Efficacy of extracting indices from large‐scale acoustic recordings to monitor biodiversity

Abstract

Passive acoustic monitoring could be a powerful way to assess biodiversity across large spatial and temporal scales. However, extracting meaningful information from recordings can be prohibitively time consuming. Acoustic indices (i.e., a mathematical summary of acoustic energy) offer a relatively rapid method for processing acoustic data and are increasingly used to characterize biological communities. We examined the relationship between acoustic indices and the diversity and abundance of biological sounds in recordings. We reviewed the acoustic-index literature and found that over 60 indices have been applied to a range of objectives with varying success. We used 36 of the most indicative indices to develop a predictive model of the diversity of animal sounds in recordings. Acoustic data were collected at 43 sites in temperate terrestrial and tropical marine habitats across the continental United States. For terrestrial recordings, random-forest models with a suite of acoustic indices as covariates predicted Shannon diversity, richness, and total number of biological sounds with high accuracy (R ≥ 0.94, mean squared error [MSE] ≤170.2). Among the indices assessed, roughness, acoustic activity, and acoustic richness contributed most to the predictive ability of models. Performance of index models was negatively affected by insect, weather, and anthropogenic sounds. For marine recordings, random-forest models poorly predicted Shannon diversity, richness, and total number of biological sounds (R ≤ 0.40, MSE ≥ 195). Our results suggest that using a combination of relevant acoustic indices in a flexible model can accurately predict the diversity of biological sounds in temperate terrestrial acoustic recordings. Thus, acoustic approaches could be an important contribution to biodiversity monitoring in some habitats.

Spatio‐temporal heterogeneity in river sounds: Disentangling micro‐ and macro‐variation in a chain of waterholes

“…In this study, we employed three acoustic indices: the acoustic complexity index ( ACI ), the spectral entropy ( H f ), and the median of amplitude envelope ( M ). These three indices were chosen because they measure different aspects of the soundscape, they have been demonstrated to efficiently represent soundscapes and have been used before in freshwater environments (Buxton, McKenna, et al, ; Desjonquères et al, ; Linke & Deretic, ; Towsey et al, ). All three indices were calculated on the whole spectrum in R using the seewave package (Sueur et al, ).…”

Spatio‐temporal scaling of biodiversity in acoustic tropical bird communities

“…The first one relies on characterizing the soundscape for each time and frequency unit by summary indices (Buxton et al 2018). Today, two approaches dominate the scene.…”

“…The first one relies on characterizing the soundscape for each time and frequency unit by summary indices (Buxton et al 2018). Of the two, the first one -soundscape analysis using indices -has been proposed as the most suitable tool to monitor the general state of habitats (Burivalova et al 2018, Buxton et al 2018, and indeed, acoustic indices have been shown to be closely related to the diversity and abundance of biological sounds across local recordings (Buxton et al 2018, Darras et al 2018). Here, the conversion of the sound data can be achieved by experts, semi-automated algorithms or machine learning techniques such as deep learning (Hill et al 2018, Ovaskainen et al 2018, Stowell et al 2018.…”

It's time to listen: there is much to be learned from the sounds of tropical ecosystems

“…, Rankin & Axel ), yet there is no consensus to date as to which are most effective, primarily due to the difficulties in generalizing across taxa and ecosystems (Buxton et al . ). Existing indices can also be sensitive to geophony such as rain, wind, and river flow, or can be skewed by certain acoustically dominant species (Staaterman et al .…”

“…Still, there are limitations to automatic approaches because they can be initially time-consuming and they require training data to create different classifiers for different species as well as programming or signal processing expertise to develop automated species identification models. At the soundscape level, many acoustic indices and soundscape analysis methods have been proposed and used for the assessment of biodiversity (e.g., Sueur et al 2008, Gasc et al 2013, Fuller et al 2015, Vega et al 2016, Rankin & Axel 2017), yet there is no consensus to date as to which are most effective, primarily due to the difficulties in generalizing across taxa and ecosystems (Buxton et al 2018). Existing indices can also be sensitive to geophony such as rain, wind, and river flow, or can be skewed by certain acoustically dominant species (Staaterman et al 2017, Linke et al 1999.…”