Classification of Echolocation Calls from 14 Species of Bat by Support Vector Machines and Ensembles of Neural Networks

Redgwell, Robert D.; Szewczak, Joseph M.; Jones, Gareth; Parsons, Stuart

doi:10.3390/a2030907

Cited by 56 publications

(49 citation statements)

References 32 publications

(35 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Bat call sequences were identified to species using a hierarchical decision engine trained on up to 72 time‐frequency and time‐amplitude parameters extracted from a library of >10 000 species‐known recordings (Parsons & Szewczak ; Redgwell et al . ) implemented in sonobat version 3.1 (Szewczak ), followed by manual vetting and confirmation of species identifications using known‐call characteristics (Szewczak et al . ).…”

Section: Methodsmentioning

confidence: 99%

Accounting for false‐positive acoustic detections of bats using occupancy models

Clement

Rodhouse

Ormsbee

et al. 2014

Journal of Applied Ecology

View full text Add to dashboard Cite

Summary1. Acoustic surveys have become a common survey method for bats and other vocal taxa. Previous work shows that bat echolocation may be misidentified, but common analytic methods, such as occupancy models, assume that misidentifications do not occur. Unless rare, such misidentifications could lead to incorrect inferences with significant management implications. 2. We fit a false-positive occupancy model to data from paired bat detector and mist-net surveys to estimate probability of presence when survey data may include false positives. We compared estimated occupancy and detection rates to those obtained from a standard occupancy model. We also derived a formula to estimate the probability that bats were present at a site given its detection history. As an example, we analysed survey data for little brown bats Myotis lucifugus from 135 sites in Washington and Oregon, USA. 3. We estimated that at an unoccupied site, acoustic surveys had a 14% chance per night of producing spurious M. lucifugus detections. Estimated detection rates were higher and occupancy rates were lower under the false-positive model, relative to a standard occupancy model. Un-modelled false positives also affected inferences about occupancy at individual sites. For example, probability of occupancy at individual sites with acoustic detections but no captures ranged from 2% to 100% under the false-positive occupancy model, but was always 100% under a standard occupancy model. 4. Synthesis and applications. Our results suggest that false positives sufficient to affect inferences may be common in acoustic surveys for bats. We demonstrate an approach that can estimate occupancy, regardless of the false-positive rate, when acoustic surveys are paired with capture surveys. Applications of this approach include monitoring the spread of WhiteNose Syndrome, estimating the impact of climate change and informing conservation listing decisions. We calculate a site-specific probability of occupancy, conditional on survey results, which could inform local permitting decisions, such as for wind energy projects. More generally, the magnitude of false positives suggests that false-positive occupancy models can improve accuracy in research and monitoring of bats and provide wildlife managers with more reliable information.

show abstract

Section: Methodsmentioning

confidence: 99%

Accounting for false‐positive acoustic detections of bats using occupancy models

Clement

Rodhouse

Ormsbee

et al. 2014

Journal of Applied Ecology

View full text Add to dashboard Cite

show abstract

“…A number of objective and quantitative methods have been used to identify bats acoustically, including discriminant function analysis (Zingg ; Vaughan, Jones & Harris ; Parsons & Jones ), support vector machines (Redgwell et al . ), artificial neural networks (ANN) (Parsons & Jones ; Redgwell et al . ) and synergetic pattern recognition (Obrist, Boesch & Flückiger ).…”

Section: Introductionmentioning

confidence: 99%

A continental‐scale tool for acoustic identification of European bats

Walters

Freeman

Collen

et al. 2012

Journal of Applied Ecology

Self Cite

163

133

View full text Add to dashboard Cite

Summary1. Acoustic methods are used increasingly to survey and monitor bat populations. However, the use of acoustic methods at continental scales can be hampered by the lack of standardized and objective methods to identify all species recorded. This makes comparable continent-wide monitoring difficult, impeding progress towards developing biodiversity indicators, transboundary conservation programmes and monitoring species distribution changes. 2. Here we developed a continental-scale classifier for acoustic identification of bats, which can be used throughout Europe to ensure objective, consistent and comparable species identifications. We selected 1350 full-spectrum reference calls from a set of 15 858 calls of 34 European species, from EchoBank, a global echolocation call library. We assessed 24 call parameters to evaluate how well they distinguish between species and used the 12 most useful to train a hierarchy of ensembles of artificial neural networks to distinguish the echolocation calls of these bat species. 3. Calls are first classified to one of five call-type groups, with a median accuracy of 97Á6%. The median species-level classification accuracy is 83Á7%, providing robust classification for most European species, and an estimate of classification error for each species. 4. These classifiers were packaged into an online tool, iBatsID, which is freely available, enabling anyone to classify European calls in an objective and consistent way, allowing standardized acoustic identification across the continent. 5. Synthesis and applications. iBatsID is the first freely available and easily accessible continental-scale bat call classifier, providing the basis for standardized, continental acoustic bat monitoring in Europe. This method can provide key information to managers and conservation planners on distribution changes and changes in bat species activity through time.

show abstract

“…In addition, visual comparisons of call profiles within our dataset with previous descriptions and reference calls from Makira found apparent correspondence with two additional species: M. oceanensis and Mosia nigrescens. However, the quantitative approach to the identification of echolocation calls offers consistent and repeatable classification of unknown calls (Redgwell et al 2009) and therefore we only note the information from visual inspection here for interest given our focus on a data-poor area.…”

Section: Data Analysesmentioning

confidence: 99%

Ultrasonic monitoring to assess the impacts of forest conversion on Solomon Island bats

Davies

Růžička

Lavery

et al. 2016

Remote Sens Ecol Conserv

View full text Add to dashboard Cite

Paleotropical islands are experiencing extensive land-use change, yet little is known about how such changes are impacting wildlife in these biodiversity hotspots. To address this knowledge gap, we characterized bat responses to forest conversion in a biodiverse, human-threatened coastal rainforest habitat on Makira, Solomon Islands. We analysed~200 h of acoustic recordings from echolocating bats in the four dominant types of land use on Makira: intact forest, secondary forest, food gardens and cacao plantations. Bat calls were identified to the species level using a supervised classification model (where labelled data are used to train the system). We examined relative activity levels and morphological traits across habitats. Relative activity levels were highest in intermediately disturbed habitats and lowest in the most heavily disturbed habitat, although these differences were not significant. There were significant differences in the mean forearm length of bat assemblages across habitats, with the highest mean forearm length found in the most open habitat (Cacao). Overall, our study constitutes the first detailed exploration of anthropogenic effects on mammalian diversity in the Solomon Islands and includes the first acoustic and morphological information for many bat species in Melanesia. We use our experience to discuss the challenges of acoustic monitoring in such a remote and poorly studied region.

show abstract

Classification of Echolocation Calls from 14 Species of Bat by Support Vector Machines and Ensembles of Neural Networks

Cited by 56 publications

References 32 publications

Accounting for false‐positive acoustic detections of bats using occupancy models

Accounting for false‐positive acoustic detections of bats using occupancy models

A continental‐scale tool for acoustic identification of European bats

Ultrasonic monitoring to assess the impacts of forest conversion on Solomon Island bats

Contact Info

Product

Resources

About