Acoustic indices as rapid indicators of avian diversity in different land-use types in an Indian biodiversity hotspot

Abstract

Passive acoustic monitoring is a potentially valuable tool in biodiversity hotspots, where surveying can occur at large scales across land conversion types. However, in order to extract meaningful biological information from resulting enormous acoustic datasets, rapid analytical techniques are required. Here we tested the ability of a suite of acoustic indices to predict avian bioacoustic activity in recordings collected from the Western Ghats, a biodiversity hotspot in southwestern India. Recordings were collected at 28 sites in a range of land-use types, from tea, coffee, and cardamom plantations to remnant forest stands. Using 36 acoustic indices we developed random forest models to predict the richness, diversity, and total number of avian vocalizations observed in recordings. We found limited evidence that acoustic indices predict the richness and total number of avian species vocalizations in recordings (R2 < 0.51). However, acoustic indices predicted the diversity of avian species vocalizations with high accuracy (R2 = 0.64, mean squared error = 0.17). Index models predicted low and high diversity best, with the highest residuals for medium diversity values and when continuous biological sounds were present (e.g., insect sounds >8 sec). The acoustic complexity index and roughness index were the most important for predicting avian vocal diversity. Avian species richness was generally higher among shade-grown crops than in the open tea plantation. Our results suggest that models incorporating acoustic indices can accurately predict low and high avian species diversity from acoustic recordings. Thus, ecoacoustics could be an important contributor to biodiversity monitoring across landscapes like the Western Ghats, which are a complex mosaic of different land-use types and face continued changes in the future.

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

“…They yield a relatively simple single measure of different aspects of a soundscape (amplitude, spectral complexity, or spectral variability). The use of acoustic indices is still relatively new, there is therefore a strong need to study their efficacy in various environments and establish evidence‐based best practices (Bradfer‐Lawrence et al, ; Buxton, Agnihotri, Robin, Goel, & Balakrishnan, ). Each index describes different attributes of the soundscape; therefore, all three indices did not reveal exactly identical results (Dema et al, ).…”

Acoustic community structure and seasonal turnover in tropical South Asian birds

“…Thus, a combination of local movements, changes in singing activity and the arrival of winter migrants appears to drive species turnover in the avian acoustic community. This turnover, and the frequent detection of common winter migrants, suggests that acoustic monitoring may provide a valuable tool to detect and monitor the annual influx of winter migrants, as well as to elucidate patterns of local movement in resident species (10,48–50). These local movements remain poorly understood for many bird species, including relativelycommon ones.…”

Passive acoustic monitoring as a potential tool to survey animal and ecosystem processes in freshwater environments

“…A diversity of species produces sounds. This acoustic diversity can be used to extract information on species richness (Buxton, Agnihotri, Robin, Goel, & Balakrishnan, ; Pieretti et al., ; Pijanowski, Farina, Gage, Dumyahn, & Krause, ; Sueur, Pavoine, et al., ); however, the relationships between acoustic and species diversity are not always straightforward (Linke, Decker, et al., ). Soundscape composition can be quantified using acoustic indices.…”

“…Although the accuracy of acoustic indices for measuring species diversity has been demonstrated in some environments (Buxton et al., ; Pieretti et al., ; Sueur, Pavoine, et al., ), they appear to need adjustment in freshwater ponds due to the low signal‐to‐noise ratios (Desjonquères et al., ), as with marine environments (Parks et al., ). To overcome this issue, audio filtering techniques could potentially help reduce noise or enhance signals of interest.…”

“…Once acoustic recordings have been analysed, an important remaining step is to relate extracted acoustic attributes to environmental data. This requires statistical analyses such as linear models (Fuller et al., ), random forest (Buxton et al., ) or neural networks (Gómez et al., ). Acoustic datasets can be challenging to analyse as they are usually time series, i.e.…”