Tools and Technology

Data Acquisition Varies by Bat Phonic Groupfor 2 Types of Bat Detectors WhenWeatherproofed and Paired in Field Settings

ZACHARY D. E. KAISER, Center for Bat Research, Outreach, and Conservation, Department of Biology, Indiana State University, Terre Haute,IN 47809, USA

JOY M. O’KEEFE,1 Center for Bat Research, Outreach, and Conservation, Department of Biology, Indiana State University, Terre Haute, IN 47809,USA

ABSTRACT Manufacturers of acoustic bat detectors use proprietary microphones with different frequencyresponses, sensitivities, and directionality. Researchers implement various waterproofing strategies to protectmicrophones from inclement weather. These factors cause different detector models to have unique samplingareas and likely results in dissimilar data recording. We tested whether SM2BATþ (SM2) and Anabat SD2(Anabat) bat detectors record dissimilar data when weatherproofed in a manner suitable for long-term passivesampling. We deployed detectors at 71 random points near Indianapolis, Indiana (USA), from May toAugust, 2012–2013. We used 458 polyvinyl chloride tubes to weatherproof directional Anabat microphonesand the stock foam shielding to cover omnidirectional SM2 microphones. Anabat and SM2 microphoneswere paired at 2-m and 5-m heights. We adjusted file parameters to make Anabat and SM2 data comparable.We identified files to phonic group (low, midrange, and Myotis) using Bat Call ID software. The effects ofdetector type, phonic group, height, and their interactions on mean files recorded per site were assessed usinggeneralized estimating equations and least-significant-differences pairwise comparisons. Anabats recordedmore low and midrange files, but fewerMyotis files per site than did SM2s.When comparing the same modelof detectors, deployment height did not affect data acquisition. Weatherproofing may limit the ability ofAnabats to recordMyotis, but Anabat microphones may have greater detection ranges for low and midrangebats. We demonstrated that the ability to record bat calls in different frequency ranges varies withmicrophone type and weatherproofing strategy, which implies that best practices for presence–absencesurveys may also vary for bats in different phonic groups. � 2015 The Wildlife Society.

KEYWORDS acoustic monitoring, Anabat SD2, bats, BCID, echolocation, generalized estimating equations, phonicgroup, SM2BATþ, ultrasonic microphone.

Insectivorous bats of eastern North America are nocturnal,volant, and exceptionally agile in cluttered landscapes;consequently, these small mammals are extremely difficultto study. Capture techniques (e.g., mist-netting), which arelabor-intensive and oftentimes low-yielding, have tradition-ally provided biologists with information about the presenceor absence of bat species within particular habitats (Kunz andKurta 1988, O’Farrell and Gannon 1999). However, withpopulation declines of many bat species across NorthAmerica due to habitat loss and degradation (Weller et al.2009), wind-energy-related fatalities (Arnett and Baerwald2013), and the white-nose syndrome fungal disease (Cryanet al. 2013), sampling via capture techniques has becomeincreasingly ineffective. Ultrasonic bat-detector technologies

provide viable alternatives for studying presence–absence orthe relative activity of bat species (Murray et al. 1999,O’Farrell and Gannon 1999, Staton and Poulton 2012). Batdetectors sample larger areas and often record higher speciesrichness than do capture techniques (Murray et al. 1999,O’Farrell and Gannon 1999). Furthermore, certain bat-detector models can record autonomously (Miller 2001),which allows numerous areas to be sampled simultaneouslywith minimal effort (Gorresen et al. 2008). Accordingly,acoustic monitoring via bat detectors has become a popularresearch method; however, limitations of this samplingtechnique, if not addressed, may result in incorrectconclusions and hinder effective management.Acoustic bat detectors suffer from unavoidable sampling

biases (Murray et al. 1999, O’Farrell and Gannon 1999,Hayes 2000). Bat detectors sample finite airspaces; thus, onlya limited area exists in which echolocating bats will bedetected (Limpens and McCracken 2004). The size of theairspace a bat detector samples (henceforth, “sampling area”)

Received: 6 June 2014; Accepted: 11 April 2015

1E-mail: joyokeefe@gmail.com

Wildlife Society Bulletin; DOI: 10.1002/wsb.572

Kaiser and O’Keefe � Data Acquisition by Bat Detectors 1

is a function of microphone design and local recordingconditions. Sampling areas are dictated by the microphone’ssensitivity (Larson and Hayes 2000), frequency response(Waters and Walsh 1994, Adams et al. 2012), directionality(Downes 1982, Waters and Walsh 1994, Limpens andMcCracken 2004), and weatherproofing (Britzke et al. 2010,Sprong et al. 2012), as well as environmental factors thataffect the transmission of sound waves through air (Limpensand McCracken 2004). Climatic conditions influence therate at which sound attenuates in air (Griffin 1971, Lawrenceand Simmons 1982, Pettersson 2004), thereby affecting thesize of the sampling area. Clutter (e.g., vegetation) createsinterference (Schnitzler and Kalko 2001, Broders et al. 2004,Obrist et al. 2011) by refracting, reflecting, or scatteringultrasounds (Pettersson 2004), and the orientation of themicrophone in relation to clutter affects the size of thesampling area (Limpens and McCracken 2004). Detectorsdeployed in uncluttered environments have larger samplingareas and record higher relative bat activity than do detectorsplaced near vegetation (Weller and Zabel 2002).The data-acquisition abilities of bat detectors are dictated

by relationships between the microphones’ sampling area andits response to varying sound-sources in dynamic environ-mental and atmospheric conditions (Limpens andMcCracken 2004). The amount of bat activity a detectorwill record varies due to the amplitude, frequency, and signaltype of the emitted ultrasounds (Downes 1982, Forbes andNewhook 1990, Limpens and McCracken 2004, Adamset al. 2012). High-frequency sounds attenuate more rapidlyand travel shorter distances in air than do low-frequencysounds (Lawrence and Simmons 1982, Parsons 1996,Pettersson 2004). Broadband frequency-modulated signalshave lower source amplitudes and, thus, travel shorterdistances than do narrowband constant-frequency or quasi-constant-frequency calls (Schnitzler and Kalko 2001,Limpens and McCracken 2004). Therefore, it may bemore difficult to detect species with high-frequencyfrequency-modulated calls (e.g., Myotis septentrionalis)than species with low-frequency quasi-constant-frequencycalls (e.g., Lasiurus cinereus); however, it is important toconsider that detectability is largely influenced by thedistance of the sound source from the microphone (Corbenand Fellers 2001) because of the effects of geometricspreading and atmospheric absorption (Lawrence andSimmons 1982, Pettersson 2004).There are several factors that affect the capabilities of bat

