Using multiple methods to assess detection probabilitiesof riparian-zone anurans implications for monitoring

Jacquelyn C GuzyAD Steven J PriceB and Michael E DorcasC

ADepartment of Biology University of Arkansas Fayetteville AR 72701 USABDepartment of Forestry University of Kentucky Lexington KY 40546 USACDepartment of Biology Davidson College Davidson NC 28035 USADCorresponding author Email jcguzyuarkedu

AbstractContext Both manual call surveys (MCS) and visual encounter surveys (VES) are popular methods used to monitor

anuranpopulationsRecent statistical developments specifically the development of occupancymodels that permit the useofdata from various survey methods to assess method-specific detection probabilities provide a rigorous framework forevaluating the effectiveness of field methods

AimTo compare species-specific detection probabilities generated byMCS andVES and to evaluate the effectiveness ofthese methods throughout the activity season of several riparian-zone anuran species

MethodsDuring 2010 and 2011 we sampled 21 sites along the Broad and Pacolet Rivers in SouthCarolina USA usingMCS and VES Anuran species were surveyed across three seasons (fall spring and summer) each year

Key results For six species MCS resulted in a higher mean probability of detection whereas VES resulted in a highermean probability of detection for four species In addition survey datewas an important influence on detection probability ofmost anurans when using MCS but largely unimportant when employing VES

Conclusions Our findings indicated that VES are as effective as MCS for detecting some species of anurans and forothers VES represent a more effective method Furthermore when using VES outside the breeding window some anuranscan be reliably detected and in some cases detected more easily than by using MCS

ImplicationsWe suggest that VES is a complimentary technique toMCS and a potentially important tool for populationmonitoring of anurans VES can provide more flexibility for anuran researchers as robust estimates of detection andoccupancy can be obtained outside a narrow breeding window

Additional keywords active search amphibian manual call survey South Carolina survey method visual encountersurvey

Received 4 March 2014 accepted 20 August 2014 published online 26 September 2014

Introduction

Manual call surveys (MCS) are a popular method used todetect anurans for ecological behavioural and conservation-related investigations (eg Blair 1961 Woolbright 1985Knutson et al 2004 Price et al 2005 Dorcas et al 2010) Inparticular using MCS to obtain anuran distribution dataprovides a relatively simple and cost-effective way to monitorpopulations across time and space (Droege and Eagle 2005Weir et al 2005 Dorcas et al 2010) A major assumption ofMCS is that peak anuran breeding windows are well understoodand surveys are well timed to coincide with anuran reproductiveactivities However a central concern in any survey methodincluding MCS is that of imperfect detection (MacKenzie et al2002 2006) Species detection probability is defined as theprobability of detecting at least one individual of a focalspecies during a sampling occasion given that individuals ofthe species are present in the area (MacKenzie et al 2002)

Although detection of a species confirms its presence lack ofdetection does not necessarily confirm absence Thus estimatesof site occupancy and abundance and their relationships withmeasured covariates (eg habitat type) along with estimatedprobabilities of colonisation and local extinction will be biasedunless methods are used that can account for imperfect detection(MacKenzie et al 2009) During MCS even within the peakbreeding season for many species calling does not occur eachnight and there are variations in calling behaviour because ofabiotic and biotic conditions which can lead to the incorrectinference of absence of a species (Gooch et al 2006 Branderet al 2007 Cook et al 2011) In addition recent studies haveshown that false positive errors (ie a species that is absent buterroneously detected) during anuran call surveys positively biasoccupancy estimates even when false positive detections occuronly 1 of the time (Miller et al 2011 2012) For these reasonsit is becoming increasingly common for MCS monitoring

CSIRO PUBLISHING

Wildlife Researchhttpdxdoiorg101071WR14038

Journal compilation CSIRO 2014 wwwpublishcsiroaujournalswr

programs to standardise survey methodologies and account forfactors that influence detection probability (Pellet and Schmidt2005 Weir et al 2005 Dorcas et al 2010)

Another common amphibian survey method is visual-encounter surveys (or active searches VES) where observersvisually search for amphibians in a designated area for aprescribed amount of time (Crump and Scott 1994) VES mayhold distinct advantages over MCS For example some speciesvocalise only in response to heavy rains (eg spadefoot toadsScaphiopus spp) call infrequently (eg gopher frogs Lithobatescapito) have relatively short breeding seasons (wood frogsLithobates sylvaticus) are uncommon (eg Pellet and Schmidt2005) or may not be audible over the louder higher-pitch callsof their contemporaries (Doan 2003 Droege and Eagle 2005)As with MCS VES are based on the assumption that allindividuals are equally detectable other major assumptions arethat there are no observer-related biases and that individuals arerecorded once during a survey (Vonesh et al 2010) Studieshave used VES to assess the habitat use by focal species(Burbrink et al 1998 Adams et al 2011) and to compare theeffectiveness of techniques such as cover-objects versus quadratmethodologies (Doan 2003 Grover 2006) and diurnal versusnocturnal searches (Heard et al 2006) but direct comparisonswith MCS have not been conducted Furthermore knowledgeof how time of year influences detection probabilities of anuransobserved via VES is not as well understood as it is for MCS

It has become increasingly common for amphibianresearchers to use multiple methods to detect species (Brownet al 2007 Mattfeldt and Grant 2007 Dahl et al 2009 Farmeret al 2009 Balas et al 2012) Some studies have expandedMCS to include vocal imitation of advertisement calls in anattempt to stimulate male frogs and improve detection alongwith other VES methods such as nocturnal spotlight surveysand diurnal visual searches (eg Heard et al 2006) Other studiesexamining optimal survey design in the tropics have used acombination of pitfall and funnel traps along with MCS andVES and automated tape recording of anuran calls (eg Parriset al 1999 Roumldel and Ernst 2004) Survey method is one of themost important considerations when designing studies Forexample when little is known about the natural history of aspecies incorporating several methods along with covariates ofinterest (eg day of year) can provide estimates of detectionprobability which allows researchers to identify the mostappropriate sampling method If detection probability isinfluenced by the time of year or weather conditions forexample survey protocols can be adjusted to minimise thechance of false absences Traditionally researchers usingmultiple survey methods either combined data from eachmethod to generate detection probabilities (ie Price et al2011) or separately generated detection probabilities for eachmethod to compare their utility (Bailey et al 2004 Mattfeldt andGrant 2007) Using multiple single-method analyses orcombined methods analysis ignores information from all butthe focal detection method (Nichols et al 2008) Recentadvances in modelling circumvent some of the drawbacks ofeither combining sampling data or conducting single-methodcomparisons Specifically multi-method models permitsimultaneous use of data from all methods for inference aboutmethod-specific detection probabilities (Nichols et al 2008)

In the present study our objectives were to evaluate twocommon methods for surveying anurans namely MCS andVES by using a statistical model that permits incorporation ofboth methods into a single model (Nichols et al 2008) Weexamined the generated detection probabilities for anuranspecies by using both methods and then examined how day ofyear influenced the effectiveness of each method Given thevariation in life history of many anurans for some species weexpected that VES would be comparable to MCS and sometimespreferable to MCS and that during certain times of the year onemethod would be preferable over the other

Materials and methodsStudy area

Study sites (n= 21) were located along the Broad and PacoletRivers in the Piedmont region of northndashcentral South Carolina(Fig 1)We used a geographic information system (ArcGIS 100Environmental SystemsResearch InstituteRedlandsCAUSA)with layers from the National Wetland Inventory (httpwwwfwsgovwetlands verified 20 July 2013) and the 2006NationalLand Cover Database (Fry et al 2011) to identify ~200 riparianwetlands within our study area We generated a circular buffer of1-km radius around each site a distance that encompasses amajority of the core terrestrial habitat used bymost anuran species(Semlitsch and Bodie 2003) and chose final study sites on thebasis of spatial independence (ie non-overlapping 1-km-radiuscircular buffers) After ground-truthing we determined that 21sites were sufficiently accessible for time-constrained activesearches (Fig 1 see Eskew et al (2012) and Hunt et al(2013) for more information on study site selection)

Data collection

During 2010 and 2011we surveyed each site usingMCS (Dorcaset al 2010) and time-constrained VES (Crump and Scott 1994)to document anurans During 2010 three anuran calling surveyswere conducted in both spring and summer at every site andduring 2011 three anuran calling surveys were conducted inwinter spring and summer at every site these seasonscorresponded to the peak breeding windows for species in ourstudy (Table 1) In total 15 MCS were conducted at each studysiteManual call surveyswere conductedbetween1845 hours and0100 hours by two experienced anuran surveyors listeningindependently for 5min and recording all species heard anydifferences in species observed were reconciled before leavingthe study site and questionable species identifications wereeliminated (Eskew et al 2012) In total there were fourexperienced observers participating in the study howeveronly two observers participated in each survey During eachMCS the number of cars passing by was recorded becausethis may interfere with the ability to detect frogs The majorityof sites were located in rural areas and had little to no road trafficand in cases with moderate traffic observers waited to begin thesurvey until all cars has passed Some survey sites were locatednearer to dams than others (see Eskew et al 2012) however thedistance was not great enough to influence observer ability todetect calling frogs (ie the noise level was always relativelylow) and we found that water releases were infrequent relativeto the timing of our surveys

B Wildlife Research J C Guzy et al

Fig 1 Location map depicting South Carolina USA Inset shows locations of each anuran study site along both the Pacolet (upper left fork) andBroad River in South Carolina USA county boundaries are delineated on the South Carolina outline and labelled on the study site inset

Table 1 Proportion of sampling units where each specieswas observed (ie naiumlve estimate) and the number of times eachspecies was detected for each survey method (ie visual encounter surveys (VES) and manual calling surveys (MCS)

In total 12 repeat surveys were conducted for each survey method

Species Common name Naiumlve estimate (no of detections) Breeding windowActive search Call survey

A americanus American toad 048 (16) 095 (40) February to AprilA fowleri Fowlerrsquos toad 090 (94) 100 (128) April to JulyA terrestris Southern toad 005 (1) 014 (4) Late February to MayA crepitans Northern cricket frog 057 (86) 071 (66) April to AugustH cinerea Green treefrog 038 (20) 076 (65) April to SeptemberH chrysoscelis Copersquos grey treefrog 090 (53) 095 (66) April to AugustP feriarum Upland chorus frog 043 (19) 090 (47) November to early AprilP crucifer Spring peeper 029 (11) 095 (102) November to AprilG carolinensis Eastern narrow-mouth toad 029 (18) 019 (4) April to OctoberL catesbeianus Bullfrog 057 (33) 067 (52) April to AugustL clamitans Green frog 067 (39) 086 (45) April to AugustL palustris Pickerel frog 043 (13) 033 (12) February to early AprilL sphenocephalus Southern leopard frog 081 (79) 062 (32) December to early AprilS holbrookii Eastern spadefoot 010 (3) 000 (0) Year round

Anuran detection probability Wildlife Research C

In addition time-constrained VES were conducted at eachsite During 2010 VES were conducted in summer (n= 3) andfall (n= 2) and during 2011 VES were conducted in spring(n = 3) summer (n= 3) and fall (n= 2) at each site we assumedthat we would detect focal species during these seasons whenactivities such as breeding ovipositing and foraging can beobserved (Dorcas and Gibbons 2008) In total 13 VES wereconducted at each study site Each survey consisted of twopersons independently searching the site for 30min to identifyanuran species A few searches were made by one person andthese lasted 1 h Each VES was conducted during daylight hours(ie 0900ndash1900 hours) and as much variation in microhabitatwas sampled as possible including surface cover objects (Voneshet al 2010) Search area was restricted to areas that we couldhear calling amphibians from during MCS surveys Anuransheard vocalising during VES were documented along withany visual observations of species All observed life-historystages were included during VES however the majority ofobservations encompassed adult and metamorphosingindividuals (as compared with observations of tadpoles) Forboth survey methods we recorded the number of days sincerainfall and included this as a sampling covariate

Data analyses

We used the model Nichols et al (2008) developed for multipledetection methods that allows simultaneous use of data fromall methods for inference about method-specific detectionprobabilities This modelling approach applies to any situationin which multiple detection methods are used in the samelocations and allows parameters to be modelled as functions ofsite or sample-specific covariates The general model comprisedone detection parameter pst which is the probability of detectionat occasion t by method s given the sample unit is occupied andthe species is present at the immediate sample station site andtwo occupancy parameters namelyy which is the probability aunit is occupied and qt which is the probability a species ispresent at the immediate sample site at occasion t given thesample unit is occupied The two occupancy parameters y andqt permit the modelling of occupancy at two different spatialscales however for the present study we focussed on theinfluence of survey method on detection probability and heldthe two occupancy parameters constant

For each species we used data from 12 VES conducted insummer fall and spring (ie peak activity period) paired withdata from 12 MCS conducted during summer winter and spring(ie peak breeding period) Because we conducted more callsurveys than active searches we eliminated call survey dates thatfell outside a particular speciesrsquo window of peak calling activity(Table 1) Peak anuran activity varies from species to speciesand we were able to fit each species into a general category ofspringndashsummerndashfall or fallndashwinterndashspring activity For examplethe northern cricket frog (Acris crepitans) was placed in thespringsummerfall activity window because it breeds fromApril to August (Table 1) and is active into the fall monthsThis exercise identified which three MCS surveys could beeliminated before data analysis for each species such thateach survey method would have the same amount of effort(ie 12 VES and 12 MCS)

To obtain estimates of the probability of detection for eachspecies for each method we used the Nichols et al (2008) modelwithin program PRESENCE 57 (Hines 2006) which estimatesparameters using maximum likelihood estimation methods Weevaluated the following eight models examining differentinfluences on detection probability of anurans (1) constant(null model with no covariates on detection probability)(2) survey method (3) day of year (lsquodatersquo) (4) number of dayssince rain (lsquorainrsquo) (5) date and rain (6) survey method and date(7) survey method and rain and (8) survey method plus dateand rain

