How well do acoustic indices measure biodiversity? Computational experiments to determine effect of sound unit shape, vocalization intensity, and frequency of vocalization occurrence on performance of acoustic indices

Abstract

Abstract Passive acoustic monitoring of biotic sounds is increasingly popular in conservation biology. Among vocalizing animals, birds are most frequently studied and multiple acoustic indices have been proposed for rapid acoustic assessment of avian diversity. Preliminary results suggest that several indices can be used as proxies for bird species richness, however it is still unclear to what extent different conditions of bird vocalizations affect the relationship between indices and bird species richness – should it matter if the bird vocalizations contain different sound unit shapes, or if the frequency of vocalization occurrence differs, or if the vocalizations are made with various intensities? In this work, seven commonly used acoustic indices were tested using three controlled computational experiments with real-world recordings to provide an objective measure of each index’s performance for answering the aforementioned questions. In the experiments, different options of sound unit shape and frequency of vocalization occurrence were precisely controlled and intensity variations were expressed as different signal-to-noise ratios (SNR) of the vocalizations. The first experiment showed that three indices (the acoustic entropy index (H), acoustic diversity index (ADI), and acoustic evenness index (AEI)) performed better than the other four, showing moderate correlations with avian species richness. The second experiment revealed that ADI for each sound unit shape tended towards a constant value with increasing frequency of vocalization occurrence while the influence from frequency of vocalization occurrence on H and AEI varied with different sound unit shape options. Our third experiment showed that the vocalization intensity affected the values of these three indices while the performance disparity among different sound unit shapes for only ADI explicitly appeared a decreasing tendency with increasing vocalization intensities. We conclude that ADI, among the tested indices, is relatively more robust with regard to bird species richness surveys when sound unit shape, frequency of vocalization occurrence, and vocalization intensity are considered. Meanwhile, since multiple indices are usually applied together to provide a comprehensive observation, the above acoustic dimensions should be taken into account especially in comparative research of different bird communities.