detectors when deployed in field settings. Experimentssuggest that detector height (Staton and Poulton 2012), theorientation and elevation of the microphone in relation toclutter (Weller and Zabel 2002), as well as microphonesensitivity (Waters andWalsh 1994, Larson andHayes 2000,Adams et al. 2012) and weatherproofing (Britzke et al. 2010,Sprong et al. 2012) influence data acquisition. Althoughstudies have shown that detectors of the same model can varyin their capabilities (Larson and Hayes 2000, Weller andZabel 2002, Britzke et al. 2010, Sprong et al. 2012), we stillhave only limited information on differences in dataacquisition across different models or brands (especially

those released in the past decade due to the rapid growth ofthe commercial market for bat detectors). Manufacturers ofacoustic bat detectors use distinct hardware components,including proprietary microphones (Waters andWalsh 1994,Pettersson 2004, Adams et al. 2012). Consequently,sampling areas and detection ranges vary between differentdetector models (Parsons 1996, Adams et al. 2012, Spronget al. 2012) and, therefore, data acquisition will also likelyvary (Waters and Walsh 1994).To our knowledge, only a few studies have compared bat

detectors produced by different manufacturers (e.g., Forbesand Newhook 1990, Waters and Walsh 1994, Parsons 1996,Adams et al. 2012, Sprong et al. 2012), and with theexception of Sprong et al. (2012), no study accounted for theeffects of weatherproofing on data acquisition. Waters andWalsh (1994) found that detectors with greater microphonesensitivities consistently recorded more bat calls per hour andParsons (1996) found significant differences among brandsin regard to minimum and maximum detection distances.More recently, D. Solick (Western Ecosystems Technology,Inc., personal communication) demonstrated that WildlifeAcoustics SM2BATþ (Wildlife Acoustics, Inc., Concord,MA) and Pettersson D500x (Pettersson Elektronik AB,Uppsala, Sweden) full-spectrum detectors could not replicatethe data-acquisition abilities of an Anabat SD1 (TitleyScientific, Inc., Columbia, MO), which is a frequencydivision detector. The differences observed were likely due toincompatible software features or microphone differences.Adams et al. (2012), using both ultrasonic playbacks and fieldrecordings, determined that variation in directionality andmicrophone frequency responses led to performance differ-ences among 5 models of modern bat detectors. In this samestudy, SM2s outperformed Anabat SD2s; however, thesesystems were not weatherproofed when field-tested. Spronget al. (2012), also using ultrasonic playbacks in a controlledsetting, demonstrated that the effects of weatherproofingvary by detector model and microphone type. Because earlierfield studies did not account for the effects of varyingweatherproofing strategies on recordings of free-flying bats,further research is required.We initiated a field study near Indianapolis, Indiana, USA,

to compare data acquisition between 2 types of automatedbat detectors: the Wildlife Acoustics SM2BATþ (hence-forth, “SM2”) and Anabat SD2 (henceforth, “Anabat”;Titley Scientific, Inc.). We chose Anabats and SM2s becausethese 2 models are widely used for passive acoustic sampling,and they vary greatly in terms of microphone design,technical specifications, and weatherproofing strategies.SM2s use omnidirectional microphones with porous,foam, weatherproof coverings, whereas Anabats use direc-tional microphones that are often weatherproofed inimpermeable polyvinyl chloride (PVC) tubes. Thus, eachhas a distinct sampling area attributable to microphonedesign, directionality, and weatherproofing. We aimed todetermine whether these 2 models, when deployed side-by-side in the field and weatherproofed in a manner suitable forlong-term passive sampling, would record similar numbers ofecholocation files per phonic group. Overall, we expected

2 Wildlife Society Bulletin � 9999

that data acquisition for these detectors would be dissimilar.Also, like Weller and Zabel (2002), we aimed to determinehow microphone elevation influences data acquisition foreach detector type. We expected that microphones elevatedhigher above understory vegetation would record more batactivity than would microphones deployed nearer to theground.

STUDY AREA

This study was conducted on 1,045 ha of property west–southwest of the Indianapolis International Airport betweenthe cities of Mooresville and Plainfield, Indiana (Fig. 1). TheEast Fork of White Lick Creek, an 84-km-long perennial

stream, flowed south through the study site (Whitaker et al.2004). Interstate 70 (I-70), running northeast–southwest,divided the study site into “northern” and “southern”sections, with State Road 67 forming the eastern andsouthern borders, and State Road 267 forming the westernborder (Fig. 1). The study site north of I-70 was limited to anarrow (50–200-m-wide), forested riparian corridor, becausethe surrounding landscape was heavily developed withairport runways and busy highways to the east, a warehousedistrict to the west, residential subdivisions to the north, andlarge electrical substations and accompanying power lines tothe south. Anthropogenic activity was commonplace andcommercial planes flew within a few hundred meters of the

Figure 1. Location of 71 acoustic sampling points (black dots) near a riparian corridor west of Indianapolis International Airport in central Indiana, USA,surveyed fromMay to August 2012 and 2013. The black polygon delineates the potential foraging area for Indiana bats (Myotis sodalis), which was derived fromcombined foraging telemetry data from 2002 to 2011.