To obtain the most accurate parameter estimation allcovariates were standardised by calculating z-scores (ie themean was subtracted from each value and then divided by thesd) before analysis Model selection was based on Akaikeinformation criteria (AIC Burnham and Anderson 2002) Weused AIC values adjusted for small samples sizes (ie AICc)and assessed fit for each model set by using the MacKenziendashBailey goodness-of-fit test (MacKenzie and Bailey 2004) Weconducted the test for 1000 bootstrap iterations on the mostparameterised models in each model set to generate estimatesof the overdispersion factor c and used the c value to ensure aconservative estimation of goodness-of-fit (ie if cgt 1 we usedQAICc values adjusted for overdispersion Burnham andAnderson 2002) We examined Akaike weights to determinethe strength of evidence for each model (Burnham and Anderson2002) We assumed models with higher weights and lowerAIC values were better able to explain variation in data andselected the models with substantial empirical support whichincludedmodelswithinD2AICc (Burnham andAnderson 2002)In cases where there was no clear lsquobestrsquo model we computedmodel-averaged estimates (Buckland et al 1997) for parametersof interest We define lsquobestrsquo to be a model within D2 AICcwhich has gt50 of the weight and where the remaining 50of the weight is spread relatively thinly among the remainingcandidate models that are also within D2 AICc Species-specificdetection probabilities were derived using the inverse logit-transformation (ie (exp (a)(1+ exp a)) of parameterestimates We estimated lower and upper 95 confidenceintervals using the delta method (Ver Hoef 2012) Because itwas impractical to summarise AIC results for eight candidatemodel sets for each species we present the top models for eachspecies

Results

We detected the following 14 anuran species American toad(Anaxyrus americanus) Fowlerrsquos toad (Anaxyrus fowleri)southern toad (Anaxyrus terrestris) northern cricket frog(A crepitans) green treefrog (Hyla cinerea) Copersquos greytreefrog (Hyla chrysoscelis) upland chorus frog (Pseudacrisferiarum) spring peeper (Pseudacris crucifer) eastern narrow-mouthed toad (Gastrophryne carolinensis) American bullfrog(Lithobates catesbeianus) green frog (Lithobates clamitans)pickerel frog (Lithobates palustris) southern leopard frog(Lithobates sphenocephalus) and eastern spadefoot(Scaphiopus holbrookii) The naiumlve occupancy and detectionestimate varied among species for both MCS and VES(Table 1) Because S holbrookii detections were so limited

D Wildlife Research J C Guzy et al

Tab

le2

Best-supp

ortedmod

elsfordetectionof

anuran

speciesba

sedon

QAIC

cselectioncriteria

The

sameeightcandidatemodelswereconstructedforeachspeciesfrom

uncorrelated

covariatesB

est-supportedmodelsareinboldA

ICA

kaikeinform

ationcriteriaDQ

AICcdifference

inQAICcrelativ

eto

thetopmod

elwmod

elweightKnu

mberof

parametersin

themod

elMetho

danuran

survey

method(V

ESor

MCS)Day

ofyeard

ayof

year

each

survey

was

conducted

Dayssincerainnu

mber

ofdays

sincerain

foragivensurvey

Species

Model

AICc

DQAICc

wK

ndash2log-lik

elihoo

dParam

eter

estim

ate(se)

Metho

dDay

ofyear

Num

berof

days

sincerain

Y-intercept

p-intercept

p-VES

p-MCS

p-VES

p-MCS

p-VES

p-MCS

Aa

mericanus

Y()Aq

()Ap

B(D

ate)

32364

000

057

431314

494

(402)

ndash287

(023)

ndashndash

-013

(009)

ndash138

(022)

ndashndash

Y()q()p(DateRain)

32554

190

022

630

754

487

(400)

ndash297

(027)

ndashndash

ndash009

(009)

ndash147

(024)

ndash064

(031)

019

(017)

Y()q()p(Method

Date)

32599

235

018

531199

392

(254)

ndashndash270

(026)

ndash323

(042)

ndash007

(008)

ndash164

(034)

ndashndash

Modelaveraged

459

(362)

ndash229

(019)

ndash049

(005)

ndash058

(008)

ndash011

(009)

ndash141

(024)

ndash014

(007)

004

(004)

Afow

leri

C()q()p(DateRain)

64863

000

073

663

063

212

(050)

ndash026

(015)

ndashndash

ndash051

(017)

124

(028)

006

(004)

084

(019)

Y()q()p(Method

DateRain)

65058

195

027

762796

206

(048)

ndashndash039

(017)

ndash002

(022)

ndash056

(017)

110

(030)

004

(015)

086

(020)

Aterrestris

Y()q()p()

5747

000

044

252

80

ndash095

(073)

ndash187

(119)

ndashndash

ndashndash

ndashndash

Y()q()p(Method)

5798

051

034

35057

ndash107

(072)

ndashndash271

(144)

ndash130

(128)

ndashndash

ndashndash

Modelaveraged

ndash078

(057)

ndash082

(052)

ndash092

(049)

ndash044

(043)

Acrepitans

Y()q()p(Method

Date)

53038

000

055

551

638

121

(029)

ndash039

(023)

ndash049

(024)

010

(008)

073

(027)

ndashndash

Y()q()p(Method)

53149

111

031

352408

108

(028)

ndash041

(024)

ndash015

(011)

ndashndash

ndashndash

Modelaveraged

100

(025)

ndash034

(020)

ndash032

(017)

006

(004)

040

(015)

ndashndash

Hcinerea

Y()q()p(Method

Date)

35484

000

054

534084

070

(041)

ndashndash181

(033)

ndash098

(045)

ndash051

(032)

292

(113)

ndashndash

Y()q()p(Date)

35521

037

045

434

471

087

(039)

ndash157

(026)

ndashndash

ndash037

(027)

296

(070)

ndashndash

Modelaveraged

077

(040)

ndash071

(012)

ndash098

(018)

-053

(024)

-044

(029)

290

(093)

ndashndash

Hchrysoscelis

Y()q()p(Date)

52682

000

058

451

632

202

(106)

ndash125

(021)

ndashndash

ndash092

(027)

081

(025)

ndashndash

Y()q()p(Method

Date)

52900

218

019

551100

200

(108)

ndashndash141

(025)

ndash109

(025)

ndash104

(032)

069

(027)

(contin

uednextpa

ge)

Anuran detection probability Wildlife Research E

Tab

le2

(con

tinued)

Species

Model

AICc

DQAICc

wK

ndash2log-lik

elihoo

dParam

eter

estim

ate(se)

Metho

dDay

ofyear

Num

berof

days

sincerain

Y-intercept

p-intercept

p-VES

p-MCS

p-VES

p-MCS

p-VES

p-MCS

Y()q()p(DateRain)

52902

220

019

651

502

205

(104)

ndash128

(020)

ndashndash

ndash093

(028)

082

(026)

ndash025

(017)

029

(016)

Modelaveraged

194

(102)

ndash097

(016)

ndash027

(005)

ndash021

(005)

ndash091

(027)

076

(025)

ndash005

(004)

006

(003)

Pferiarum

C()q()p(Metho

dDate)

30614

000

097

529

214

102

(032)

ndashndash251

(024)

ndash201

(066)

035

(024)

ndash333

(050)

ndashndash

Pcrucifer

C()q()p(Metho

dDate)

41265

000

085

539

865

087

(033)

ndashndash260

(034)

ndash120

(036)

046

(032)

ndash273

(100)

ndashndash

Y()q()p(Date)

41641

376

013

440

591

071

(020)

ndash211

(024)

ndashndash

049

(028)

ndash495

(105)

ndashndash

Gcarolinensis

C()q()p(Metho

d)15

168

000

093

313

768

ndash041

(030)

ndashndash056

(041)

ndash237

(061)

ndashndash

ndashndash

Lcatesbeianus

Y()q()p(Date)

43926

000

044

442

876

153

(052)

ndash128

(019)

ndashndash

-019

(015)

074

(025)

ndashndash

Y()q()p(Method)

44010

084

029

343269

158

(052)

ndashndash143

(023)

ndash082

(021)

ndashndash

ndashndash

Y()q()p(Method

Date)

44118

192

017

542718

153

(052)

ndashndash144

(023)

ndash107

(025)

ndash015

(021)

058

(027)

ndashndash

Modelaveraged

139

(047)

ndash056

(008)

ndash066

(011)

ndash042

(010)

ndash011

(010)

042

(016)

Lclamita

nsC()q()p(Date)

44153

000

058

443

103

178

(061)

ndash144

(019)

ndashndash

ndash029

(021)

072

(024)

ndashndash

Y()q()p(DateRain)

44376

223

019

642

576

179

(060)

ndash147

(019)

ndashndash

ndash031

(021)

074

(024)

042

(018)

003

(018)

Lp

alustris

C()p()

20071

000

065

219

604

101

(067)

ndash226

(041)

ndashndash

ndashndash

ndashndash

Y()p(Metho

d)20

341

270

017

319

600

101

(070)

ndash222

(046)

ndash231

(046)

ndashndash

ndashndash

Lsph

enocepha

lus

C()q()p(Metho

dDate)

48414

000

078

547

014

163

(056)

ndashndash025

(022)

ndash249

(041)

ndash008

(006)

ndash110

(036)

ndashndash

Y()q()p(Method)

48804

390

011

348063

166

(056)

ndashndash026

(022)

ndash154

(023)

ndashndash

ndashndash

AConstantprobability

ofoccupancy(ieY

()q())

BProbabilityof

detection

which

varies

bymodelforeach

specieswith

orwith

outinclusionof

covariates

thatim

provedetectionestim

ates

F Wildlife Research J C Guzy et al

method-specific detection probabilities for this species were notpossible to estimate

The best supported model for 10 anuran species indicatedthat detection varied by survey method (ie VES or MCSTable 2 Appendix 1) for four of these species (P feriarumP crucifer G carolinensis and L sphenocephalus) detectionvaried by survey method and for six species the best-supportedmodel was determined from model averaging (A americanusA terrestris A crepitans H cinerea H chrysoscelis andL catesbeianus) For the remaining three species (A fowleriL palustris and L clamitans) survey method was unimportantand the best-supported model suggested either a constantprobability of detection (L palustris) or suggested thatdetection varied by date (L clamitans) or date and days sincerainfall (A fowleri Table 2)

For six species MCS resulted in a higher mean (se)probability of detection than did VES (A terrestris MCS 039(038) vs VES 028 015) H cinerea MCS 037 (017) vs VES027 (005)H chrysoscelisMCS045 (011) vsVES043 (008)P feriarum MCS 012 (004) vs VES 008 (001) P cruciferMCS 023 (007) vs VES 007 (001) and L catesbeianus MCS040 (009) vs VES 034 (006) Fig 2) Conversely for fourspecies VES resulted in a higher mean probability of detectionthan did MCS (A americanus VES 038 (004) vs MCS 036(005) A crepitans VES 060 (036) vs MCS 038 (018)G carolinensis VES 036 (027) vs MCS 009 (002) andL sphenocephalus VES 044 (039) vs MCS 008 (001)Fig 2) For three species however both methods were equallyeffective and detection probabilities were identical (A fowleri044 (025) L clamitans 046 (008) and L palustris 009(002) Fig 2)

In addition to survey method date was an important factorinfluencing detection probability and was included in thebest-supported model for 10 species (A americanusA fowleri A crepitans H cinerea H chrysoscelisP feriarum P crucifer L catesbeianus L clamitans andL sphenocephalus Table 2) For all summer-breeding frogsexcept A crepitans detection probability during the VESdecreased with increasing day of year (ie decreased fromspring to fall) which corresponded to between 12 April and18 November each year (Figs 3 4) Specifically from mid-April to mid-November of each year the estimated probabilityof detection during VES increased from 058 to 060 forA crepitans and decreased from 040 to 030 forL catesbeianus from 078 to 032 for A fowleri from 045 to019 for H cinerea from 073 to 048 for L clamitans and from080 to 017 for H chrysoscelis (Figs 3 4) For all summer-breeding frogs detection probability during the MCS callingwindow increased with increasing day of year whichcorresponds to between 9 April and 28 June each year (Figs 34) Specifically from early April to late June of each year theestimated probability of detection during MCS increasedfrom 025 to 060 for A crepitans from 022 to 060 forL catesbeianus from 021 to 095 for A fowleri from 0002to099 forH cinerea from022 to 080 forL clamitans and from020 to 078 for H chrysoscelis (Figs 3 4)

For all winter-breeding frogs detection probability duringthe active-search survey window varied slightly withincreasing day of year (Fig 5) the estimated probability ofdetection during VES decreased from 050 to 040 forL sphenocephalus and from 041 to 031 for A americanusand increased from 005 to 018 for P feriarum and from 003 to

100

090

080

070

060

050

040

030

020

010

Est

imat

ed p

roba

bilit

y of

det

ectio

n

0

A americanus

A fowleri

A terre

stris

A crepitans

H cinerea

H chrysoscelis

P feriarum

P crucifer

G carolinensis

L catesbeianus

L clamitans

L palustris

L sphenocephalus

Fig 2 Estimated conditional probability of detection (1 se) for each anuran species along the Broad and PacoletRivers South Carolina USA Estimates are based on repeat surveys using two different survey methods (MCS or VES)

Anuran detection probability Wildlife Research G

018 for P crucifer (Fig 5) For all winter-breeding frogsdetection probability during the calling-survey time spandecreased with increasing day of year (Fig 5) Specifically

from mid-February to mid-May of each year the estimatedprobability of detection during MCS decreased from 038 to002 for L sphenocephalus from 098 to 001 for P feriarum

