Relationship between acoustic indices, length of recordings and processing time: a methodological test

Abstract

Ecoacoustic approaches have the potential to provide rapid biodiversity assessments and avoid costly fieldwork. Their use in biodiversity studies for improving management and conservation of natural landscapes has grown considerably in recent years. Standardised methods for sampling acoustic information that deliver reliable and consistent results within and between ecosystems are still lacking. Sampling frequency and duration are particularly important considerations because shorter, intermittent recordings mean recorder batteries last longer and data processing is less computationally intensive, but a smaller proportion of the available soundscape is sampled. Here, we compare acoustic indices and processing time for subsamples of increasing duration clipped from 94 one-hour recordings, to test how different acoustic indices behave, in order to identify the minimum sample length required. Our results suggest that short recordings distributed across the survey period accurately represent acoustic patterns, while optimizing data collection and processing. ACI and H are the most stable indices, showing an ideal sampling schedule of ten 1-minute samples in an hour. Although ADI, AEI and NDSI well represent acoustic patterns under the same sampling schedule, these are more robust under continuous recording formats. Such targeted subsampling could greatly reduce data storage and computational power requirements in large-scale and long-term projects.

Linking acoustic diversity to compositional and configurational heterogeneity in mosaic landscapes

“…Our study follows previous recommendations in Passive Acoustic Monitoring and soundscape assessment methodology that rotating recorders across the landscape is the most cost-efficient design for the best trade-off in sound data acquisition at larger spatial and temporal scales (Sugai et al 2020;Drake et al 2021), while keeping the volume of data storage under a reasonable threshold (Cifuentes et al 2021;Wood et al 2021). While in tropical forest biomes, the minimal recording time period required to stabilize the variance in acoustic indices across time for a given site is ca 120hr (Bradfer-Lawrence et al 2019), in temperate and semi-arid biomes where intra-day variation is often higher than inter-day variation due to higher seasonality in acoustic activity (Gasc et al 2018), continuous recording across 24-48hr is generally accurate if the relevant season is targeted for surveys (Metcalf et al 2021).…”

“…A discontinuous recording schedule was set to record 30 minute per hour (30 min on / 30 min off) during a continuous time period of several 24-hr diel periods in a row for each site (Burivalova et al 2018). We used relatively short recording periods compared to the ones conducted in tropical biomes, to allow sampling the acoustic diversity of multiple sites by rotating the recorders across the landscape (Sugai et al 2020;Cifuentes et al 2021), while still being within the peak seasonal period of breeding bird vocal activity in both study regions.…”

Pond Acoustic Sampling Scheme: A draft protocol for rapid acoustic data collection in small waterbodies

“…To represent potential variation across the waterbody, each 10-min sample should be divided into ten 1-min subsamples recorded in different mesohabitats around the edge of the pond (Figure 1). The 1-min recording length has become common practice for ecoacoustic research, used in many studies (e.g., Bayne et al, 2017;Campos-Cerqueira et al, 2020;Eldridge et al, 2018;Farina et al, 2011;Farina & Gage, 2017;Fuller et al, 2015;Gottesman et al, 2018;Pieretti et al, 2015;Wimmer et al, 2013), and has benefits over longer recording periods in terms of acoustic index accuracy, and computational requirements (Cifuentes et al, 2021).…”

Biotic sound SNR influence analysis on acoustic indices

“…Despite ambitious global targets to reduce biodiversity loss (Tittensor et al, 2014), pressure on biodiversity has increased notably (Butchart et al, 2010;Dröge et al, 2021) over the past four decades. Quantifying biological diversity is fundamental for setting priorities for conservation (Mittermeier et al, 1998;Brooks et al, 2006), especially in the current period of dramatic biodiversity loss (Brooks et al, 2006;Ceballos et al, 2015;Cifuentes et al, 2021). Traditionally, this has relied on detailed species inventories, which however often require an extensive, costly sampling effort, especially in high-biodiversity areas (Cifuentes et al, 2021).…”