Kaiser and O’Keefe � Data Acquisition by Bat Detectors 3

tree canopy throughout the night. Agricultural landsdominated the study site south of I-70, although parcelsof land have been permanently set aside as forested wildliferefuges, replanted wetlands or forests, or local parks availablefor recreation (Sparks et al. 2009). Overall, the area in andaround the study site had been heavily modified by humanactivities and consisted of numerous fragmented land-covertypes, including farmland, wetlands, pasture–grasslands,residential–commercial developments, forested ripariancorridors, and remnant deciduous forests.Since 1996, the study site has been home to a long-term

study of the summer foraging and roosting ecology of thefederally endangered Indiana bat (Myotis sodalis; Whitakeret al. 2004). From 1997 to 2013, 9 bat species have beenregularly captured in mist nets in the study area, includingbig brown bats (Eptesicus fuscus), silver-haired bats (Lasio-nycteris noctivagans), hoary bats (Lasiurus cinereus), easternred bats (Lasiurus borealis), evening bats (Nycticeius humer-alis), tri-colored bats (Perimyotis subflavus), northern long-eared bats (M. septentrionalis), little brown bats (M.lucifugus), and Indiana bats (Whitaker et al. 2004, O’Keefeet al. 2014).

MATERIALS AND METHODS

We combined 95% minimum convex polygons, derived fromforaging telemetry data collected for Indiana bats in previousyears (2002–2011; Sparks et al. 2005, J. Whitaker, IndianaState University, personal communication), to create a singlepolygon in ArcMap 10.0 that outlined potential Indiana batforaging habitat within our study site (Fig. 1). UsingArcMap, we generated random points within this polygon torepresent potential sampling locations. We used an aerialmap to manually discard random points that wereinaccessible (e.g., located on private property, buildings,roadways, or within agricultural fields). The remainingrandom points were numbered sequentially and samplingorder was determined using a random number generator. Ifsampling locations were determined to be inadequate whenin the field, the next random point was used. Our samplingeffort was generally restricted because private property andhuman activity was so commonplace within the study site.Therefore, we discarded locations where bat detectorequipment was in jeopardy of being stolen or tamperedwith (e.g., public parks) or damaged (e.g., active cowpastures). From 16 May to 6 August 2012 and 15 May to 7August 2013, we deployed bat detectors at 71 random pointswithin this polygon, sampling 35 sites in 2012 and 36 sites in2013. We sampled in several land-cover types, includingwetland–riparian areas (31% of our sampling effort),deciduous forests (28%), pasture–grasslands (21%), replantedforests<20 years old (16%), and farmlands (4%).We did notsample land-cover types in proportion to their availability onthe landscape.We used omnidirectional SMX-USmicrophones (Wildlife

Acoustics, Inc., Concord, MA) and directional High Mount(Titley Scientific, Columbia, MO) microphones with theSM2s and Anabats, respectively. Wildlife Acoustics releaseda directional horn for the SMX-US microphone after we

began collecting data in 2012 and, thus, we were unable totest this directional set-up during our study. To reducesampling-area variation among the Anabat units wedeployed, each season we calibrated microphones inaccordance with Larson and Hayes (2000) using a constant40-kHz tone emitted from an Anabat Chirper (TitleyScientific). Anabats and microphones were purchased in2011 and sensitivity settings ranged from 5.5 to 6.5. At thetime of this study, there were no published methods forcalibrating SM2 microphones; however, we purchased alldetectors prior to beginning fieldwork in 2012 and used newSM2 microphones each year. Anabats recorded zero-crossing data with a division ratio of 8. SM2s recordedmonaural, full-spectrum data with a bit depth of 16, asampling rate of 384,000 kHz, and a WAC0 audiocompression. The following settings were used with theSM2, trigger: trig left¼ 18 dB, trig win left¼ 2.0 s; gain:48 dB, left microphone¼þ0.0 dB; high-pass filter: HPFleft¼ fs/24; and low-pass filter: LPF left¼OFF. Weprogrammed detectors to passively record from predusk(2000 hr EDT) until postdawn (0800 hr EDT) each nightfor 2 consecutive detector-nights at each sampling point. Adetector-night was defined as one uninterrupted night ofrecording. We included data in our analyses only if a fulldetector-night of sampling was achieved; and in 2012 and2013 we discarded 2 nights and 16 nights of sampling,respectively, because of Anabat malfunctions. Consequently,we only included data from 33 of 36 sites sampled in 2013 inour analysis.We paired one Anabat with one SM2 at each of 35 points

in 2012, and 2 Anabats with 2 SM2s at each of 36 points in2013. To replicate current protocols for passive acousticmonitoring studies, we used what we have observed (viapublished research, e.g., Britzke et al. 2010, and presenta-tions at regional and national bat conferences) to be the mostpopular weatherproofing solution for each detector type. Weused the foam shielding included with SMX-US micro-phones and placed Anabat microphones in 458 PVC tubes(Britzke et al. 2010). To record the best calls possible, weselected a point with the least amount of clutter �25m fromthe random point and oriented microphones towardpotential bat flyways (Larson and Hayes 2000). From 16May to 6 August 2012, we employed a double-observermethod similar to Duchamp et al. (2006), with one SM2 andone Anabat paired and programmed to record simultaneous-ly at each sampling site. Each microphone was connected toits respective detector with a 10-m cable and elevated 2m ona metal post (Weller and Zabel 2002). Each Anabatmicrophone was positioned 15 cm above the SM2 micro-phone to reduce the Anabat’s PVC enclosure from blocking alarge portion of the SM2’s sampling area; we assumed thisdistance was negligible in terms of differences in samplingspace for the 2 microphones (Fig. 2). From 15 May to 7August 2013, we used a similar set-up, but we added a secondSM2–Anabat pairing at 5m above ground, again withAnabat microphones positioned 15 cm above SM2 micro-phones. On account of concerns about theft or disturbance toour research efforts in the open environment where we

4 Wildlife Society Bulletin � 9999

worked, we did not feel comfortable testing a microphoneheight>5m.We used identical microphone orientations forthe SM2–Anabat pairings at both 2-m and 5-m heights, andwe randomly determined each coupling of detectors prior todeployment. To simplify discussion, 2-m microphoneelevations are referred to as “low” and 5-m microphoneelevations are referred to as “high.”