10MCS

Est

imat

ed p

roba

bilit

y of

det

ectio

n

VES

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

9Apr

29Apr

19May

8Jun

28Jun

12Apr

1Jun

21Jul

19Sep

18Nov

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

Fig 3 Effect of date by survey method (MCS or VES) on conditional detection probability ofsummer-breeding anuran species along the Broad and Pacolet Rivers South Carolina USA Solid linesrepresent the relationship between date and detection probability and dashed lines are the 95 CIsfor estimates of the covariate effect

H Wildlife Research J C Guzy et al

from 089 to 005 for A americanus and from 098 to 001 forP crucifer (Fig 5)

Discussion

Our study has provided insight for monitoring anurans byimplementing a modelling framework where data from bothMCS and VES are evaluated simultaneously resulting inmethod-specific detection probabilities Our results providedstrong evidence of variation in anuran detection probabilitiesamong two different detection methods Specifically for 10 of13 species the best-supported model indicated that detectionvaried by survey method Consistent with previous studies (egGooch et al 2006 Steelman and Dorcas 2010) calling surveyswere effective at detecting a wide range of anurans Visualencounter surveys however were also effective at detecting awide range of anuran species and several species had a higherdetection probability with VES thanwithMCSVisual encountersurveys are likely to be under-used as a monitoring tool foranurans perhaps because of a lack of information on theirutility Instead VES are often used along with other methodsin surveys for focal or rare species where the intent is gearedtoward complete coverage of all life stages to increase detectionprobability (eg Drost and Fellers 1996 Burbrink et al 1998Lips 1999 Lehtinen and Galatowitsch 2001)

Although both MCS and VES are effective methods forcertain species in our study one method was most effective

Specifically VES represent a more effective method fordetecting A crepitans G carolinensis and L sphenocephaluswhich is perhaps reflective of variation in species life historyFor example G carolinensis and L sphenocephalus like mostanurans can often be observed via vocalisations made duringbreeding however the breeding windows are wide (particularlyfor L sphenocephalus) calling by both species is sporadic andfor G carolinensis calling is usually initiated only by heavyrains (Brandt 1936 Dodd 2013) In addition calls of bothG carolinensis and L sphenocephalus are easily masked bylouder species (eg H cinerea and P crucifer Conant andCollins 1998) Likely because both G carolinensis andL sphenocephalus are active during the non-breeding periodin a wide variety of aquatic and terrestrial habitats (Mitchelland Lannoo 2005 Dodd 2013) VES was a more reliable methodof detection Acris crepitans another species with higherdetection probabilities during VES is active and conspicuousduring the non-breeding season Several other anuran speciesmight also be more reliably detected using VES if theirvocalisations are difficult to observe and their non-breedinghabits permit detection (ie little grass frogs (Pseudacrisocularis) or species that exhibit explosive andor short-livedbreeding periods (eg Scaphiopus spp L sylvaticusL capito) In these cases precisely timing MCS during thisbrief window is challenging (Dostine et al 2013) and oftennot ideal because anuran call surveys are less effective duringheavy rain (Weir et al 2005)

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

9Apr

29Apr

19May

8Jun

28Jun

12Apr

1Jun

21Jul

19Sep

18Nov

MCS VES

Est

imat

ed p

roba

bilit

y of

det

ectio

n

Fig 4 Effect of date by survey method (MCS or VES) on conditional detection probability of summer-breeding anuran species along the Broad and Pacolet Rivers South Carolina USA Solid lines represent therelationship between date and detection probability and dashed lines are the 95 CIs for estimates of thecovariate effect

Anuran detection probability Wildlife Research I

We found that MCS represented a more effectivedetection method for H cinerea P feriarum P crucifer andL catesbeianus These species are common vocal and calls can

be heard from a considerable distance (Dorcas et al 2007) Fora few species however both survey methods were equallyeffective (ie for A fowleri L clamitans and L palustris) In

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

14Feb

10Mar

4Apr

24Apr

14May

12Apr

1Jun

21Jul

19Sep

18Nov

Est

imat

ed p

roba

bilit

y of

det

ectio

n

MCS VES

Fig 5 Effect of date by survey method (MCS or VES) on conditional detection probability of winter-breeding anuran species along the Broad and Pacolet Rivers South Carolina USA Solid lines representthe relationship between date and detection probability and dashed lines are the 95 CIs for estimates ofthe covariate effect

J Wildlife Research J C Guzy et al

addition given the marginally larger detection probabilities ofA americanus using VES and of H chrysoscelis using MCSfor these species we also consider both survey methods to besimilarly effective For both A americanus and A fowleriMCS and VES detection probabilities were very similar MCSare effective because males of each species produce loud easilyrecognisable calls (Dorcas et al 2007) but VES are also effectiveperhaps because of their conspicuous tadpoles which school inlarge groups (Lefcort 1998) and the tendency for large numbersof metamorphs to forage for several weeks around the edge ofthe aquatic habitat fromwhich they emerged (Beck and Congdon1999 Dodd 2013) Likewise VES had detection probabilitiessimilar to those of MCS for L clamitans L palustris andH chrysoscelis for L clamitans this may be because whenstartled they emit conspicuous alarm calls as they leap into thewater (Dorcas et al 2007) and L palustrismay call in any monthof the year in the southern United States (Conant and Collins1998) Finally for H chrysoscelis its tendency to callsporadically during the daytime likely accounts for thesimilarity in effectiveness of both survey methods

Datewas an important factor influencing detection probabilityof anurans particularly regarding call surveys For summer-breeding anurans detection probability steadily increasedtoward the end of the summer survey windows and likewisefor winter-breeding anurans detection probability was highesttoward the beginning of the winter survey windows Previousstudies have documented strong patterns in seasonal breedingactivity and breeding windows are well established (Droege andEagle 2005) However for VES the potential influence of dateto increase or decrease the effectiveness of this survey methodis not well known Our results indicated that for VES for mostspecies detection probability was relatively constant throughoutthe sampling window (ie AprilndashNovember each year) althoughfor a few species (A fowleri H chrysoscelis H cinerea andL clamitans) detection probability steadily decreased into thefall months Furthermore although we showed that VES arebetter for some species we also highlighted that VES canprovide adequate occupancy estimation even when conductedoutside the breeding window For example for L clamitans andA fowleri VES conducted in early April resulted in a 60probability of detection and during MCS it was not until lateJune toward the end of the breeding season that detectionprobabilities reached 60 Similarly for H chrysoscelis VESconducted in early April resulted in an 80 probability ofdetection and for A crepitans detection probabilities remained~60 during the VES survey window for both of thesespecies detection probabilities comparable to 80 and 60respectively did not occur until late June Therefore we suggestthat spring and fall VES can detect species that vocalise mostlyduring the summer months

Management implications

The goals of a particular study should dictate the appropriatesurvey method however results of the present study highlightedthe importance of estimating method-specific detectionprobabilities which should prove especially useful in thedesign of multispecies surveys We recommend the use MCSfor anurans with broad calling windows and conspicuous

vocalisations and VES for anurans that either vocalisesporadically or discreetly Furthermore we recommend the useof VES throughout the activity season not just the breedingseason because we found that survey date was a largelyunimportant predictor of detection using VES Our results alsosuggested that VES can provide more flexibility for anuranresearchers because robust estimates of detection andoccupancy can be obtained outside a narrow breeding window

Acknowledgements

This research was supported with funding provided by the Broad RiverMitigation Trust Fund administered through the South CarolinaDepartment of Natural Resources Additional funding was provided by theDukeEndowment through theDavidsonResearch Initiative and theDavidsonCollege Department of Biology We thank E Eskew S Hunt L WitczakC Williams A Domske D Millican C Oldham R Bauer B Abbuhl andC Ruder for assistance in the field and J Micancin for his help in identifyingcricket frog vocal recordings For assistance with logistics related to theprojectwe thankGVaughanSBennettBPerry andVVejdaniFinallywethank D Ramsey and two anonymous reviewers for statistical assistanceand comments which greatly improved this manuscript

References

Adams M J Pearl C A Galvan S andMcCreary B (2011) Non-nativespecies impacts on pond occupancy by an anuran The Journal of WildlifeManagement 75 30ndash35 doi101002jwmg29

Bailey L L Simons T R and Pollock K H (2004) Estimating siteoccupancy and species detection probability parameters for terrestrialsalamandersEcologicalApplications14 692ndash702 doi10189003-5012

BalasC JEulissNH andMushetDM (2012) Influenceof conservationprograms on amphibians using seasonal wetlands in the prairie potholeregion Wetlands 32 333ndash345 doi101007s13157-012-0269-9

Beck CW andCongdon J D (1999) Effects of individual variation in ageand size at metamorphosis on growth and survivorship of southern toad(Bufo terrestris)metamorphsCanadian Journal of Zoology77 944ndash951doi101139cjz-77-6-944

Blair W F (1961) Calling and spawning seasons in a mixed population ofanurans Ecology 42 99ndash110 doi1023071933272

Brander S M Royle J A and Eames M (2007) Evaluation of the statusof anurans on a refuge in suburbanMaryland Journal of Herpetology 4152ndash60 doi1016700022-1511(2007)41[52EOTSOA]20CO2

Brandt B B (1936) The frogs and toads of eastern North Carolina Copeia1936 215ndash223 doi1023071436326

Brown G W Scroggie M P Smith M J and Steane D (2007) Anevaluation of methods for assessing the population status of thethreatened Alpine tree frog Litoria verreauxii alpina in southeasternAustralia Copeia 3 765ndash770 doi1016430045-8511(2007)2007[765AEOMFA]20CO2

BucklandSTBurnhamKP andAugustinNH (1997)Model selectionan integral part of inference Biometrics 53 603ndash618 doi1023072533961

Burbrink F T Phillips C A and Heske E J (1998) A riparian zone insouthern Illinois as a potential dispersal corridor for reptiles andamphibians Biological Conservation 86 107ndash115 doi101016S0006-3207(98)00054-8

Burnham K P and Anderson D R (2002) lsquoModel Selection andMultimodel Inference a Practical Information-theoretic Approachrsquo2nd edn (Springer-Verlag New York)

Conant R and Collins J T (1998) lsquoA Field Guide to Amphibians andReptiles Eastern andCentralNorthAmericarsquo 3rd edn (HoughtonMifflinCompany Boston MA)

CookR P Tupper TA Paton PWC andTimmBC (2011) Effects oftemperature and temporal factors on anuran detection probabilities at

Anuran detection probability Wildlife Research K

In addition time-constrained VES were conducted at eachsite During 2010 VES were conducted in summer (n= 3) andfall (n= 2) and during 2011 VES were conducted in spring(n = 3) summer (n= 3) and fall (n= 2) at each site we assumedthat we would detect focal species during these seasons whenactivities such as breeding ovipositing and foraging can beobserved (Dorcas and Gibbons 2008) In total 13 VES wereconducted at each study site Each survey consisted of twopersons independently searching the site for 30min to identifyanuran species A few searches were made by one person andthese lasted 1 h Each VES was conducted during daylight hours(ie 0900ndash1900 hours) and as much variation in microhabitatwas sampled as possible including surface cover objects (Voneshet al 2010) Search area was restricted to areas that we couldhear calling amphibians from during MCS surveys Anuransheard vocalising during VES were documented along withany visual observations of species All observed life-historystages were included during VES however the majority ofobservations encompassed adult and metamorphosingindividuals (as compared with observations of tadpoles) Forboth survey methods we recorded the number of days sincerainfall and included this as a sampling covariate

Data analyses

We used the model Nichols et al (2008) developed for multipledetection methods that allows simultaneous use of data fromall methods for inference about method-specific detectionprobabilities This modelling approach applies to any situationin which multiple detection methods are used in the samelocations and allows parameters to be modelled as functions ofsite or sample-specific covariates The general model comprisedone detection parameter pst which is the probability of detectionat occasion t by method s given the sample unit is occupied andthe species is present at the immediate sample station site andtwo occupancy parameters namelyy which is the probability aunit is occupied and qt which is the probability a species ispresent at the immediate sample site at occasion t given thesample unit is occupied The two occupancy parameters y andqt permit the modelling of occupancy at two different spatialscales however for the present study we focussed on theinfluence of survey method on detection probability and heldthe two occupancy parameters constant

For each species we used data from 12 VES conducted insummer fall and spring (ie peak activity period) paired withdata from 12 MCS conducted during summer winter and spring(ie peak breeding period) Because we conducted more callsurveys than active searches we eliminated call survey dates thatfell outside a particular speciesrsquo window of peak calling activity(Table 1) Peak anuran activity varies from species to speciesand we were able to fit each species into a general category ofspringndashsummerndashfall or fallndashwinterndashspring activity For examplethe northern cricket frog (Acris crepitans) was placed in thespringsummerfall activity window because it breeds fromApril to August (Table 1) and is active into the fall monthsThis exercise identified which three MCS surveys could beeliminated before data analysis for each species such thateach survey method would have the same amount of effort(ie 12 VES and 12 MCS)

To obtain estimates of the probability of detection for eachspecies for each method we used the Nichols et al (2008) modelwithin program PRESENCE 57 (Hines 2006) which estimatesparameters using maximum likelihood estimation methods Weevaluated the following eight models examining differentinfluences on detection probability of anurans (1) constant(null model with no covariates on detection probability)(2) survey method (3) day of year (lsquodatersquo) (4) number of dayssince rain (lsquorainrsquo) (5) date and rain (6) survey method and date(7) survey method and rain and (8) survey method plus dateand rain

