Impact of acoustic index parameters on soundscape comparisons

Abstract

Unprecedented ecosystem changes have prompted prioritisation of scalable and informative environmental monitoring. Ecoacoustics derives information from passive acoustic monitoring of soundscapes as a proxy to describe ecological processes. Ecoacoustic indices are often used to characterise specific aspects of the acoustic environment, for example the Acoustic Diversity Index potentially reflecting local biophony. Increasing application of these indices raises concerns about the ecological relationship between the soundscape and the index values. The need to quantify their variance against validated ecosystem traits is becoming more appreciated. For several commonly used indices, the impacts of the parameter settings and sampling regimes on output values have not been fully explored, despite the potential to bias results. To address this, we conducted three sensitivity analyses on acoustic index parameters. The first explores raw output from the Acoustic Complexity Index at various parameter settings, on a single recording. The second models the output of four common indices (Bioacoustics Index and Acoustic Complexity, Diversity and Evenness Indexes) using Bayesian mixed effects models on two datasets and explores how comparative differences change with parameter settings. The third explores the effect of the pitch of simulated animal calls on the output of the four indices. The analyses demonstrate that parameter selection alone can influence the magnitude and direction of difference between acoustic index outputs, in turn affecting comparative interpretation. These effects are also dependent on the soundscape data (e.g. sampling frequency, target frequency bands). We provide insights into drivers of these effects and recommendations to better understand the implications of index variance for within- and between-study comparisons. Recommendations include identifying frequency bands relevant to the research question, selecting parameters appropriately and using sensitivity analyses to determine if parameter selection is influencing interpretation of results. Further work is required to ecologically validate acoustic indices and their relationships to soundscapes. Standardised methods are needed to identify the scale to which differences in the ecosystem can be resolved and compared.