Acoustic identification of bats in the eastern United States: A comparison of parametric and nonparametric methods

Abstract

Ultrasonic detectors are widely used to survey bats in ecological studies. To evaluate efficacy of acoustic identification, we compiled a library of search phase calls from across the eastern United States using the Anabat system. The call library included 1,846 call sequences of 12 species recorded from 14 states. We determined accuracy rates using 3 parametric and 4 nonparametric classification functions for acoustic identification. The 2 most flexible classification functions also were the most accurate: neural networks (overall classification accuracy = 0.94) and mixture discriminant analysis incorporating an adaptive regression model (overall classification accuracy = 0.93). Flexible nonparametric methods offer substantial benefits when discriminating among closely related species and may preclude the need to group species with similar calls. We demonstrate that quantitative methods provide an effective technique to acoustically identify bats in the eastern United States with known accuracy rates. © 2011 The Wildlife Society.

Comparing supervised learning methods for classifying sex, age, context and individual Mudi dogs from barking

“…Gaussian mixture models have also been used for individual animal recognition in birds (Cheng et al 2010). Bat species have been acoustically identified using artificial neural networks (Parsons 2001;Britzke et al 2011), discriminant function analysis (Parsons and Jones 2000; Britzke et al 2011), classification trees (Adams et al 2010), k-nearest neighbors (Britzke et al 2011) as well as other classifiers (random forests and support vector machines) whose behavior has been compared (Armitage and Ober 2010). Artificial neural networks have been used to discriminate between the sounds of different animals within a group of British insect species (Orthoptera), including crickets and grasshoppers (Chesmore 2001).…”

The Problem of Low Agreement among Automated Identification Programs for Acoustical Surveys of Bats

“…Several efforts have explored approaches to optimize bat species identification (Parsons and Jones 2000, Skowronski and Harris 2006, Britzke et al 2011). There are also warnings that identifications may prove problematic because bat calls are not as distinctive as bird calls (Barclay 1999).…”

“…However, we do not address this method here; rather, we concentrate on assessing the accuracy of the identification of call sequences by looking at the agreement among software packages. (Skowronski and Harris 2006, Skowronski and Fenton 2008, Britzke et al 2011, we restricted our analysis to these widely available packages.…”

Insectivorous Bats and Silviculture: Balancing Timber Production and Bat Conservation

“…Second, the response to clearcut harvests between Myotis species varies both within and among species (Patriquin and Barclay 2003), with some increase in activity associated with linear edge habitats at the periphery of cuts but reduced activity in the centre of harvested stands, except where residual patches are left behind (Hogberg et al 2002). As our ability to distinguish among Myotis species increases with technological advances in acoustic detectors and software packages (Britzke et al 2011), resolution among the full suite of Myotis bats in North America should become possible allowing for a more in-depth and complete evaluation of bat response to edge effects in actively managed forests.…”

“…Forests with reduced tree density and vegetative clutter permit higher levels of light penetration, with this increased exposure hypothesised to enhance the suitability of live and dead trees for roosting by bark-and cavity-roosting bats in temperate climates (Boyles and Aubrey 2006). Further, LiDAR studies demonstrate that reduced clutter in the mid-and understory layers of forests is correlated with higher levels of activity by low-frequency (≤34 kHz) open-space bats (Britzke et al 2011;Dodd et al 2013). However, closed-space bat species that glean insects from vegetation and manoeuvre well within clutter benefit from a relatively dense understorey and higher tree densities, which can act as sources of insect prey (FuentesMontemayor et al 2013).…”