To obtain the most accurate parameter estimation allcovariates were standardised by calculating z-scores (ie themean was subtracted from each value and then divided by thesd) before analysis Model selection was based on Akaikeinformation criteria (AIC Burnham and Anderson 2002) Weused AIC values adjusted for small samples sizes (ie AICc)and assessed fit for each model set by using the MacKenziendashBailey goodness-of-fit test (MacKenzie and Bailey 2004) Weconducted the test for 1000 bootstrap iterations on the mostparameterised models in each model set to generate estimatesof the overdispersion factor c and used the c value to ensure aconservative estimation of goodness-of-fit (ie if cgt 1 we usedQAICc values adjusted for overdispersion Burnham andAnderson 2002) We examined Akaike weights to determinethe strength of evidence for each model (Burnham and Anderson2002) We assumed models with higher weights and lowerAIC values were better able to explain variation in data andselected the models with substantial empirical support whichincludedmodelswithinD2AICc (Burnham andAnderson 2002)In cases where there was no clear lsquobestrsquo model we computedmodel-averaged estimates (Buckland et al 1997) for parametersof interest We define lsquobestrsquo to be a model within D2 AICcwhich has gt50 of the weight and where the remaining 50of the weight is spread relatively thinly among the remainingcandidate models that are also within D2 AICc Species-specificdetection probabilities were derived using the inverse logit-transformation (ie (exp (a)(1+ exp a)) of parameterestimates We estimated lower and upper 95 confidenceintervals using the delta method (Ver Hoef 2012) Because itwas impractical to summarise AIC results for eight candidatemodel sets for each species we present the top models for eachspecies

Results

We detected the following 14 anuran species American toad(Anaxyrus americanus) Fowlerrsquos toad (Anaxyrus fowleri)southern toad (Anaxyrus terrestris) northern cricket frog(A crepitans) green treefrog (Hyla cinerea) Copersquos greytreefrog (Hyla chrysoscelis) upland chorus frog (Pseudacrisferiarum) spring peeper (Pseudacris crucifer) eastern narrow-mouthed toad (Gastrophryne carolinensis) American bullfrog(Lithobates catesbeianus) green frog (Lithobates clamitans)pickerel frog (Lithobates palustris) southern leopard frog(Lithobates sphenocephalus) and eastern spadefoot(Scaphiopus holbrookii) The naiumlve occupancy and detectionestimate varied among species for both MCS and VES(Table 1) Because S holbrookii detections were so limited

D Wildlife Research J C Guzy et al

Tab

le2

Best-supp

ortedmod

elsfordetectionof

anuran

speciesba

sedon

QAIC

cselectioncriteria

The

sameeightcandidatemodelswereconstructedforeachspeciesfrom

uncorrelated

covariatesB

est-supportedmodelsareinboldA

ICA

kaikeinform

ationcriteriaDQ

AICcdifference

inQAICcrelativ

eto

thetopmod

elwmod

elweightKnu

mberof

parametersin

themod

elMetho

danuran

survey

method(V

ESor

MCS)Day

ofyeard

ayof

year

each

survey

was

conducted

Dayssincerainnu

mber

ofdays

sincerain

foragivensurvey

Species

Model

AICc

DQAICc

wK

ndash2log-lik

elihoo

dParam

eter

estim

ate(se)

Metho

dDay

ofyear

Num

berof

days

sincerain

Y-intercept

p-intercept

p-VES

p-MCS

p-VES

p-MCS

p-VES

p-MCS

Aa

mericanus

Y()Aq

()Ap

B(D

ate)

32364

000

057

431314

494

(402)

ndash287

(023)

ndashndash

-013

(009)

ndash138

(022)

ndashndash

Y()q()p(DateRain)

32554

190

022

630

754

487

(400)

ndash297

(027)

ndashndash

ndash009

(009)

ndash147

(024)

ndash064

(031)

019

(017)

Y()q()p(Method

Date)

32599

235

018

531199

392

(254)

ndashndash270

(026)

ndash323

(042)

ndash007

(008)

ndash164

(034)

ndashndash

Modelaveraged

459

(362)

ndash229

(019)

ndash049

(005)

ndash058

(008)

ndash011

(009)

ndash141

(024)

ndash014

(007)

004

(004)

Afow

leri

C()q()p(DateRain)

64863

000

073

663

063

212

(050)

ndash026

(015)

ndashndash

ndash051

(017)

124

(028)

006

(004)

084

(019)

Y()q()p(Method

DateRain)

65058

195

027

762796

206

(048)

ndashndash039

(017)

ndash002

(022)

ndash056

(017)

110

(030)

004

(015)

086

(020)

Aterrestris

Y()q()p()

5747

000

044

252

80

ndash095

(073)

ndash187

(119)

ndashndash

ndashndash

ndashndash

Y()q()p(Method)

5798

051

034

35057

ndash107

(072)

ndashndash271

(144)

ndash130

(128)

ndashndash

ndashndash

Modelaveraged

ndash078

(057)

ndash082

(052)

ndash092

(049)

ndash044

(043)

Acrepitans

Y()q()p(Method

Date)

53038

000

055

551

638

121

(029)

ndash039

(023)

ndash049

(024)

010

(008)

073

(027)

ndashndash

Y()q()p(Method)

53149

111

031

352408

108

(028)

ndash041

(024)

ndash015

(011)

ndashndash

ndashndash

Modelaveraged

100

(025)

ndash034

(020)

ndash032

(017)

006

(004)

040

(015)

ndashndash

Hcinerea

Y()q()p(Method

Date)

35484

000

054

534084

070

(041)

ndashndash181

(033)

ndash098

(045)

ndash051

(032)

292

(113)

ndashndash

Y()q()p(Date)

35521

037

045

434

471

087

(039)

ndash157

(026)

ndashndash

ndash037

(027)

296

(070)

ndashndash

Modelaveraged

077

(040)

ndash071

(012)

ndash098

(018)

-053

(024)

-044

(029)

290

(093)

ndashndash

Hchrysoscelis

Y()q()p(Date)

52682

000

058

451

632

202

(106)

ndash125

(021)

ndashndash

ndash092

(027)

081

(025)

ndashndash

Y()q()p(Method

Date)

52900

218

019

551100

200

(108)

ndashndash141

(025)

ndash109

(025)

ndash104

(032)

069

(027)

(contin

uednextpa

ge)

Anuran detection probability Wildlife Research E

Tab

le2

(con

tinued)

Species

Model

AICc

DQAICc

wK

ndash2log-lik

elihoo

dParam

eter

estim

ate(se)

Metho

dDay

ofyear

Num

berof

days

sincerain

Y-intercept

p-intercept

p-VES

p-MCS

p-VES

p-MCS

p-VES

p-MCS

Y()q()p(DateRain)

52902

220

019

651

502

205

(104)

ndash128

(020)

ndashndash

ndash093

(028)

082

(026)

ndash025

(017)

029

(016)

Modelaveraged

194

(102)

ndash097

(016)

ndash027

(005)

ndash021

(005)

ndash091

(027)

076

(025)

ndash005

(004)

006

(003)

Pferiarum

C()q()p(Metho

dDate)

30614

000

097

529

214

102

(032)

ndashndash251

(024)

ndash201

(066)

035

(024)

ndash333

(050)

ndashndash

Pcrucifer

C()q()p(Metho

dDate)

41265

000

085

539

865

087

(033)

ndashndash260

(034)

ndash120

(036)

046

(032)

ndash273

(100)

ndashndash

Y()q()p(Date)

41641

376

013

440

591

071

(020)

ndash211

(024)

ndashndash

049

(028)

ndash495

(105)

ndashndash

Gcarolinensis

C()q()p(Metho

d)15

168

000

093

313

768

ndash041

(030)

ndashndash056

(041)

ndash237

(061)

ndashndash

ndashndash

Lcatesbeianus

Y()q()p(Date)

43926

000

044

442

876

153

(052)

ndash128

(019)

ndashndash

-019

(015)

074

(025)

ndashndash

Y()q()p(Method)

44010

084

029

343269

158

(052)

ndashndash143

(023)

ndash082

(021)

ndashndash

ndashndash

Y()q()p(Method

Date)

44118

192

017

542718

153

(052)

ndashndash144

(023)

ndash107

(025)

ndash015

(021)

058

(027)

ndashndash

Modelaveraged

139

(047)

ndash056

(008)

ndash066

(011)

ndash042

(010)

ndash011

(010)

042

(016)

Lclamita

nsC()q()p(Date)

44153

000

058

443

103

178

(061)

ndash144

(019)

ndashndash

ndash029

(021)

072

(024)

ndashndash

Y()q()p(DateRain)

44376

223

019

642

576

179

(060)

ndash147

(019)

ndashndash

ndash031

(021)

074

(024)

042

(018)

003

(018)

Lp

alustris

C()p()

20071

000

065

219

604

101

(067)

ndash226

(041)

ndashndash

ndashndash

ndashndash

Y()p(Metho

d)20

341

270

017

319

600

101

(070)

ndash222

(046)

ndash231

(046)

ndashndash

ndashndash

Lsph

enocepha

lus

C()q()p(Metho

dDate)

48414

000

078

547

014

163

(056)

ndashndash025

(022)

ndash249

(041)

ndash008

(006)

ndash110

(036)

ndashndash

Y()q()p(Method)

48804

390

011

348063

166

(056)

ndashndash026

(022)

ndash154

(023)

ndashndash

ndashndash

AConstantprobability

ofoccupancy(ieY

()q())

BProbabilityof

detection

which

varies

bymodelforeach

specieswith

orwith

outinclusionof

covariates

thatim

provedetectionestim

ates

F Wildlife Research J C Guzy et al

method-specific detection probabilities for this species were notpossible to estimate

The best supported model for 10 anuran species indicatedthat detection varied by survey method (ie VES or MCSTable 2 Appendix 1) for four of these species (P feriarumP crucifer G carolinensis and L sphenocephalus) detectionvaried by survey method and for six species the best-supportedmodel was determined from model averaging (A americanusA terrestris A crepitans H cinerea H chrysoscelis andL catesbeianus) For the remaining three species (A fowleriL palustris and L clamitans) survey method was unimportantand the best-supported model suggested either a constantprobability of detection (L palustris) or suggested thatdetection varied by date (L clamitans) or date and days sincerainfall (A fowleri Table 2)

For six species MCS resulted in a higher mean (se)probability of detection than did VES (A terrestris MCS 039(038) vs VES 028 015) H cinerea MCS 037 (017) vs VES027 (005)H chrysoscelisMCS045 (011) vsVES043 (008)P feriarum MCS 012 (004) vs VES 008 (001) P cruciferMCS 023 (007) vs VES 007 (001) and L catesbeianus MCS040 (009) vs VES 034 (006) Fig 2) Conversely for fourspecies VES resulted in a higher mean probability of detectionthan did MCS (A americanus VES 038 (004) vs MCS 036(005) A crepitans VES 060 (036) vs MCS 038 (018)G carolinensis VES 036 (027) vs MCS 009 (002) andL sphenocephalus VES 044 (039) vs MCS 008 (001)Fig 2) For three species however both methods were equallyeffective and detection probabilities were identical (A fowleri044 (025) L clamitans 046 (008) and L palustris 009(002) Fig 2)

In addition to survey method date was an important factorinfluencing detection probability and was included in thebest-supported model for 10 species (A americanusA fowleri A crepitans H cinerea H chrysoscelisP feriarum P crucifer L catesbeianus L clamitans andL sphenocephalus Table 2) For all summer-breeding frogsexcept A crepitans detection probability during the VESdecreased with increasing day of year (ie decreased fromspring to fall) which corresponded to between 12 April and18 November each year (Figs 3 4) Specifically from mid-April to mid-November of each year the estimated probabilityof detection during VES increased from 058 to 060 forA crepitans and decreased from 040 to 030 forL catesbeianus from 078 to 032 for A fowleri from 045 to019 for H cinerea from 073 to 048 for L clamitans and from080 to 017 for H chrysoscelis (Figs 3 4) For all summer-breeding frogs detection probability during the MCS callingwindow increased with increasing day of year whichcorresponds to between 9 April and 28 June each year (Figs 34) Specifically from early April to late June of each year theestimated probability of detection during MCS increasedfrom 025 to 060 for A crepitans from 022 to 060 forL catesbeianus from 021 to 095 for A fowleri from 0002to099 forH cinerea from022 to 080 forL clamitans and from020 to 078 for H chrysoscelis (Figs 3 4)

For all winter-breeding frogs detection probability duringthe active-search survey window varied slightly withincreasing day of year (Fig 5) the estimated probability ofdetection during VES decreased from 050 to 040 forL sphenocephalus and from 041 to 031 for A americanusand increased from 005 to 018 for P feriarum and from 003 to

100

090

080

070

060

050

040

030

020

010

Est

imat

ed p

roba

bilit

y of

det

ectio

n

0

A americanus

A fowleri

A terre

stris

A crepitans

H cinerea

H chrysoscelis

P feriarum

P crucifer

G carolinensis

L catesbeianus

L clamitans

L palustris

L sphenocephalus

Fig 2 Estimated conditional probability of detection (1 se) for each anuran species along the Broad and PacoletRivers South Carolina USA Estimates are based on repeat surveys using two different survey methods (MCS or VES)

Anuran detection probability Wildlife Research G

018 for P crucifer (Fig 5) For all winter-breeding frogsdetection probability during the calling-survey time spandecreased with increasing day of year (Fig 5) Specifically

from mid-February to mid-May of each year the estimatedprobability of detection during MCS decreased from 038 to002 for L sphenocephalus from 098 to 001 for P feriarum

10MCS

Est

imat

ed p

roba

bilit

y of

det

ectio

n

VES

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

9Apr

29Apr

19May

8Jun

28Jun

12Apr

1Jun

21Jul

19Sep

18Nov

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

Fig 3 Effect of date by survey method (MCS or VES) on conditional detection probability ofsummer-breeding anuran species along the Broad and Pacolet Rivers South Carolina USA Solid linesrepresent the relationship between date and detection probability and dashed lines are the 95 CIsfor estimates of the covariate effect

H Wildlife Research J C Guzy et al

from 089 to 005 for A americanus and from 098 to 001 forP crucifer (Fig 5)

