‐specific (Carrico et al, 2020; Scarpelli et al, 2021).…”

Using acoustic indices in ecology: Guidance on study design, analyses and interpretation

“…Each index only reflects one aspect of the soundscape, so researchers can consider several indices in combination to reflect multiple facets at once (Bradfer-Lawrence et al, 2019;Eldridge et al, 2018;Scarpelli et al, 2021;Sueur et al, 2014;Towsey, Zhang, et al, 2014).…”

“…Researchers should consider whether they can create subsets of data or simulated datasets that will enable power analyses (Bradfer-Lawrence et al, 2019;Wood et al, 2021). Scarpelli et al, 2021;Towsey et al, 2018;Znidersic et al, 2020) but temporal or spatial non-independence can also be problematic with these approaches (Colegrave & Ruxton, 2018;Forstmeier et al, 2017).…”

“…Hill et al, 2018), facilitating the collection of vast audio data sets (Roe et al, 2021). While audio recordings can be readily collected, the analysis methods needed to garner robust ecological insights from acoustic data are still in development (Deichmann et al, 2018; Scarpelli et al, 2021; Vella et al, 2022; Wimmer et al, 2013).…”

Evaluating the socio-environmental impacts of monoculture cashew expansion in the Northern Western Ghats, India

“…Acoustics is rapidly gaining popularity, however using only acoustic indices could lead to missed inferences due to the high vocalization diversity in biodiverse landscapes, since no index currently captures the overall information contained within soundscapes (Buxton et al 2018a). Using acoustic indices for monitoring biodiversity faces its share of criticism such as spatial autocorrelation, no consensus on which indices are best suited for specific taxa, and the lack of compatibility across different sampling habitats (Scarpelli et al, 2021). Previous research has shown that using a combination of We used ADI, AEI, NDSI, BI, and H because these indices have been shown to be related to metrics such as bird presence, species richness and abundances in studies in forest-production landscapes (Mammides et al, 2017).…”