Acoustic indices for biodiversity assessments: Analyses of bias based on simulated bird assemblages and recommendations for field surveys

a r t i c l e i n f o Rapid acoustic surveys aim to estimate biodiversity based on the diversity of sounds produced by animal communities , and interest in this approach among conservation planners is increasing. Several indices of acoustic diversity have been proposed as proxies for species richness. However, in the field, the animal activity may be responsible for acoustic diversity to varying degrees. To evaluate how measures of acoustic diversity may depart from actual species richness, we selected seven acoustic indices and applied them to simulated recordings of bird assemblages under different field conditions. For a given sound, defined here as the song composition emitted by a species assemblage, we determined if the indices were i) only driven by animal sounds, ii) insensitive to species identity, and iii) independent of species evenness. Under the field conditions evaluated, none of the indices fulfilled the three criteria necessary for a perfect proxy of species richness. However, some indices may be appropriate as a measure of biodiversity under a more broad definition including phylogenetic and/or functional aspects of diversity. We provide recommendations for the application of these indices for biodiversity measurement under field conditions, such as the application of appropriate audio filters, the increase of the repetition rate of the recordings, and the identification of the main taxonomic groups occurring in the recorded communities. Exhaustively describing all living organisms, from bacteria to large animals and plants, seems to be an impossible target despite the launch of all taxa biodiversity inventories (ATBIs) (Granjou et al. Although an exact measure of biodiversity is currently out of reach, it is still possible to generate estimates from a subset of organisms. Traditionally, species inventories are created through sampling and then extrapolating the data over larger areas and time scales (Sutherland, 2006), but an accurate extrapolation requires a sampling protocol with a large spatial and/or temporal component , which requires lengthy and difficult fieldwork. In addition, the sampling and identification of different taxa depend on a high degree of expert taxonomic knowledge; even a drastic increase in the number of taxonomists might not meet this demand for expertise (Costello et al., 2013). For example, a recent study reported an average time of 21 years between the collection of a specimen in the field and a formal species description and name (Fontaine et al., 2012). This time lag is especially prevalent for under-sampled areas, …