Discussion

Our study has provided insight for monitoring anurans byimplementing a modelling framework where data from bothMCS and VES are evaluated simultaneously resulting inmethod-specific detection probabilities Our results providedstrong evidence of variation in anuran detection probabilitiesamong two different detection methods Specifically for 10 of13 species the best-supported model indicated that detectionvaried by survey method Consistent with previous studies (egGooch et al 2006 Steelman and Dorcas 2010) calling surveyswere effective at detecting a wide range of anurans Visualencounter surveys however were also effective at detecting awide range of anuran species and several species had a higherdetection probability with VES thanwithMCSVisual encountersurveys are likely to be under-used as a monitoring tool foranurans perhaps because of a lack of information on theirutility Instead VES are often used along with other methodsin surveys for focal or rare species where the intent is gearedtoward complete coverage of all life stages to increase detectionprobability (eg Drost and Fellers 1996 Burbrink et al 1998Lips 1999 Lehtinen and Galatowitsch 2001)

Although both MCS and VES are effective methods forcertain species in our study one method was most effective

Specifically VES represent a more effective method fordetecting A crepitans G carolinensis and L sphenocephaluswhich is perhaps reflective of variation in species life historyFor example G carolinensis and L sphenocephalus like mostanurans can often be observed via vocalisations made duringbreeding however the breeding windows are wide (particularlyfor L sphenocephalus) calling by both species is sporadic andfor G carolinensis calling is usually initiated only by heavyrains (Brandt 1936 Dodd 2013) In addition calls of bothG carolinensis and L sphenocephalus are easily masked bylouder species (eg H cinerea and P crucifer Conant andCollins 1998) Likely because both G carolinensis andL sphenocephalus are active during the non-breeding periodin a wide variety of aquatic and terrestrial habitats (Mitchelland Lannoo 2005 Dodd 2013) VES was a more reliable methodof detection Acris crepitans another species with higherdetection probabilities during VES is active and conspicuousduring the non-breeding season Several other anuran speciesmight also be more reliably detected using VES if theirvocalisations are difficult to observe and their non-breedinghabits permit detection (ie little grass frogs (Pseudacrisocularis) or species that exhibit explosive andor short-livedbreeding periods (eg Scaphiopus spp L sylvaticusL capito) In these cases precisely timing MCS during thisbrief window is challenging (Dostine et al 2013) and oftennot ideal because anuran call surveys are less effective duringheavy rain (Weir et al 2005)

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

9Apr

29Apr

19May

8Jun

28Jun

12Apr

1Jun

21Jul

19Sep

18Nov

MCS VES

Est

imat

ed p

roba

bilit

y of

det

ectio

n

Fig 4 Effect of date by survey method (MCS or VES) on conditional detection probability of summer-breeding anuran species along the Broad and Pacolet Rivers South Carolina USA Solid lines represent therelationship between date and detection probability and dashed lines are the 95 CIs for estimates of thecovariate effect

Anuran detection probability Wildlife Research I

We found that MCS represented a more effectivedetection method for H cinerea P feriarum P crucifer andL catesbeianus These species are common vocal and calls can

be heard from a considerable distance (Dorcas et al 2007) Fora few species however both survey methods were equallyeffective (ie for A fowleri L clamitans and L palustris) In

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

14Feb

10Mar

4Apr

24Apr

14May

12Apr

1Jun

21Jul

19Sep

18Nov

Est

imat

ed p

roba

bilit

y of

det

ectio

n

MCS VES

Fig 5 Effect of date by survey method (MCS or VES) on conditional detection probability of winter-breeding anuran species along the Broad and Pacolet Rivers South Carolina USA Solid lines representthe relationship between date and detection probability and dashed lines are the 95 CIs for estimates ofthe covariate effect

J Wildlife Research J C Guzy et al

addition given the marginally larger detection probabilities ofA americanus using VES and of H chrysoscelis using MCSfor these species we also consider both survey methods to besimilarly effective For both A americanus and A fowleriMCS and VES detection probabilities were very similar MCSare effective because males of each species produce loud easilyrecognisable calls (Dorcas et al 2007) but VES are also effectiveperhaps because of their conspicuous tadpoles which school inlarge groups (Lefcort 1998) and the tendency for large numbersof metamorphs to forage for several weeks around the edge ofthe aquatic habitat fromwhich they emerged (Beck and Congdon1999 Dodd 2013) Likewise VES had detection probabilitiessimilar to those of MCS for L clamitans L palustris andH chrysoscelis for L clamitans this may be because whenstartled they emit conspicuous alarm calls as they leap into thewater (Dorcas et al 2007) and L palustrismay call in any monthof the year in the southern United States (Conant and Collins1998) Finally for H chrysoscelis its tendency to callsporadically during the daytime likely accounts for thesimilarity in effectiveness of both survey methods

Datewas an important factor influencing detection probabilityof anurans particularly regarding call surveys For summer-breeding anurans detection probability steadily increasedtoward the end of the summer survey windows and likewisefor winter-breeding anurans detection probability was highesttoward the beginning of the winter survey windows Previousstudies have documented strong patterns in seasonal breedingactivity and breeding windows are well established (Droege andEagle 2005) However for VES the potential influence of dateto increase or decrease the effectiveness of this survey methodis not well known Our results indicated that for VES for mostspecies detection probability was relatively constant throughoutthe sampling window (ie AprilndashNovember each year) althoughfor a few species (A fowleri H chrysoscelis H cinerea andL clamitans) detection probability steadily decreased into thefall months Furthermore although we showed that VES arebetter for some species we also highlighted that VES canprovide adequate occupancy estimation even when conductedoutside the breeding window For example for L clamitans andA fowleri VES conducted in early April resulted in a 60probability of detection and during MCS it was not until lateJune toward the end of the breeding season that detectionprobabilities reached 60 Similarly for H chrysoscelis VESconducted in early April resulted in an 80 probability ofdetection and for A crepitans detection probabilities remained~60 during the VES survey window for both of thesespecies detection probabilities comparable to 80 and 60respectively did not occur until late June Therefore we suggestthat spring and fall VES can detect species that vocalise mostlyduring the summer months

Management implications

The goals of a particular study should dictate the appropriatesurvey method however results of the present study highlightedthe importance of estimating method-specific detectionprobabilities which should prove especially useful in thedesign of multispecies surveys We recommend the use MCSfor anurans with broad calling windows and conspicuous

vocalisations and VES for anurans that either vocalisesporadically or discreetly Furthermore we recommend the useof VES throughout the activity season not just the breedingseason because we found that survey date was a largelyunimportant predictor of detection using VES Our results alsosuggested that VES can provide more flexibility for anuranresearchers because robust estimates of detection andoccupancy can be obtained outside a narrow breeding window

Acknowledgements

This research was supported with funding provided by the Broad RiverMitigation Trust Fund administered through the South CarolinaDepartment of Natural Resources Additional funding was provided by theDukeEndowment through theDavidsonResearch Initiative and theDavidsonCollege Department of Biology We thank E Eskew S Hunt L WitczakC Williams A Domske D Millican C Oldham R Bauer B Abbuhl andC Ruder for assistance in the field and J Micancin for his help in identifyingcricket frog vocal recordings For assistance with logistics related to theprojectwe thankGVaughanSBennettBPerry andVVejdaniFinallywethank D Ramsey and two anonymous reviewers for statistical assistanceand comments which greatly improved this manuscript

References

Adams M J Pearl C A Galvan S andMcCreary B (2011) Non-nativespecies impacts on pond occupancy by an anuran The Journal of WildlifeManagement 75 30ndash35 doi101002jwmg29

Bailey L L Simons T R and Pollock K H (2004) Estimating siteoccupancy and species detection probability parameters for terrestrialsalamandersEcologicalApplications14 692ndash702 doi10189003-5012

BalasC JEulissNH andMushetDM (2012) Influenceof conservationprograms on amphibians using seasonal wetlands in the prairie potholeregion Wetlands 32 333ndash345 doi101007s13157-012-0269-9

Beck CW andCongdon J D (1999) Effects of individual variation in ageand size at metamorphosis on growth and survivorship of southern toad(Bufo terrestris)metamorphsCanadian Journal of Zoology77 944ndash951doi101139cjz-77-6-944

Blair W F (1961) Calling and spawning seasons in a mixed population ofanurans Ecology 42 99ndash110 doi1023071933272

Brander S M Royle J A and Eames M (2007) Evaluation of the statusof anurans on a refuge in suburbanMaryland Journal of Herpetology 4152ndash60 doi1016700022-1511(2007)41[52EOTSOA]20CO2

Brandt B B (1936) The frogs and toads of eastern North Carolina Copeia1936 215ndash223 doi1023071436326

Brown G W Scroggie M P Smith M J and Steane D (2007) Anevaluation of methods for assessing the population status of thethreatened Alpine tree frog Litoria verreauxii alpina in southeasternAustralia Copeia 3 765ndash770 doi1016430045-8511(2007)2007[765AEOMFA]20CO2

BucklandSTBurnhamKP andAugustinNH (1997)Model selectionan integral part of inference Biometrics 53 603ndash618 doi1023072533961

Burbrink F T Phillips C A and Heske E J (1998) A riparian zone insouthern Illinois as a potential dispersal corridor for reptiles andamphibians Biological Conservation 86 107ndash115 doi101016S0006-3207(98)00054-8

Burnham K P and Anderson D R (2002) lsquoModel Selection andMultimodel Inference a Practical Information-theoretic Approachrsquo2nd edn (Springer-Verlag New York)

Conant R and Collins J T (1998) lsquoA Field Guide to Amphibians andReptiles Eastern andCentralNorthAmericarsquo 3rd edn (HoughtonMifflinCompany Boston MA)

CookR P Tupper TA Paton PWC andTimmBC (2011) Effects oftemperature and temporal factors on anuran detection probabilities at

Anuran detection probability Wildlife Research K

Tab

le2

Best-supp

ortedmod

elsfordetectionof

anuran

speciesba

sedon

QAIC

cselectioncriteria

The

sameeightcandidatemodelswereconstructedforeachspeciesfrom

uncorrelated

covariatesB

est-supportedmodelsareinboldA

ICA

kaikeinform

ationcriteriaDQ

AICcdifference

inQAICcrelativ

eto

thetopmod

elwmod

elweightKnu

mberof

parametersin

themod

elMetho

danuran

survey

method(V

ESor

MCS)Day

ofyeard

ayof

year

each

survey

was

conducted

Dayssincerainnu

mber

ofdays

sincerain

foragivensurvey

Species

Model

AICc

DQAICc

wK

ndash2log-lik

elihoo

dParam

eter

estim

ate(se)

Metho

dDay

ofyear

Num

berof

days

sincerain

Y-intercept

p-intercept

p-VES

p-MCS

p-VES

p-MCS

p-VES

p-MCS

Aa

mericanus

Y()Aq

()Ap

B(D

ate)

32364

000

057

431314

494

(402)

ndash287

(023)

ndashndash

-013

(009)

ndash138

(022)

ndashndash

Y()q()p(DateRain)

32554

190

022

630

754

487

(400)

ndash297

(027)

ndashndash

ndash009

(009)

ndash147

(024)

ndash064

(031)

019

(017)

Y()q()p(Method

Date)

32599

235

018

531199

392

(254)

ndashndash270

(026)

ndash323

(042)

ndash007

(008)

ndash164

(034)

ndashndash

Modelaveraged

459

(362)

ndash229

(019)

ndash049

(005)

ndash058

(008)

ndash011

(009)

ndash141

(024)

ndash014

(007)

004

(004)

Afow

leri

C()q()p(DateRain)

64863

000

073

663

063

212

(050)

ndash026

(015)

ndashndash

ndash051

(017)

124

(028)

006

(004)

084

(019)

Y()q()p(Method

DateRain)

65058

195

027

762796

206

(048)

ndashndash039

(017)

ndash002

(022)

ndash056

(017)

110

(030)

004

(015)

086

(020)

Aterrestris

Y()q()p()

5747

000

044

252

80

ndash095

(073)

ndash187

(119)

ndashndash

ndashndash

ndashndash

Y()q()p(Method)

5798

051

034

35057

ndash107

(072)

ndashndash271

(144)

ndash130

(128)

ndashndash

ndashndash

Modelaveraged

ndash078

(057)

ndash082

(052)

ndash092

(049)

ndash044

(043)

Acrepitans

Y()q()p(Method

Date)

53038

000

055

551

638

121

(029)

ndash039

(023)

ndash049

(024)

010

(008)

073

(027)

ndashndash

Y()q()p(Method)

53149

111

031

352408

108

(028)

ndash041

(024)

ndash015

(011)

ndashndash

ndashndash

Modelaveraged

100

(025)

ndash034

(020)

ndash032

(017)

006

(004)

040

(015)

ndashndash

Hcinerea

Y()q()p(Method

Date)

35484

000

054

534084

070

(041)

ndashndash181

(033)

ndash098

(045)

ndash051

(032)

292

(113)

ndashndash

Y()q()p(Date)

35521

037

045

434

471

087

(039)

ndash157

(026)

ndashndash

ndash037

(027)

296

(070)

ndashndash

Modelaveraged

077

(040)

ndash071

(012)

ndash098

(018)

-053

(024)

-044

(029)

290

(093)

ndashndash

Hchrysoscelis

Y()q()p(Date)

52682

000

058

451

632

202

(106)

ndash125

(021)

ndashndash

ndash092

(027)

081

(025)

ndashndash

Y()q()p(Method

Date)

52900

218

019

551100

200

(108)

ndashndash141

(025)

ndash109

(025)

ndash104

(032)

069

(027)

(contin

uednextpa

ge)

Anuran detection probability Wildlife Research E

Tab

le2

(con

tinued)

Species

Model

AICc

DQAICc

wK

ndash2log-lik

elihoo

dParam

eter

estim

ate(se)