Echolocation Data AnalysisFor this study, a bat call, also referred to as a “file,” is definedas a series of�3 consecutive echolocation pulses emitted by asingle bat (Ford et al. 2005). A pulse is defined as a singleecholocation sound wave within the larger bat call. UsingWAC2WAV 3.3.0 (Wildlife Acoustics, Inc.) software, weconverted full-spectrum SM2 data to zero-crossing formatwith parameters identical to those of the Anabat (divisionratio¼ 8, max. file duration¼ 15 sec, and min. time betweencalls¼ 2 sec). We applied the SMX-US-to-UT filter inWAC2WAV to achieve a flatter frequency response, assuggested byWildlife Acoustics (J. King,Wildlife Acoustics,Inc., personal communication). We used Bat Call ID 2.5b(henceforth, “BCID”; Bat Call Identification, Inc., KansasCity, MO) automated software to identify bat call files to 1 of

3 phonic groups (low, midrange, or Myotis; Romeling et al.2012). The software referred to a call library for bats thatoccur in Indiana (1,546 calls of 9 species), required aminimum of 3 pulses within 15 seconds for file identification(Romeling et al. 2012), and reported a conservative �90%phonic-group confidence level. BCID categorized files intophonic groups based on a clustering algorithm and callparameters such as duration, minimum frequency, slope atthe flattest portion of the call, and frequency at the knee ofthe call (R. Allen, Bat Call Identification, Inc., personalcommunication; e.g., Romeling et al. 2012). The low phonicgroup (min. call frequencies<30 kHz) contained E. fuscus, L.noctivagans, and L. cinereus; the midrange phonic group(non-Myotis bats with min. frequency >30, but <50 kHz)contained L. borealis, N. humeralis, and P. subflavus; and theMyotis phonic group (min. call frequencies between 30 and60 kHz) contained M. lucifugus, M. septentrionalis, and M.sodalis (Romeling et al. 2012). We excludedM. leibii andM.grisescens from BCID analyses because these species have notbeen regularly captured with mist nets in the study site overthe past 15 years (Whitaker et al. 2006, J. O’Keefe,unpublished data). We excluded files that could not beidentified to phonic group by BCID because of insufficientpulse counts or poor call quality.

Statistical AnalysesWe conducted separate analyses on the data we collected in2012 and 2013. For 2012, we compared low Anabatmicrophones versus low SM2 microphones (i.e., bothelevated 2m). For each phonic group, we tallied the totalnumber of files recorded per night for each detector typefrom the BCID outputs. We then calculated the mean filecount per night per phonic group per sampling site for eachdetector. We used generalized estimating equation models,accounting for repeated measures by site, with a negativebinomial distribution and log link to test the effects of phonicgroup, detector type, and their interaction on mean file countper site per detector type (response variable).We conducted asimilar analysis for 2013 data, but added height and itscorresponding 2-way and 3-way interactions as independentfactors. We used least-significant-differences pairwisecomparisons to compare responses for significant tests.We used SPSS 20.0.0 (IBM Corporation, Armonk, NY) forall tests and assessed significance at a¼ 0.05.

RESULTS

Anabats recorded 10,092 identifiable bat calls over the courseof the study, whereas SM2s recorded 5,888 identifiable batcalls (Table 1). We recorded more identifiable files for bothdetector types in2013 (whenwedeployeddetectors atboth2mand5m;Table 1), but in both years a greater proportion of fileswere identifiable from Anabat recordings. For SM2s, BCIDcould not identify 43.1–51.7% of all files recorded, whereasBCIDcould not identify 38.6–44.6%of allAnabat files.Thesepercentages do not include nonbat files (e.g., noise), becausethese files were discarded by BCID’s default filter.In 2012, data acquisition varied by detector type, phonic

group, and detector type� phonic group (P� 0.002;

Figure 2. Acoustic sampling points were surveyed for bats from May toAugust 2012 and 2013 in central Indiana, USA. Acoustic microphones wereset up such that Anabat microphones housed in 458 polyvinyl chloride tubeswere 15 cm above SM2 microphones. We paired one Anabat microphonewith one SM2 microphone at 2m above ground in 2012 and, in 2013, weadded a second SM2-Anabat pairing at 5m above ground.

Kaiser and O’Keefe � Data Acquisition by Bat Detectors 5

Table 2). Anabats recorded more files from all phonic groupscombined (�x¼ 13.72� 3.14 files/site/night) than SM2s(9.05� 2.28). Low-phonic-group calls were recorded mostoften by both detectors, followed by midrange calls, and thenMyotis (Fig. 3A). However, least-significant-differencespairwise comparisons showed that differences were onlystatistically significant for low versus Myotis (P¼ 0.008) andmidrange versus Myotis (P¼ 0.011). Odds ratios indicatedthat, relative to Myotis, low- and midrange-phonic-groupcalls were 3.8 times and 2.1 times more likely to be recorded,respectively. When considering the detector type� phonicgroup interaction, Anabats recorded more low (�x difference¼ 22.3� 9.9, P¼ 0.023) and midrange calls (�x difference¼ 14.3� 4.4, P¼ 0.001) than SM2s, but recorded fewerMyotis calls than SM2s (�x difference¼�1.8� 0.7,P¼ 0.013; Fig. 3A).In 2013, detector type was not a significant factor in

explaining differences in data acquisition (P¼ 0.754;Table 2). Anabats and SM2s recorded, on average, similarmean file counts per site per night (7.26� 1.24 and7.03� 1.43, respectively). However, data acquisition didvary by phonic group and detector type� phonic group in2013 (P< 0.001; Table 2). As in 2012, low-phonic-groupcalls were recorded most often by both detectors, followed bymidrange calls, and thenMyotis. Significant differences wereobserved between each phonic group pair (Fig. 3B). RelativetoMyotis, the odds of recording low- and midrange-phonic-group calls were 14.7 and 3.6 times greater, respectively;whereas odds of recording low-phonic-group calls were 2.6times greater relative to midrange calls. The detectortype� phonic group effect was similar to 2012; Anabats

recordedmore low (�x difference¼ 10.7� 3.6, P¼ 0.003) andmidrange calls (�x difference¼ 6.6� 1.5, P< 0.001) thanSM2s, but recorded fewer Myotis calls than SM2s (�xdifference¼�1.2� 0.4, P¼ 0.004; Fig. 3B). Height and its2-way and 3-way interactions were not significant factorsexplaining differences in data acquisition (Table 2). Whencomparing between detector types, Anabats and SM2sperformed similarly at both 2-m and 5-m heights (Table 1;Fig. 4).

