Sounding out ecoacoustic metrics: Avian species richness is predicted by acoustic indices in temperate but not tropical habitats

Abstract

Affordable, autonomous recording devices facilitate large scale acoustic monitoring and Rapid Acoustic Survey is emerging as a cost-effective approach to ecological monitoring; the success of the approach rests on the development of computational methods by which biodiversity metrics can be automatically derived from remotely collected audio data. Dozens of indices have been proposed to date, but systematic validation against classical, in situ diversity measures are lacking. This study conducted the most comprehensive comparative evaluation to date of the relationship between avian species diversity and a suite of acoustic indices. Acoustic surveys were carried out across habitat gradients in temperate and tropical biomes. Baseline avian species richness and subjective multi-taxa biophonic density estimates were established through aural counting by expert ornithologists. 26 acoustic indices were calculated and compared to observed variations in species diversity. Five acoustic diversity indices (Bioacoustic Index, Acoustic Diversity Index, Acoustic Evenness Index, Acoustic Entropy, and the Normalised Difference Sound Index) were assessed as well as three simple acoustic descriptors (Root-mean-square, Spectral centroid and Zero-crossing rate). Highly significant correlations, of up to 65%, between acoustic indices and avian species richness were observed across temperate habitats, supporting the use of automated acoustic indices in biodiversity monitoring where a single vocal taxon dominates. Significant, weaker correlations were observed in neotropical habitats which host multiple non-avian vocalizing species. Multivariate classification analyses demonstrated that each habitat has a very distinct soundscape and that AIs track observed differences in habitat-dependent community composition. Multivariate analyses of the relative predictive power of AIs show that compound indices are more powerful predictors of avian species richness than any single index and simple descriptors are significant contributors to avian diversity prediction in multi-taxa tropical environments. Our results support the use of community level acoustic indices as a proxy for species richness and point to the potential for tracking subtler habitat-dependent changes in community composition. Recommendations for the design of compound indices for multi-taxa community composition appraisal are put forward, with consideration for the requirements of next generation, low power remote monitoring networks.

Impact of stand variables on characteristics of avian soundscape in common oak (Quercus robur L.) forests

“…(Machado et al, 2017). The results of studies conducted at sites of varying degrees of degradation (from forests to agricultural land) in southeastern England and northwestern Ecuador showed that the greatest correlation (up to 65%) between acoustic indices and the number of bird species was observed in a temperate climate (Eldridge et al, 2018). However, the results of a number of studies failed to indicate a significant correlation of acoustic indices with the number of species.…”

“…The integration of several acoustic parameters allowed to characterize and identify habitats with varying degrees of disturbance (mature forest, secondary growth and pasture) in Colombia (Gόmez et al, 2018) accurately predicted the diversity of birds in India (Buxton et al, 2018), etc. A multivariate analysis of the relative predictive power of acoustic indices shows that composite indices are more powerful predictors of bird species richness than any particular index (Eldridge et al, 2018).…”

Combining machine learning and a universal acoustic feature-set yields efficient automated monitoring of ecosystems

“…Each of the above features was computed over 1 s of audio and concatenated to create a five-dimensional feature vector. This is referred to as a compound index in standard eco-acoustic studies 25 .…”

“…entropy of the audio waveform 24 ) rather than a machine-learned, general acoustic fingerprint. Again, these indices can predict key ecological indicators in local contexts [25][26][27] , but they often fail to discriminate even large ecological gradients 28,29 , and behave unpredictably when transferred to new environments 30 .…”

Ecoacoustics in the rain: understanding acoustic indices under the most common geophonic source in tropical rainforests

“…In recent years, acoustic indices have been widely used in ecological studies in highly pluvious tropical environments (e.g. Eldridge et al 2018;G omez et al 2018;Jorge et al 2018;Bradfer-Lawrence et al 2019). Although rainfall occurs seasonally in these environments, it can also vary unpredictably at small spatial and temporal scales (Poveda et al 2005;Rapp and Silma 2012).…”