Metho

dDay

ofyear

Num

berof

days

sincerain

Y-intercept

p-intercept

p-VES

p-MCS

p-VES

p-MCS

p-VES

p-MCS

Y()q()p(DateRain)

52902

220

019

651

502

205

(104)

ndash128

(020)

ndashndash

ndash093

(028)

082

(026)

ndash025

(017)

029

(016)

Modelaveraged

194

(102)

ndash097

(016)

ndash027

(005)

ndash021

(005)

ndash091

(027)

076

(025)

ndash005

(004)

006

(003)

Pferiarum

C()q()p(Metho

dDate)

30614

000

097

529

214

102

(032)

ndashndash251

(024)

ndash201

(066)

035

(024)

ndash333

(050)

ndashndash

Pcrucifer

C()q()p(Metho

dDate)

41265

000

085

539

865

087

(033)

ndashndash260

(034)

ndash120

(036)

046

(032)

ndash273

(100)

ndashndash

Y()q()p(Date)

41641

376

013

440

591

071

(020)

ndash211

(024)

ndashndash

049

(028)

ndash495

(105)

ndashndash

Gcarolinensis

C()q()p(Metho

d)15

168

000

093

313

768

ndash041

(030)

ndashndash056

(041)

ndash237

(061)

ndashndash

ndashndash

Lcatesbeianus

Y()q()p(Date)

43926

000

044

442

876

153

(052)

ndash128

(019)

ndashndash

-019

(015)

074

(025)

ndashndash

Y()q()p(Method)

44010

084

029

343269

158

(052)

ndashndash143

(023)

ndash082

(021)

ndashndash

ndashndash

Y()q()p(Method

Date)

44118

192

017

542718

153

(052)

ndashndash144

(023)

ndash107

(025)

ndash015

(021)

058

(027)

ndashndash

Modelaveraged

139

(047)

ndash056

(008)

ndash066

(011)

ndash042

(010)

ndash011

(010)

042

(016)

Lclamita

nsC()q()p(Date)

44153

000

058

443

103

178

(061)

ndash144

(019)

ndashndash

ndash029

(021)

072

(024)

ndashndash

Y()q()p(DateRain)

44376

223

019

642

576

179

(060)

ndash147

(019)

ndashndash

ndash031

(021)

074

(024)

042

(018)

003

(018)

Lp

alustris

C()p()

20071

000

065

219

604

101

(067)

ndash226

(041)

ndashndash

ndashndash

ndashndash

Y()p(Metho

d)20

341

270

017

319

600

101

(070)

ndash222

(046)

ndash231

(046)

ndashndash

ndashndash

Lsph

enocepha

lus

C()q()p(Metho

dDate)

48414

000

078

547

014

163

(056)

ndashndash025

(022)

ndash249

(041)

ndash008

(006)

ndash110

(036)

ndashndash

Y()q()p(Method)

48804

390

011

348063

166

(056)

ndashndash026

(022)

ndash154

(023)

ndashndash

ndashndash

AConstantprobability

ofoccupancy(ieY

()q())

BProbabilityof

detection

which

varies

bymodelforeach

specieswith

orwith

outinclusionof

covariates

thatim

provedetectionestim

ates

F Wildlife Research J C Guzy et al

method-specific detection probabilities for this species were notpossible to estimate

The best supported model for 10 anuran species indicatedthat detection varied by survey method (ie VES or MCSTable 2 Appendix 1) for four of these species (P feriarumP crucifer G carolinensis and L sphenocephalus) detectionvaried by survey method and for six species the best-supportedmodel was determined from model averaging (A americanusA terrestris A crepitans H cinerea H chrysoscelis andL catesbeianus) For the remaining three species (A fowleriL palustris and L clamitans) survey method was unimportantand the best-supported model suggested either a constantprobability of detection (L palustris) or suggested thatdetection varied by date (L clamitans) or date and days sincerainfall (A fowleri Table 2)

For six species MCS resulted in a higher mean (se)probability of detection than did VES (A terrestris MCS 039(038) vs VES 028 015) H cinerea MCS 037 (017) vs VES027 (005)H chrysoscelisMCS045 (011) vsVES043 (008)P feriarum MCS 012 (004) vs VES 008 (001) P cruciferMCS 023 (007) vs VES 007 (001) and L catesbeianus MCS040 (009) vs VES 034 (006) Fig 2) Conversely for fourspecies VES resulted in a higher mean probability of detectionthan did MCS (A americanus VES 038 (004) vs MCS 036(005) A crepitans VES 060 (036) vs MCS 038 (018)G carolinensis VES 036 (027) vs MCS 009 (002) andL sphenocephalus VES 044 (039) vs MCS 008 (001)Fig 2) For three species however both methods were equallyeffective and detection probabilities were identical (A fowleri044 (025) L clamitans 046 (008) and L palustris 009(002) Fig 2)

In addition to survey method date was an important factorinfluencing detection probability and was included in thebest-supported model for 10 species (A americanusA fowleri A crepitans H cinerea H chrysoscelisP feriarum P crucifer L catesbeianus L clamitans andL sphenocephalus Table 2) For all summer-breeding frogsexcept A crepitans detection probability during the VESdecreased with increasing day of year (ie decreased fromspring to fall) which corresponded to between 12 April and18 November each year (Figs 3 4) Specifically from mid-April to mid-November of each year the estimated probabilityof detection during VES increased from 058 to 060 forA crepitans and decreased from 040 to 030 forL catesbeianus from 078 to 032 for A fowleri from 045 to019 for H cinerea from 073 to 048 for L clamitans and from080 to 017 for H chrysoscelis (Figs 3 4) For all summer-breeding frogs detection probability during the MCS callingwindow increased with increasing day of year whichcorresponds to between 9 April and 28 June each year (Figs 34) Specifically from early April to late June of each year theestimated probability of detection during MCS increasedfrom 025 to 060 for A crepitans from 022 to 060 forL catesbeianus from 021 to 095 for A fowleri from 0002to099 forH cinerea from022 to 080 forL clamitans and from020 to 078 for H chrysoscelis (Figs 3 4)

For all winter-breeding frogs detection probability duringthe active-search survey window varied slightly withincreasing day of year (Fig 5) the estimated probability ofdetection during VES decreased from 050 to 040 forL sphenocephalus and from 041 to 031 for A americanusand increased from 005 to 018 for P feriarum and from 003 to

100

090

080

070

060

050

040

030

020

010

Est

imat

ed p

roba

bilit

y of

det

ectio

n

0

A americanus

A fowleri

A terre

stris

A crepitans

H cinerea

H chrysoscelis

P feriarum

P crucifer

G carolinensis

L catesbeianus

L clamitans

L palustris

L sphenocephalus

Fig 2 Estimated conditional probability of detection (1 se) for each anuran species along the Broad and PacoletRivers South Carolina USA Estimates are based on repeat surveys using two different survey methods (MCS or VES)

Anuran detection probability Wildlife Research G

018 for P crucifer (Fig 5) For all winter-breeding frogsdetection probability during the calling-survey time spandecreased with increasing day of year (Fig 5) Specifically

from mid-February to mid-May of each year the estimatedprobability of detection during MCS decreased from 038 to002 for L sphenocephalus from 098 to 001 for P feriarum

10MCS

Est

imat

ed p

roba

bilit

y of

det

ectio

n

VES

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

9Apr

29Apr

19May

8Jun

28Jun

12Apr

1Jun

21Jul

19Sep

18Nov

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

Fig 3 Effect of date by survey method (MCS or VES) on conditional detection probability ofsummer-breeding anuran species along the Broad and Pacolet Rivers South Carolina USA Solid linesrepresent the relationship between date and detection probability and dashed lines are the 95 CIsfor estimates of the covariate effect

H Wildlife Research J C Guzy et al

from 089 to 005 for A americanus and from 098 to 001 forP crucifer (Fig 5)

Discussion

Our study has provided insight for monitoring anurans byimplementing a modelling framework where data from bothMCS and VES are evaluated simultaneously resulting inmethod-specific detection probabilities Our results providedstrong evidence of variation in anuran detection probabilitiesamong two different detection methods Specifically for 10 of13 species the best-supported model indicated that detectionvaried by survey method Consistent with previous studies (egGooch et al 2006 Steelman and Dorcas 2010) calling surveyswere effective at detecting a wide range of anurans Visualencounter surveys however were also effective at detecting awide range of anuran species and several species had a higherdetection probability with VES thanwithMCSVisual encountersurveys are likely to be under-used as a monitoring tool foranurans perhaps because of a lack of information on theirutility Instead VES are often used along with other methodsin surveys for focal or rare species where the intent is gearedtoward complete coverage of all life stages to increase detectionprobability (eg Drost and Fellers 1996 Burbrink et al 1998Lips 1999 Lehtinen and Galatowitsch 2001)

Although both MCS and VES are effective methods forcertain species in our study one method was most effective

Specifically VES represent a more effective method fordetecting A crepitans G carolinensis and L sphenocephaluswhich is perhaps reflective of variation in species life historyFor example G carolinensis and L sphenocephalus like mostanurans can often be observed via vocalisations made duringbreeding however the breeding windows are wide (particularlyfor L sphenocephalus) calling by both species is sporadic andfor G carolinensis calling is usually initiated only by heavyrains (Brandt 1936 Dodd 2013) In addition calls of bothG carolinensis and L sphenocephalus are easily masked bylouder species (eg H cinerea and P crucifer Conant andCollins 1998) Likely because both G carolinensis andL sphenocephalus are active during the non-breeding periodin a wide variety of aquatic and terrestrial habitats (Mitchelland Lannoo 2005 Dodd 2013) VES was a more reliable methodof detection Acris crepitans another species with higherdetection probabilities during VES is active and conspicuousduring the non-breeding season Several other anuran speciesmight also be more reliably detected using VES if theirvocalisations are difficult to observe and their non-breedinghabits permit detection (ie little grass frogs (Pseudacrisocularis) or species that exhibit explosive andor short-livedbreeding periods (eg Scaphiopus spp L sylvaticusL capito) In these cases precisely timing MCS during thisbrief window is challenging (Dostine et al 2013) and oftennot ideal because anuran call surveys are less effective duringheavy rain (Weir et al 2005)

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

9Apr

29Apr

19May

8Jun

28Jun

12Apr

1Jun

21Jul

19Sep

18Nov

MCS VES

Est

imat

ed p

roba

bilit

y of

det

ectio

n

Fig 4 Effect of date by survey method (MCS or VES) on conditional detection probability of summer-breeding anuran species along the Broad and Pacolet Rivers South Carolina USA Solid lines represent therelationship between date and detection probability and dashed lines are the 95 CIs for estimates of thecovariate effect

Anuran detection probability Wildlife Research I

We found that MCS represented a more effectivedetection method for H cinerea P feriarum P crucifer andL catesbeianus These species are common vocal and calls can

be heard from a considerable distance (Dorcas et al 2007) Fora few species however both survey methods were equallyeffective (ie for A fowleri L clamitans and L palustris) In

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

14Feb

10Mar

4Apr

24Apr

14May

12Apr

1Jun

21Jul

19Sep

18Nov

Est

imat

ed p

roba

bilit

y of

det

ectio

n

MCS VES

Fig 5 Effect of date by survey method (MCS or VES) on conditional detection probability of winter-breeding anuran species along the Broad and Pacolet Rivers South Carolina USA Solid lines representthe relationship between date and detection probability and dashed lines are the 95 CIs for estimates ofthe covariate effect

J Wildlife Research J C Guzy et al

addition given the marginally larger detection probabilities ofA americanus using VES and of H chrysoscelis using MCSfor these species we also consider both survey methods to besimilarly effective For both A americanus and A fowleriMCS and VES detection probabilities were very similar MCSare effective because males of each species produce loud easilyrecognisable calls (Dorcas et al 2007) but VES are also effectiveperhaps because of their conspicuous tadpoles which school inlarge groups (Lefcort 1998) and the tendency for large numbersof metamorphs to forage for several weeks around the edge ofthe aquatic habitat fromwhich they emerged (Beck and Congdon1999 Dodd 2013) Likewise VES had detection probabilitiessimilar to those of MCS for L clamitans L palustris andH chrysoscelis for L clamitans this may be because whenstartled they emit conspicuous alarm calls as they leap into thewater (Dorcas et al 2007) and L palustrismay call in any monthof the year in the southern United States (Conant and Collins1998) Finally for H chrysoscelis its tendency to callsporadically during the daytime likely accounts for thesimilarity in effectiveness of both survey methods

Datewas an important factor influencing detection probabilityof anurans particularly regarding call surveys For summer-breeding anurans detection probability steadily increasedtoward the end of the summer survey windows and likewisefor winter-breeding anurans detection probability was highesttoward the beginning of the winter survey windows Previousstudies have documented strong patterns in seasonal breedingactivity and breeding windows are well established (Droege andEagle 2005) However for VES the potential influence of dateto increase or decrease the effectiveness of this survey methodis not well known Our results indicated that for VES for mostspecies detection probability was relatively constant throughoutthe sampling window (ie AprilndashNovember each year) althoughfor a few species (A fowleri H chrysoscelis H cinerea andL clamitans) detection probability steadily decreased into thefall months Furthermore although we showed that VES arebetter for some species we also highlighted that VES canprovide adequate occupancy estimation even when conductedoutside the breeding window For example for L clamitans andA fowleri VES conducted in early April resulted in a 60probability of detection and during MCS it was not until lateJune toward the end of the breeding season that detectionprobabilities reached 60 Similarly for H chrysoscelis VESconducted in early April resulted in an 80 probability ofdetection and for A crepitans detection probabilities remained~60 during the VES survey window for both of thesespecies detection probabilities comparable to 80 and 60respectively did not occur until late June Therefore we suggestthat spring and fall VES can detect species that vocalise mostlyduring the summer months