DISCUSSION

As expected, the omnidirectional SMX-US microphone ofthe SM2BATþ and the directional High Mount micro-phone of the Anabat SD2 recorded different mean file countsper site for all phonic groups. Although we did not directlytest for the source of this variability, the observed data-acquisition differences are likely attributable to a combina-tion of directionality (Waters and Walsh 1994), theinteraction between frequency response and directionality(Adams et al. 2012), and weatherproofing differences(Britzke et al. 2010, Sprong et al. 2012). Overall, Anabatsrecorded more identifiable bat calls than SM2s whenweatherproofed, but the relative performance of eachdetector type varied by phonic group. Although our resultscontrast with earlier comparative studies of bat detectors(e.g., Adams et al. 2012), the observed disparities likely relateto differences in methods and study environments. Weobserved that slight differences in detector height maymatter less than differences in detector models whendesigning passive acoustic studies.

Table 1. Total number of identifiable bat files and mean (�SE) bat files recorded per night by phonic group for Anabat SD2 and SM2BATþ acousticdetectors deployed at sites near the Indianapolis International Airport in central Indiana, USA, May to August 2012 and 2013. Anabat and SM2 data arepresented for low (2-m) and high (5-m) microphone elevations.

Total bat files recorded Mean bat files recorded (�SE)

Phonic group Phonic group

Year Mic ht Detector No. of sites sampled No. of nights sampled Low Mid Myotis Low SE Mid SE Myotis SE

2012 Low Anabat 35 68 1,812 1,606 164 39.8 12.1 23.7 7.2 2.7 0.7SM2 35 68 771 619 263 17.5 7.7 9.3 3.0 4.5 1.2

2013 Low Anabat 33 61 2,337 1,012 46 35.8 10.2 14.1 3.8 0.8 0.2SM2 33 61 1,288 636 152 20.1 7.1 8.7 3.7 2.4 0.8

High Anabat 33 63 2,111 965 39 31.4 7.9 13.9 2.8 0.8 0.2SM2 33 63 1,622 444 93 26.0 9.4 6.3 1.7 1.8 0.4

Table 2. Significance tests for generalized estimating equation models for differences in data acquisition by bat detectors deployed near the IndianapolisInternational Airport in central Indiana, USA, May to August 2012 and 2013.

2012 2013

(2-m mics; 35 sites) (2-m and 5-m mics; 33 sites)

Effect df Wald x2 P df Wald x2 P

Detector type 1 9.326 0.002 1 0.098 0.754Phonic group 2 29.144 <0.001 2 75.898 <0.001Detector type� phonic group 2 76.725 <0.002 2 28.216 <0.001Height 1 0.375 0.540Height� detector type 1 0.186 0.666Height� phonic group 2 2.594 0.273Height� phonic group� detector type 2 3.393 0.183

6 Wildlife Society Bulletin � 9999

Both Anabats and SM2s recorded low-phonic-group filesmost often, followed by midrange bat calls, and lastlyMyotis.Differences in detection rates for these 3 phonic groups mayrelate to both the capabilities of bat detectors and todisproportionate detectability and species abundance amongphonic groups. Low-frequency ultrasounds travel farther in airand attenuate less rapidly than high-frequency ultrasounds(Lawrence and Simmons 1982, Parsons 1996, Pettersson

2004). Furthermore, broadband frequency-modulated signalsused by Myotis bats have lower source amplitudes and travelshorter distances than the quasi-constant-frequency orfrequency-modulated calls used by low and midrangefrequency bats (Schnitzler and Kalko 2001, Limpens andMcCracken 2004). Consequently, when surveying withacoustic bat detectors, we are more likely to detect low-phonic-group bats than high-frequency bats such as Myotisspecies (Adams et al. 2012). It is possible that differences in therelative abundance of bats in each phonic group may alsopartially explain the observed variation by phonic group. Forexample, low-phonic-group bats (i.e., primarily E. fuscus) arecaptured most often within our study site, followed bymidrange species (i.e., L. borealis and P. subflavus), and lastlyMyotis species (O’Keefe et al. 2014). However, we note thatthere are likely differences in the probability of detecting eachphonic group via capture surveys, becausewe avoided samplingin high-clutter areas where Myotis bats might be more active(Aldrich and Rautenbach 1987, Owen et al. 2004), and somespeciesmaybe less susceptible to capture thanothers (O’Farrelland Gannon 1999).Overall, the observed data-acquisition inconsistencies

between the 2 detector models we tested are likelyattributable to microphone sensitivity, frequency response,and directionality differences (Waters and Walsh 1994,Adams et al. 2012), but weatherproofing strategies alsocreate variability (Britzke et al. 2010, Sprong et al. 2012, C.Corben, www.hoarybat.com, Columbia, Missouri, personalcommunication), especially in regard to recording specificphonic groups. When Anabats and SM2s are weath-erproofed in a manner typical of most long-term passive

Figure 3. Mean bat call files recorded per phonic group per site fromMay to August 2012 (A) and 2013 (B) for Anabat SD2 and SM2BATþ acoustic detectorsdeployed at sites near the Indianapolis International Airport in central Indiana, USA. In each year, all possible pairwise comparisons were significantly differentfor the detector type� phonic group interaction (P < 0.05), as indicated by the different letters above each bar.

Figure 4. Mean bat call files recorded per phonic group by detector heightfor Anabat SD2 and SM2BATþ acoustic detectors deployed at sites near theIndianapolis International Airport in central Indiana, USA, May toAugust 2013. We did not observe significant effects of detector height, norof its 2-way or 3-way interactions with detector type and phonic group.

