Acoustic identification of Mexican bats based on taxonomic and ecological constraints on call design

Abstract

Summary Monitoring global biodiversity is critical for understanding responses to anthropogenic change, but biodiversity monitoring is often biased away from tropical, megadiverse areas that are experiencing more rapid environmental change. Acoustic surveys are increasingly used to monitor biodiversity change, especially for bats as they are important indicator species and most use sound to detect, localise and classify objects. However, using bat acoustic surveys for monitoring poses several challenges, particularly in megadiverse regions. Many species lack reference recordings, some species have high call similarity or differ in call detectability, and quantitative classification tools, such as machine learning algorithms, have rarely been applied to data from these areas. Here, we collate a reference call library for bat species that occur in a megadiverse country, Mexico. We use 4685 search‐phase calls from 1378 individual sequences of 59 bat species to create automatic species identification tools generated by machine learning algorithms (Random Forest). We evaluate the improvement in species‐level classification rates gained by using hierarchical classifications, reflecting either taxonomic or ecological constraints (guilds) on call design, and examine how classification rate accuracy changes at different hierarchical levels (family, genus and guild). Species‐level classification of calls had a mean accuracy of 66%, and the use of hierarchies improved mean species‐level classification accuracy by up to 6% (species within families 72%, species within genera 71·2% and species within guilds 69·1%). Classification accuracy to family, genus and guild‐level was 91·7%, 77·8% and 82·5%, respectively. The bioacoustic identification tools we have developed are accurate for rapid biodiversity assessments in a megadiverse region and can also be used effectively to classify species at broader taxonomic or ecological levels. This flexibility increases their usefulness when there are incomplete species reference recordings and also offers the opportunity to characterise and track changes in bat community structure. Our results show that bat bioacoustic surveys in megadiverse countries have more potential than previously thought to monitor biodiversity changes and can be used to direct further developments of bioacoustic monitoring programs in Mexico.

Testing the performances of automated identification of bat echolocation calls: A request for prudence

“…Recently, different automated approaches, typically employing multivariate sets of spectral and temporal variables of bat calls, have been attempted with variable results (e.g. Parsons and Jones, 2000;Walters et al, 2012;Zamora-Gutierrez et al, 2016). Freeware and commercial software used to speed up the screening of long recordings, select echolocation calls and identify species have recently appeared and are used extensively.…”

Weather conditions determine attenuation and speed of sound: Environmental limitations for monitoring and analyzing bat echolocation

“…Many bats use echolocation, an active sensory system (Nelson & MacIver, 2006), as main remote sense for perceiving their environment. Researchers take advantage of this continuous stream of echolocation calls to acoustically monitor bat presence/absence, activity and behavior and, in certain cases, to identify species (Andreassen, Surlykke, & Hallam, 2014;Walters et al, 2012;Zamora-Gutierrez et al, 2016). Beyond academic research, these data are used for risk assessment (e.g., at wind turbines; e.g., Newson et al, 2017), to infer population density, diversity, and vulnerability of bats (Clement, Rodhouse, Ormsbee, Szewczak, & Nichols, 2014;Meyer et al, 2011), and to inform risk mitigation and conservation strategies (Meyer, 2015), often based on automatic call analysis software (Russo & Voigt, 2016;Rydell, Nyman, Eklöf, Jones, & Russo, 2017).…”

Emerging opportunities and challenges for passive acoustics in ecological assessment and monitoring

“…Perhaps the most fundamental knowledge gap for PAM is the limited availability of comprehensive, expert‐verified species call databases for reference and training data. Much remains unknown about the intra‐ and interspecific call diversity of even well‐studied taxa (Kershenbaum et al., ), and ground‐truthed call databases are difficult and laborious to assemble, requiring the collection of high‐quality audio recordings of animals identified to species either visually or through capture (e.g., Zamora‐Gutierrez et al., ). Where such verified datasets exist they are biased towards vertebrates (particularly cetaceans, bats, and birds), with especially scarce resources for anurans and invertebrates (Lehmann, Frommolt, Lehmann, & Riede, ; Penone et al., ) and regions outside Europe and North America, despite the urgent need for tools to facilitate monitoring of subtropical and tropical habitats (Zamora‐Gutierrez et al., ).…”

“…Much remains unknown about the intra‐ and interspecific call diversity of even well‐studied taxa (Kershenbaum et al., ), and ground‐truthed call databases are difficult and laborious to assemble, requiring the collection of high‐quality audio recordings of animals identified to species either visually or through capture (e.g., Zamora‐Gutierrez et al., ). Where such verified datasets exist they are biased towards vertebrates (particularly cetaceans, bats, and birds), with especially scarce resources for anurans and invertebrates (Lehmann, Frommolt, Lehmann, & Riede, ; Penone et al., ) and regions outside Europe and North America, despite the urgent need for tools to facilitate monitoring of subtropical and tropical habitats (Zamora‐Gutierrez et al., ). These gaps translate into equivalent biases in classifier availability, and to our knowledge no widely available tools exist for distinguishing intraspecific acoustic behaviours (e.g., social from echolocation calls in cetaceans and bats) (Figure e), although machine learning methods have successfully been applied to analysis of bat acoustic social behaviour (Prat et al., ).…”