Management implications

The goals of a particular study should dictate the appropriatesurvey method however results of the present study highlightedthe importance of estimating method-specific detectionprobabilities which should prove especially useful in thedesign of multispecies surveys We recommend the use MCSfor anurans with broad calling windows and conspicuous

vocalisations and VES for anurans that either vocalisesporadically or discreetly Furthermore we recommend the useof VES throughout the activity season not just the breedingseason because we found that survey date was a largelyunimportant predictor of detection using VES Our results alsosuggested that VES can provide more flexibility for anuranresearchers because robust estimates of detection andoccupancy can be obtained outside a narrow breeding window

Acknowledgements

This research was supported with funding provided by the Broad RiverMitigation Trust Fund administered through the South CarolinaDepartment of Natural Resources Additional funding was provided by theDukeEndowment through theDavidsonResearch Initiative and theDavidsonCollege Department of Biology We thank E Eskew S Hunt L WitczakC Williams A Domske D Millican C Oldham R Bauer B Abbuhl andC Ruder for assistance in the field and J Micancin for his help in identifyingcricket frog vocal recordings For assistance with logistics related to theprojectwe thankGVaughanSBennettBPerry andVVejdaniFinallywethank D Ramsey and two anonymous reviewers for statistical assistanceand comments which greatly improved this manuscript

References

Adams M J Pearl C A Galvan S andMcCreary B (2011) Non-nativespecies impacts on pond occupancy by an anuran The Journal of WildlifeManagement 75 30ndash35 doi101002jwmg29

Bailey L L Simons T R and Pollock K H (2004) Estimating siteoccupancy and species detection probability parameters for terrestrialsalamandersEcologicalApplications14 692ndash702 doi10189003-5012

BalasC JEulissNH andMushetDM (2012) Influenceof conservationprograms on amphibians using seasonal wetlands in the prairie potholeregion Wetlands 32 333ndash345 doi101007s13157-012-0269-9

Beck CW andCongdon J D (1999) Effects of individual variation in ageand size at metamorphosis on growth and survivorship of southern toad(Bufo terrestris)metamorphsCanadian Journal of Zoology77 944ndash951doi101139cjz-77-6-944

Blair W F (1961) Calling and spawning seasons in a mixed population ofanurans Ecology 42 99ndash110 doi1023071933272

Brander S M Royle J A and Eames M (2007) Evaluation of the statusof anurans on a refuge in suburbanMaryland Journal of Herpetology 4152ndash60 doi1016700022-1511(2007)41[52EOTSOA]20CO2

Brandt B B (1936) The frogs and toads of eastern North Carolina Copeia1936 215ndash223 doi1023071436326

Brown G W Scroggie M P Smith M J and Steane D (2007) Anevaluation of methods for assessing the population status of thethreatened Alpine tree frog Litoria verreauxii alpina in southeasternAustralia Copeia 3 765ndash770 doi1016430045-8511(2007)2007[765AEOMFA]20CO2

BucklandSTBurnhamKP andAugustinNH (1997)Model selectionan integral part of inference Biometrics 53 603ndash618 doi1023072533961

Burbrink F T Phillips C A and Heske E J (1998) A riparian zone insouthern Illinois as a potential dispersal corridor for reptiles andamphibians Biological Conservation 86 107ndash115 doi101016S0006-3207(98)00054-8

Burnham K P and Anderson D R (2002) lsquoModel Selection andMultimodel Inference a Practical Information-theoretic Approachrsquo2nd edn (Springer-Verlag New York)

Conant R and Collins J T (1998) lsquoA Field Guide to Amphibians andReptiles Eastern andCentralNorthAmericarsquo 3rd edn (HoughtonMifflinCompany Boston MA)

CookR P Tupper TA Paton PWC andTimmBC (2011) Effects oftemperature and temporal factors on anuran detection probabilities at

Anuran detection probability Wildlife Research K

Tab

le2

(con

tinued)

Species

Model

AICc

DQAICc

wK

ndash2log-lik

elihoo

dParam

eter

estim

ate(se)

Metho

dDay

ofyear

Num

berof

days

sincerain

Y-intercept

p-intercept

p-VES

p-MCS

p-VES

p-MCS

p-VES

p-MCS

Y()q()p(DateRain)

52902

220

019

651

502

205

(104)

ndash128

(020)

ndashndash

ndash093

(028)

082

(026)

ndash025

(017)

029

(016)

Modelaveraged

194

(102)

ndash097

(016)

ndash027

(005)

ndash021

(005)

ndash091

(027)

076

(025)

ndash005

(004)

006

(003)

Pferiarum

C()q()p(Metho

dDate)

30614

000

097

529

214

102

(032)

ndashndash251

(024)

ndash201

(066)

035

(024)

ndash333

(050)

ndashndash

Pcrucifer

C()q()p(Metho

dDate)

41265

000

085

539

865

087

(033)

ndashndash260

(034)

ndash120

(036)

046

(032)

ndash273

(100)

ndashndash

Y()q()p(Date)

41641

376

013

440

591

071

(020)

ndash211

(024)

ndashndash

049

(028)

ndash495

(105)

ndashndash

Gcarolinensis

C()q()p(Metho

d)15

168

000

093

313

768

ndash041

(030)

ndashndash056

(041)

ndash237

(061)

ndashndash

ndashndash

Lcatesbeianus

Y()q()p(Date)

43926

000

044

442

876

153

(052)

ndash128

(019)

ndashndash

-019

(015)

074

(025)

ndashndash

Y()q()p(Method)

44010

084

029

343269

158

(052)

ndashndash143

(023)

ndash082

(021)

ndashndash

ndashndash

Y()q()p(Method

Date)

44118

192

017

542718

153

(052)

ndashndash144

(023)

ndash107

(025)

ndash015

(021)

058

(027)

ndashndash

Modelaveraged

139

(047)

ndash056

(008)

ndash066

(011)

ndash042

(010)

ndash011

(010)

042

(016)

Lclamita

nsC()q()p(Date)

44153

000

058

443

103

178

(061)

ndash144

(019)

ndashndash

ndash029

(021)

072

(024)

ndashndash

Y()q()p(DateRain)

44376

223

019

642

576

179

(060)

ndash147

(019)

ndashndash

ndash031

(021)

074

(024)

042

(018)

003

(018)

Lp

alustris

C()p()

20071

000

065

219

604

101

(067)

ndash226

(041)

ndashndash

ndashndash

ndashndash

Y()p(Metho

d)20

341

270

017

319

600

101

(070)

ndash222

(046)

ndash231

(046)

ndashndash

ndashndash

Lsph

enocepha

lus

C()q()p(Metho

dDate)

48414

000

078

547

014

163

(056)

ndashndash025

(022)

ndash249

(041)

ndash008

(006)

ndash110

(036)

ndashndash

Y()q()p(Method)

48804

390

011

348063

166

(056)

ndashndash026

(022)

ndash154

(023)

ndashndash

ndashndash

AConstantprobability

ofoccupancy(ieY

()q())

BProbabilityof

detection

which

varies

bymodelforeach

specieswith

orwith

outinclusionof

covariates

thatim

provedetectionestim

ates

F Wildlife Research J C Guzy et al

method-specific detection probabilities for this species were notpossible to estimate

The best supported model for 10 anuran species indicatedthat detection varied by survey method (ie VES or MCSTable 2 Appendix 1) for four of these species (P feriarumP crucifer G carolinensis and L sphenocephalus) detectionvaried by survey method and for six species the best-supportedmodel was determined from model averaging (A americanusA terrestris A crepitans H cinerea H chrysoscelis andL catesbeianus) For the remaining three species (A fowleriL palustris and L clamitans) survey method was unimportantand the best-supported model suggested either a constantprobability of detection (L palustris) or suggested thatdetection varied by date (L clamitans) or date and days sincerainfall (A fowleri Table 2)

For six species MCS resulted in a higher mean (se)probability of detection than did VES (A terrestris MCS 039(038) vs VES 028 015) H cinerea MCS 037 (017) vs VES027 (005)H chrysoscelisMCS045 (011) vsVES043 (008)P feriarum MCS 012 (004) vs VES 008 (001) P cruciferMCS 023 (007) vs VES 007 (001) and L catesbeianus MCS040 (009) vs VES 034 (006) Fig 2) Conversely for fourspecies VES resulted in a higher mean probability of detectionthan did MCS (A americanus VES 038 (004) vs MCS 036(005) A crepitans VES 060 (036) vs MCS 038 (018)G carolinensis VES 036 (027) vs MCS 009 (002) andL sphenocephalus VES 044 (039) vs MCS 008 (001)Fig 2) For three species however both methods were equallyeffective and detection probabilities were identical (A fowleri044 (025) L clamitans 046 (008) and L palustris 009(002) Fig 2)

In addition to survey method date was an important factorinfluencing detection probability and was included in thebest-supported model for 10 species (A americanusA fowleri A crepitans H cinerea H chrysoscelisP feriarum P crucifer L catesbeianus L clamitans andL sphenocephalus Table 2) For all summer-breeding frogsexcept A crepitans detection probability during the VESdecreased with increasing day of year (ie decreased fromspring to fall) which corresponded to between 12 April and18 November each year (Figs 3 4) Specifically from mid-April to mid-November of each year the estimated probabilityof detection during VES increased from 058 to 060 forA crepitans and decreased from 040 to 030 forL catesbeianus from 078 to 032 for A fowleri from 045 to019 for H cinerea from 073 to 048 for L clamitans and from080 to 017 for H chrysoscelis (Figs 3 4) For all summer-breeding frogs detection probability during the MCS callingwindow increased with increasing day of year whichcorresponds to between 9 April and 28 June each year (Figs 34) Specifically from early April to late June of each year theestimated probability of detection during MCS increasedfrom 025 to 060 for A crepitans from 022 to 060 forL catesbeianus from 021 to 095 for A fowleri from 0002to099 forH cinerea from022 to 080 forL clamitans and from020 to 078 for H chrysoscelis (Figs 3 4)

For all winter-breeding frogs detection probability duringthe active-search survey window varied slightly withincreasing day of year (Fig 5) the estimated probability ofdetection during VES decreased from 050 to 040 forL sphenocephalus and from 041 to 031 for A americanusand increased from 005 to 018 for P feriarum and from 003 to

100

090

080

070

060

050

040

030

020

010

Est

imat

ed p

roba

bilit

y of

det

ectio

n

0

A americanus

A fowleri

A terre

stris

A crepitans

H cinerea

H chrysoscelis

P feriarum

P crucifer

G carolinensis

L catesbeianus

L clamitans

L palustris

L sphenocephalus

Fig 2 Estimated conditional probability of detection (1 se) for each anuran species along the Broad and PacoletRivers South Carolina USA Estimates are based on repeat surveys using two different survey methods (MCS or VES)

Anuran detection probability Wildlife Research G

018 for P crucifer (Fig 5) For all winter-breeding frogsdetection probability during the calling-survey time spandecreased with increasing day of year (Fig 5) Specifically

from mid-February to mid-May of each year the estimatedprobability of detection during MCS decreased from 038 to002 for L sphenocephalus from 098 to 001 for P feriarum

10MCS

Est

imat

ed p

roba

bilit

y of

det

ectio

n

VES

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

9Apr

29Apr

19May

8Jun

28Jun

12Apr

1Jun

21Jul

19Sep

18Nov

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

Fig 3 Effect of date by survey method (MCS or VES) on conditional detection probability ofsummer-breeding anuran species along the Broad and Pacolet Rivers South Carolina USA Solid linesrepresent the relationship between date and detection probability and dashed lines are the 95 CIsfor estimates of the covariate effect

H Wildlife Research J C Guzy et al

from 089 to 005 for A americanus and from 098 to 001 forP crucifer (Fig 5)

Discussion

Our study has provided insight for monitoring anurans byimplementing a modelling framework where data from bothMCS and VES are evaluated simultaneously resulting inmethod-specific detection probabilities Our results providedstrong evidence of variation in anuran detection probabilitiesamong two different detection methods Specifically for 10 of13 species the best-supported model indicated that detectionvaried by survey method Consistent with previous studies (egGooch et al 2006 Steelman and Dorcas 2010) calling surveyswere effective at detecting a wide range of anurans Visualencounter surveys however were also effective at detecting awide range of anuran species and several species had a higherdetection probability with VES thanwithMCSVisual encountersurveys are likely to be under-used as a monitoring tool foranurans perhaps because of a lack of information on theirutility Instead VES are often used along with other methodsin surveys for focal or rare species where the intent is gearedtoward complete coverage of all life stages to increase detectionprobability (eg Drost and Fellers 1996 Burbrink et al 1998Lips 1999 Lehtinen and Galatowitsch 2001)

Although both MCS and VES are effective methods forcertain species in our study one method was most effective

Specifically VES represent a more effective method fordetecting A crepitans G carolinensis and L sphenocephaluswhich is perhaps reflective of variation in species life historyFor example G carolinensis and L sphenocephalus like mostanurans can often be observed via vocalisations made duringbreeding however the breeding windows are wide (particularlyfor L sphenocephalus) calling by both species is sporadic andfor G carolinensis calling is usually initiated only by heavyrains (Brandt 1936 Dodd 2013) In addition calls of bothG carolinensis and L sphenocephalus are easily masked bylouder species (eg H cinerea and P crucifer Conant andCollins 1998) Likely because both G carolinensis andL sphenocephalus are active during the non-breeding periodin a wide variety of aquatic and terrestrial habitats (Mitchelland Lannoo 2005 Dodd 2013) VES was a more reliable methodof detection Acris crepitans another species with higherdetection probabilities during VES is active and conspicuousduring the non-breeding season Several other anuran speciesmight also be more reliably detected using VES if theirvocalisations are difficult to observe and their non-breedinghabits permit detection (ie little grass frogs (Pseudacrisocularis) or species that exhibit explosive andor short-livedbreeding periods (eg Scaphiopus spp L sylvaticusL capito) In these cases precisely timing MCS during thisbrief window is challenging (Dostine et al 2013) and oftennot ideal because anuran call surveys are less effective duringheavy rain (Weir et al 2005)