Kaiser and O’Keefe � Data Acquisition by Bat Detectors 7

acoustic monitoring studies, our results suggest that SMX-US microphones detect more Myotis bat activity, but lesslow- and midrange-phonic-group bat activity, than doAnabats. The omnidirectional SMX-US is capable ofrecording echolocating bats with flight paths above, below,and behind the microphone. The greater directionality of anAnabat microphone (relative to the SMX-US) limits itsability to sample areas not directly in front of the microphone(Sprong et al. 2012) and the PVC weatherproofing enclosurelikely further exacerbates this phenomenon. For example,Sprong et al. (2012) found that Anabat SD2 microphoneswithout weatherproofing could not detect nearby (i.e., 1mfrom microphone) ultrasounds broadcasted 1808 behind themicrophone. Additionally, the foam weatherproofing of theSMX-US is more permeable to sound than the PVC and,thus, the SM2s used in our study were likely more exposed toultrasounds in the environment. A less directional and moreexposed microphone, such as the SMX-US, shouldtheoretically have a better chance of recording a low-intensityMyotis call (or any bat call) before the call attenuates(Waters and Walsh 1994). This correlates with our findingthat SM2s recorded moreMyotis than did Anabats; however,resonance created by the PVC tube likely further inhibitedthe Anabat’s ability to record sounds above 40 kHz (i.e.,Myotis bats in our study site; Corben 2011). This does notexplain why SM2s, on average, recorded fewer low- andmidrange-phonic-group files than did Anabats. Relative toMyotis, low- and midrange-phonic-group calls possesshigher source amplitudes, travel farther in air, and are easierto detect with bat detectors at greater distances (Lawrenceand Simmons 1982, Corben 2004). The fact that fewer lowand midrange bats were recorded by SM2s suggests thatAnabat microphones may be more sensitive to theseparticular frequency ranges, and are capable of recordingthese “louder” ultrasounds at distances outside the samplingarea of SM2s (Waters and Walsh 1994, Limpens andMcCracken 2004). In fact, Adams et al. (2012) observed thatSM2s had a greater rate of attenuation for low-frequencyultrasounds (e.g., 25 kHz) relative to Anabats, whichsupports the idea that the interaction between frequencyresponse and directionality influences data acquisition.Our results differed from Solick et al. (2011) and Adams

et al. (2012) in 2 ways: 1) we recorded more files withAnabats, whereas those studies recorded more files withSM2s; and 2) we recorded more low- and midrange-frequency calls with Anabats, whereas Adams et al. (2012)recorded more low- and midrange-frequency calls (e.g.,25 kHz and 55 kHz) with SM2s. It seems unlikely thatSMX-US microphones would sample farther than Anabats,because omnidirectional microphones are typically lesssensitive than their directional counterparts (Waters andWalsh 1994, Limpens and MacCracken 2004, Adams et al.2012). Adams et al. (2012) attributed variation in dataacquisition to the abilities of specific detectors to detectcertain frequency ranges at varying distances from themicrophone. In the study by Adams et al. (2012), Anabatsdid not record signals >85 kHz, whereas SM2s did; andAnabats detected low-frequency ultrasounds (e.g., 25 kHz)

at greater distances than SM2s. This corresponds with ourfinding that Anabats recorded fewer high-frequency Myotiscalls, but seemed to have greater detection ranges for low-frequency sounds when compared with SM2s.We attribute the disparity between our results and those of

Solick et al. (2011) and Adams et al. (2012) to differencesamong our studies with respect to weatherproofing,deployment height–microphone orientation, gain–triggersettings (D. Solick, personal communication; Adams et al.2012), data postprocessing, the use of artificial ultrasoniccalls by Adams et al. (2012), and the fact that each study wasconducted in a different environment (Parsons 1996). Weweatherproofed Anabats, paired microphones on a verticalpost and not in a horizontal plane, postprocessed datadifferently (e.g., WAC0 conversion to zero-crossing formatusing WAC2WAV software), and unlike Adams et al.(2012) we did not use synthetic calls. In addition, each ofthese studies was conducted in a different fine-scale spaceand climatic region. Though necessary, making comparisonsacross studies that occur in different field settings ischallenging because the rate at which sound attenuates—and thus, the sampling area—will vary on account ofdifferences in climate and vegetation (Parsons 1996, C.Corben, personal communication); further, local batassemblages also may be dissimilar.Height (2m vs. 5m) did not affect the total number of bat

files recorded by either SM2s or Anabats. We expectedmicrophones elevated at greater heights above theunderstory vegetation would record significantly morebat activity (Weller and Zabel 2002, Staton and Poulton2012); however, mean file counts per site were similar forboth detector types. We expected high SM2s to outperformlow SM2s, because many of the calls collected in 2012 with2-m microphones were laden with noise, perhaps due toclose proximity to understory vegetation. However, 2-mand 5-m SM2 microphones performed similarly in 2013,suggesting that these microphones either had overlappingsampling areas or they are just inherently sensitive toenvironmental noise when deployed using their defaultgain settings. Anabats at 2m and 5m heights alsoperformed similarly, which further suggests that a distanceof 3m between microphones may be insubstantial fordiscerning differences in bat activity due to overlappingsampling areas. It is likely that we would have observedmore pronounced differences in bat activity if we had beenable to compare microphones deployed at more dissimilarheights (Staton and Poulton 2012).We postprocessed full-spectrum SM2 data to zero-crossing

format usingWAC2WAV 3.3.0 software in order to match thefile parameters of Anabats, but we do not know whetherpostprocessing data from one file format to another affects totalfile counts. Originally, we intended to conduct a spectral analysisand automatic identification of SM2 data using Sonobat 3.1NE(SonoBat, Arcata, CA); however, after one season of datacollection, we noticed that SM2 calls were burdened with noiseand audio distortion in the form of clipping. We struggled toidentify these data, and some bat calls were visually indiscerniblefromnoise.Wepostprocessed full-spectrumfiles tozero-crossing

8 Wildlife Society Bulletin � 9999

format (required by BCID) in WAC2WAV, hoping BCIDcould better distinguish bat calls from noise. When testing the“skip noise” feature in WAC2WAV, many bat call files wereclassified as noise. Therefore, we did not implement this feature,but rather relied on the default BCID filter to remove filescontaining noise. In 2013, we had the option to record directly inzero-crossing formatwithSM2s,whichwould have resolved thispotential postprocessing issue. However, we were instructed byWildlife Acoustics that SM2s detect more bats in triggered wavmode versus zero-crossingmode, sowe continuedwith our 2012methods. It is no longer possible to test the effects ofpostprocessingwithWAC2WAV,becausethis softwarepackagehas recently been retired by Wildlife Acoustics. Instead, werecommend recording in the format best suited to your researchquestion and avoiding file conversions, if possible.Acoustic bat detectors sample small areas over relatively

narrow periods of time and data acquisition among differentmodels may vary significantly (Waters and Walsh 1994,Parsons 1996, Solick et al. 2011, Adams et al. 2012, thisstudy). However, different models, microphones, andweatherproofing strategies will continue to be used, ofteninterchangeably, in bat research across the globe. Without acomplete understanding of the capabilities and limitationsof modern bat detectors, studies designed to assess batactivity will yield inaccurate estimates and may undersamplesome phonic groups or bat species. It is likely that passiveacoustic sampling will continue to play larger roles in futurebat-monitoring efforts, so decisions about acoustic equip-ment and deployment could have significant effects onspecies management. For instance, the U.S. Fish andWildlife Service has recently implemented major changes tothe summer survey protocol for the federally endangeredIndiana bat (USFWS 2015) by requiring passive acousticsampling using only directional microphones (e.g., AnabatSD2, among other models). In this case, the stock SMX-USomnidirectional microphone (without Wildlife Acoustic’sdirectional horn) could not be used, even though our resultssuggest this microphone detects more Myotis bat activitythan directional Anabat microphones housed in PVC. Onthe other hand, Anabats may be better suited for acousticsurveys at wind energy facilities if the target species are low-and midrange-phonic-group bats (e.g., L. noctivagans, L.cinereus, and L. borealis), which comprise the bulk of wind-energy-related fatalities in North America (Arnett andBaerwald 2013). These are merely 2 examples, but if wewant to better portray reality with acoustic sampling efforts,it is imperative that further comparative studies beconducted to determine the capabilities and limitationsof different bat-detector models and how each performsrelative to another. It is also crucial to develop astandardized approach for weatherproofing systems, hard-ware–software settings, and deployment techniques, ifpossible. It is unlikely that any single methodology willuniversally satisfy the research or management goals of allacoustic studies (Adams et al. 2012), but minimizingvariation will undoubtedly be beneficial. Therefore, asrecommended by Adams et al. (2012), we also encourageauthors to present detailed information about their

methods, especially technical hardware–software settingsused with bat detectors.Lastly, deploying a combination of bat detector models

may be the best option for reducing bias and obtaining moreaccurate results during acoustic surveys. We stress that ourresults represent one location and one bat assemblage.Results will likely vary with location, species assemblage,habitat type, and weather. Also, we recognize that the scopeof this study was limited as a result of comparing only 2different bat-detector systems. Including several otherbrands of detectors would have enhanced this study;however, we were limited financially in acquiring suchequipment. Bat biologists must make every effort tounderstand and reduce our sampling biases when studyingthese small, volant, nocturnal mammals. Additionalcomparative studies will continue to inform best practicesfor bat research, which will ultimately aid in themanagement and conservation of these valuable wildlifespecies.

ACKNOWLEDGMENTS

We thank the Indianapolis Airport Authority for funding andproperty access. Special thanks to B. Walters and J. Helms(Indiana State University), D. Solick (WEST, Inc.), S.Amelon (U.S. Forest Service), C. R. Allen (Bat Call ID), andC. Corben for logistical help, advice on project design,software, and hardware support. For field assistance, we thankG. Auteri, J. Cox, A. Gondran, D. Howard, M. Johnson, V.Kuczynska,A.Nolder,C. Schweizer, J.Weber, andS.Wiram.We thank Drs. C. Murphy and S. Lima for comments onexperimental design, analyses, and earlier drafts of thismanuscript. We also thank S. DeStefano and 2 anonymousreviewers who provided comments on this manuscript.

LITERATURE CITEDAdams, A.M.,M. K. Jantzen, R.M.Hamilton, andM. B. Fenton. 2012.Doyou hear what I hear? Implications of detector selection for acousticmonitoring of bats. Methods in Ecology and Evolution 3:992�998.

Aldrich, H. D. N., and L. L. Rautenbach. 1987. Morphology, echolocationand resource partitioning in insectivorous bats. Journal of Animal Ecology56:763�778.

Arnett, E. B., and E. Baerwald. 2013. Impacts of wind energy developmenton bats: implications for conservations. Pages 435–456 in R. A. Adamsand S. C. Pederson, editors. Bat ecology, evolution, and conservation.Springer Science Press, New York, New York, USA.

Britzke, E. R., B. A. Slack, M. P. Armstrong, and S. C. Loeb. 2010. Effectsof orientation and weatherproofing on the detection of bat echolocationcalls. Journal of Fish and Wildlife Management 1:136�141.

Broders, H. G., C. S. Findlay, and L. Zheng. 2004. Effects of clutter onecholocation call structure of Myotis septentrionalis and M. lucifugus.Journal of Mammalogy 85:273�281.

Corben, C. 2004. Zero-crossings analysis for bat identification: an overview.Pages 95�107 in R. M. Brigham, E. K. V. Kalko, G. Jones, S. Parsons,and H. J. G. A. Limpens, editors. Bat echolocation research: tools,techniques and analysis. Bat Conservation International, Austin, Texas,USA.

Corben, C. 2011. Weather protection. http://users.lmi.net/corben/Weather%20Protection.htm. Accessed on 29 Sep 2014.

Corben, C., and G. M. Fellers. 2001. Choosing the ‘correct’ bat detector—areply. Acta Chiropterologica 3:253�256.

Cryan, P. M., C. U. Meteyer, J. G. Boyles, and D. S. Blehert. 2013. White-nose syndrome in bats: illuminating the darkness. BMC Biology 11:47.

Downes, C. 1982. A comparison of sensitivities of three bat detectors.Journal of Mammalogy 63:345�347.

Kaiser and O’Keefe � Data Acquisition by Bat Detectors 9

Duchamp, J. E., M. Yates, R. M. Muzika, and R. K. Swihart. 2006.Estimating probabilities of detection for bat echolocation calls: anapplication of the double observer method. Wildlife Society Bulletin34:408�412.

Forbes, B., and E. M. Newhook. 1990. A comparison of the performance ofthree models of bat detectors. Journal of Mammalogy 71:108�110.

Ford, W. M., M. A. Menzel, J. L. Rodrigue, J. M. Menzel, and J. B.Johnson. 2005. Relating bat species presence to simple habitat measures ina central Appalachian forest. Biological Conservation 126:528�539.

Gorresen, P. M., A. C. Miles, C. M. Todd, F. J. Bonaccorso, and T. J.Weller. 2008. Assessing bat detectability and occupancy with multipleautomated echolocation detectors. Journal of Mammalogy 89:11�17.

Griffin, D. R. 1971. The importance of atmospheric attenuation for theecholocation of bats (Chiroptera). Animal Behaviour 19:55�61.

Hayes, J. P. 2000. Assumptions and practical considerations in the designand interpretation of echolocation-monitoring studies. Acta Chiropter-ologica 2:225�236.

Kunz, T. H., and A. Kurta. 1988. Capture methods and holding devices.Pages 1�29 in T. H. Kunz, editor. Ecological and behavioral methods forthe study of bats. Smithsonian Institution Press,Washington, D.C., USA.

Larson, D. J., and J. P. Hayes. 2000. Variability in sensitivity of Anabat II batdetectors and a method of calibration. Acta Chiropterologica 2:209�213.

Lawrence, B. D., and J. A. Simmons. 1982. Measurements of atmosphericattenuation at ultrasonic frequencies and the significance for echolocatingbats. Journal of Acoustical Society of America 71:585�590.

Limpens, H. J. G. A., and G. F.McCracken. 2004. Choosing a bat detector:theoretical and practical aspects. Pages 28–37 in R. M. Brigham, E. K. V.Kalko, G. Jones, S. Parsons, H. J. G. A. Limpens, editors. Batecholocation research: tools, techniques and analysis. Bat ConservationInternational, Austin, Texas, USA.

Miller, B. W. 2001. A method for detecting relative activity of free flyingbats using a new activity index for acoustic monitoring. ActaChiropterologica 3:93�105.

Murray, K. L., E. R. Britzke, and B. M. Hadley. 1999. Surveying batcommunities: a comparison between mist nets and the Anabat II batdetector system. Acta Chiropterologica 1:105�112.

Obrist, M. K., E. Rathey, F. Bontadina, A. Martinoli, M. Conedera, P.Christe, and M. Moretti. 2011. Response of bat species to sylvo-pastoralabandonment. Forest Ecology and Management 261:789�798.

O’Farrell, M. J., and W. L. Gannon. 1999. A comparison of acoustic versuscapture techniques for the inventory of bats. Journal of Mammalogy80:24�30.

O’Keefe, J. M., S. M. Bergeson, C. Byrne, L. K. Castor, Z. D. Kaiser, B. L.Walters, and J. O. Whitaker, Jr. 2014. 2013 monitoring program for theIndiana myotis (Myotis sodalis) near the Six Points Interchange inHendricks and Marion Counties, Indiana as required under the SixPoints Interchange Habitat Conservation Plan. Center for Bat Research,Outreach, and Conservation. Indiana State University, Terre Haute,USA.

Owen, S. F., M. A. Menzel, and J. W. Edwards. 2004. Bat activity inharvested and intact forest stands in the Allegheny Mountains. NorthernJournal of Applied Forestry 21:154�159.

Parsons, S. 1996. A comparison of the performance of a brand of broad-bandand several brands of narrow-band bat detectors in two different habitattypes. Bioacoustics 7:33�43.

Pettersson, L. 2004. The properties of sound and bat detectors. Pages 9�12in R. M. Brigham, E. K. V. Kalko, G. Jones, S. Parsons, H. J. G. A.Limpens, editors. Bat echolocation research: tools, techniques andanalysis. Bat Conservation International, Austin, Texas, USA.

Romeling, S., C. R. Allen, and L. W. Robbins. 2012. Acoustically detectingIndiana bats: how long does it take? Bat Research News 53:51�58.

Schnitzler, H., and E. K. V. Kalko. 2001. Echolocation by insect-eating bats.BioScience 51:557�569.

Solick, D., C. Nations, and J. Gruver. 2011. Activities rates and call qualityby full-spectrum detectors. Bat Research News 52:59.

Sparks, D.W., V.W. Brack, Jr., J. O.Whitaker, Jr., and R. Lotspeich. 2009.Reconciliation ecology and the Indiana bat at Indianapolis InternationalAirport. Pages 1�15 in P. B. Larauge and M. E. Castille, editors.Airports: performance, risks, and problems. Nova Science Hauppauge,New York, New York, USA.

Sparks, D. W., C. M. Ritzi, J. E. Duchamp, and J. O. Whitaker, Jr. 2005.Foraging habitat of the Indiana bat (Myotis sodalis) at an urban-ruralinterface. Journal of Mammalogy 86:713–718.

Sprong, L., M. Keith, and E. C. J. Seamark. 2012. Assessing the effect ofwaterproofing on three different bat detectors. African Bat ConservationNews, ISSN 1812–1268.

Staton, T., and S. Poulton. 2012. Seasonal variation in bat activity in relationto detector height: a case study. Acta Chiropterologica 14:401–408.

U.S. Fish and Wildlife Service [USFWS]. 2015. Range-wide Indiana batsummer survey guidelines, April 2015. http://www.fws.gov/midwest/endangered/mammals/inba/surveys/pdf/2015IndianaBatSummerSurveyGuidelines01April2015.pdf Accessed 5Apr 2015.

Waters, D. A., and A. L. Walsh. 1994. The influence of bat detector brandon the quantitative estimation of bat activity. Bioacoustics 5:205�221.

Weller, T. J., P. M. Cryan, and T. J. O’Shea. 2009. Broadening the focus ofbat conservation and research in the USA for the 21st century. EndangeredSpecies Research 8:129�145.

Weller, T. J., and C. J. Zabel. 2002. Variation in bat detections due todetector orientation in a forest. Wildlife Society Bulletin 30:922�930.

Whitaker, J. O., Jr., D. W. Sparks, and V. W. Brack. 2004. Bats of theIndianapolis International Airport area, 1991�2001. Proceedings of theIndiana Academy of Science 113:151�161.

Whitaker, J. O., Jr., D. W. Sparks, and V. W. Brack, Jr. 2006. Use ofartificial roost structures by bats at Indianapolis International Airport.Environmental Management 38:28�36.

Associate Editor: DeStefano.

10 Wildlife Society Bulletin � 9999