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

9Apr

29Apr

19May

8Jun

28Jun

12Apr

1Jun

21Jul

19Sep

18Nov

MCS VES

Est

imat

ed p

roba

bilit

y of

det

ectio

n

Fig 4 Effect of date by survey method (MCS or VES) on conditional detection probability of summer-breeding anuran species along the Broad and Pacolet Rivers South Carolina USA Solid lines represent therelationship between date and detection probability and dashed lines are the 95 CIs for estimates of thecovariate effect

Anuran detection probability Wildlife Research I

We found that MCS represented a more effectivedetection method for H cinerea P feriarum P crucifer andL catesbeianus These species are common vocal and calls can

be heard from a considerable distance (Dorcas et al 2007) Fora few species however both survey methods were equallyeffective (ie for A fowleri L clamitans and L palustris) In

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

14Feb

10Mar

4Apr

24Apr

14May

12Apr

1Jun

21Jul

19Sep

18Nov

Est

imat

ed p

roba

bilit

y of

det

ectio

n

MCS VES

Fig 5 Effect of date by survey method (MCS or VES) on conditional detection probability of winter-breeding anuran species along the Broad and Pacolet Rivers South Carolina USA Solid lines representthe relationship between date and detection probability and dashed lines are the 95 CIs for estimates ofthe covariate effect

J Wildlife Research J C Guzy et al

addition given the marginally larger detection probabilities ofA americanus using VES and of H chrysoscelis using MCSfor these species we also consider both survey methods to besimilarly effective For both A americanus and A fowleriMCS and VES detection probabilities were very similar MCSare effective because males of each species produce loud easilyrecognisable calls (Dorcas et al 2007) but VES are also effectiveperhaps because of their conspicuous tadpoles which school inlarge groups (Lefcort 1998) and the tendency for large numbersof metamorphs to forage for several weeks around the edge ofthe aquatic habitat fromwhich they emerged (Beck and Congdon1999 Dodd 2013) Likewise VES had detection probabilitiessimilar to those of MCS for L clamitans L palustris andH chrysoscelis for L clamitans this may be because whenstartled they emit conspicuous alarm calls as they leap into thewater (Dorcas et al 2007) and L palustrismay call in any monthof the year in the southern United States (Conant and Collins1998) Finally for H chrysoscelis its tendency to callsporadically during the daytime likely accounts for thesimilarity in effectiveness of both survey methods

Datewas an important factor influencing detection probabilityof anurans particularly regarding call surveys For summer-breeding anurans detection probability steadily increasedtoward the end of the summer survey windows and likewisefor winter-breeding anurans detection probability was highesttoward the beginning of the winter survey windows Previousstudies have documented strong patterns in seasonal breedingactivity and breeding windows are well established (Droege andEagle 2005) However for VES the potential influence of dateto increase or decrease the effectiveness of this survey methodis not well known Our results indicated that for VES for mostspecies detection probability was relatively constant throughoutthe sampling window (ie AprilndashNovember each year) althoughfor a few species (A fowleri H chrysoscelis H cinerea andL clamitans) detection probability steadily decreased into thefall months Furthermore although we showed that VES arebetter for some species we also highlighted that VES canprovide adequate occupancy estimation even when conductedoutside the breeding window For example for L clamitans andA fowleri VES conducted in early April resulted in a 60probability of detection and during MCS it was not until lateJune toward the end of the breeding season that detectionprobabilities reached 60 Similarly for H chrysoscelis VESconducted in early April resulted in an 80 probability ofdetection and for A crepitans detection probabilities remained~60 during the VES survey window for both of thesespecies detection probabilities comparable to 80 and 60respectively did not occur until late June Therefore we suggestthat spring and fall VES can detect species that vocalise mostlyduring the summer months

Management implications

The goals of a particular study should dictate the appropriatesurvey method however results of the present study highlightedthe importance of estimating method-specific detectionprobabilities which should prove especially useful in thedesign of multispecies surveys We recommend the use MCSfor anurans with broad calling windows and conspicuous

vocalisations and VES for anurans that either vocalisesporadically or discreetly Furthermore we recommend the useof VES throughout the activity season not just the breedingseason because we found that survey date was a largelyunimportant predictor of detection using VES Our results alsosuggested that VES can provide more flexibility for anuranresearchers because robust estimates of detection andoccupancy can be obtained outside a narrow breeding window

Acknowledgements

This research was supported with funding provided by the Broad RiverMitigation Trust Fund administered through the South CarolinaDepartment of Natural Resources Additional funding was provided by theDukeEndowment through theDavidsonResearch Initiative and theDavidsonCollege Department of Biology We thank E Eskew S Hunt L WitczakC Williams A Domske D Millican C Oldham R Bauer B Abbuhl andC Ruder for assistance in the field and J Micancin for his help in identifyingcricket frog vocal recordings For assistance with logistics related to theprojectwe thankGVaughanSBennettBPerry andVVejdaniFinallywethank D Ramsey and two anonymous reviewers for statistical assistanceand comments which greatly improved this manuscript

References

Adams M J Pearl C A Galvan S andMcCreary B (2011) Non-nativespecies impacts on pond occupancy by an anuran The Journal of WildlifeManagement 75 30ndash35 doi101002jwmg29

Bailey L L Simons T R and Pollock K H (2004) Estimating siteoccupancy and species detection probability parameters for terrestrialsalamandersEcologicalApplications14 692ndash702 doi10189003-5012

BalasC JEulissNH andMushetDM (2012) Influenceof conservationprograms on amphibians using seasonal wetlands in the prairie potholeregion Wetlands 32 333ndash345 doi101007s13157-012-0269-9

Beck CW andCongdon J D (1999) Effects of individual variation in ageand size at metamorphosis on growth and survivorship of southern toad(Bufo terrestris)metamorphsCanadian Journal of Zoology77 944ndash951doi101139cjz-77-6-944

Blair W F (1961) Calling and spawning seasons in a mixed population ofanurans Ecology 42 99ndash110 doi1023071933272

Brander S M Royle J A and Eames M (2007) Evaluation of the statusof anurans on a refuge in suburbanMaryland Journal of Herpetology 4152ndash60 doi1016700022-1511(2007)41[52EOTSOA]20CO2

Brandt B B (1936) The frogs and toads of eastern North Carolina Copeia1936 215ndash223 doi1023071436326

Brown G W Scroggie M P Smith M J and Steane D (2007) Anevaluation of methods for assessing the population status of thethreatened Alpine tree frog Litoria verreauxii alpina in southeasternAustralia Copeia 3 765ndash770 doi1016430045-8511(2007)2007[765AEOMFA]20CO2

BucklandSTBurnhamKP andAugustinNH (1997)Model selectionan integral part of inference Biometrics 53 603ndash618 doi1023072533961

Burbrink F T Phillips C A and Heske E J (1998) A riparian zone insouthern Illinois as a potential dispersal corridor for reptiles andamphibians Biological Conservation 86 107ndash115 doi101016S0006-3207(98)00054-8

Burnham K P and Anderson D R (2002) lsquoModel Selection andMultimodel Inference a Practical Information-theoretic Approachrsquo2nd edn (Springer-Verlag New York)

Conant R and Collins J T (1998) lsquoA Field Guide to Amphibians andReptiles Eastern andCentralNorthAmericarsquo 3rd edn (HoughtonMifflinCompany Boston MA)

CookR P Tupper TA Paton PWC andTimmBC (2011) Effects oftemperature and temporal factors on anuran detection probabilities at

Anuran detection probability Wildlife Research K

method-specific detection probabilities for this species were notpossible to estimate

The best supported model for 10 anuran species indicatedthat detection varied by survey method (ie VES or MCSTable 2 Appendix 1) for four of these species (P feriarumP crucifer G carolinensis and L sphenocephalus) detectionvaried by survey method and for six species the best-supportedmodel was determined from model averaging (A americanusA terrestris A crepitans H cinerea H chrysoscelis andL catesbeianus) For the remaining three species (A fowleriL palustris and L clamitans) survey method was unimportantand the best-supported model suggested either a constantprobability of detection (L palustris) or suggested thatdetection varied by date (L clamitans) or date and days sincerainfall (A fowleri Table 2)

For six species MCS resulted in a higher mean (se)probability of detection than did VES (A terrestris MCS 039(038) vs VES 028 015) H cinerea MCS 037 (017) vs VES027 (005)H chrysoscelisMCS045 (011) vsVES043 (008)P feriarum MCS 012 (004) vs VES 008 (001) P cruciferMCS 023 (007) vs VES 007 (001) and L catesbeianus MCS040 (009) vs VES 034 (006) Fig 2) Conversely for fourspecies VES resulted in a higher mean probability of detectionthan did MCS (A americanus VES 038 (004) vs MCS 036(005) A crepitans VES 060 (036) vs MCS 038 (018)G carolinensis VES 036 (027) vs MCS 009 (002) andL sphenocephalus VES 044 (039) vs MCS 008 (001)Fig 2) For three species however both methods were equallyeffective and detection probabilities were identical (A fowleri044 (025) L clamitans 046 (008) and L palustris 009(002) Fig 2)

In addition to survey method date was an important factorinfluencing detection probability and was included in thebest-supported model for 10 species (A americanusA fowleri A crepitans H cinerea H chrysoscelisP feriarum P crucifer L catesbeianus L clamitans andL sphenocephalus Table 2) For all summer-breeding frogsexcept A crepitans detection probability during the VESdecreased with increasing day of year (ie decreased fromspring to fall) which corresponded to between 12 April and18 November each year (Figs 3 4) Specifically from mid-April to mid-November of each year the estimated probabilityof detection during VES increased from 058 to 060 forA crepitans and decreased from 040 to 030 forL catesbeianus from 078 to 032 for A fowleri from 045 to019 for H cinerea from 073 to 048 for L clamitans and from080 to 017 for H chrysoscelis (Figs 3 4) For all summer-breeding frogs detection probability during the MCS callingwindow increased with increasing day of year whichcorresponds to between 9 April and 28 June each year (Figs 34) Specifically from early April to late June of each year theestimated probability of detection during MCS increasedfrom 025 to 060 for A crepitans from 022 to 060 forL catesbeianus from 021 to 095 for A fowleri from 0002to099 forH cinerea from022 to 080 forL clamitans and from020 to 078 for H chrysoscelis (Figs 3 4)

For all winter-breeding frogs detection probability duringthe active-search survey window varied slightly withincreasing day of year (Fig 5) the estimated probability ofdetection during VES decreased from 050 to 040 forL sphenocephalus and from 041 to 031 for A americanusand increased from 005 to 018 for P feriarum and from 003 to

100

090

080

070

060

050

040

030

020

010

Est

imat

ed p

roba

bilit

y of

det

ectio

n

0

A americanus

A fowleri

A terre

stris

A crepitans

H cinerea

H chrysoscelis

P feriarum

P crucifer

G carolinensis

L catesbeianus

L clamitans

L palustris

L sphenocephalus

Fig 2 Estimated conditional probability of detection (1 se) for each anuran species along the Broad and PacoletRivers South Carolina USA Estimates are based on repeat surveys using two different survey methods (MCS or VES)

Anuran detection probability Wildlife Research G

018 for P crucifer (Fig 5) For all winter-breeding frogsdetection probability during the calling-survey time spandecreased with increasing day of year (Fig 5) Specifically

from mid-February to mid-May of each year the estimatedprobability of detection during MCS decreased from 038 to002 for L sphenocephalus from 098 to 001 for P feriarum

10MCS

Est

imat

ed p

roba

bilit

y of

det

ectio

n

VES

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

9Apr

29Apr

19May

8Jun

28Jun

12Apr

1Jun

21Jul

19Sep

18Nov

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

Fig 3 Effect of date by survey method (MCS or VES) on conditional detection probability ofsummer-breeding anuran species along the Broad and Pacolet Rivers South Carolina USA Solid linesrepresent the relationship between date and detection probability and dashed lines are the 95 CIsfor estimates of the covariate effect

H Wildlife Research J C Guzy et al

from 089 to 005 for A americanus and from 098 to 001 forP crucifer (Fig 5)

Discussion

Our study has provided insight for monitoring anurans byimplementing a modelling framework where data from bothMCS and VES are evaluated simultaneously resulting inmethod-specific detection probabilities Our results providedstrong evidence of variation in anuran detection probabilitiesamong two different detection methods Specifically for 10 of13 species the best-supported model indicated that detectionvaried by survey method Consistent with previous studies (egGooch et al 2006 Steelman and Dorcas 2010) calling surveyswere effective at detecting a wide range of anurans Visualencounter surveys however were also effective at detecting awide range of anuran species and several species had a higherdetection probability with VES thanwithMCSVisual encountersurveys are likely to be under-used as a monitoring tool foranurans perhaps because of a lack of information on theirutility Instead VES are often used along with other methodsin surveys for focal or rare species where the intent is gearedtoward complete coverage of all life stages to increase detectionprobability (eg Drost and Fellers 1996 Burbrink et al 1998Lips 1999 Lehtinen and Galatowitsch 2001)

Although both MCS and VES are effective methods forcertain species in our study one method was most effective

Specifically VES represent a more effective method fordetecting A crepitans G carolinensis and L sphenocephaluswhich is perhaps reflective of variation in species life historyFor example G carolinensis and L sphenocephalus like mostanurans can often be observed via vocalisations made duringbreeding however the breeding windows are wide (particularlyfor L sphenocephalus) calling by both species is sporadic andfor G carolinensis calling is usually initiated only by heavyrains (Brandt 1936 Dodd 2013) In addition calls of bothG carolinensis and L sphenocephalus are easily masked bylouder species (eg H cinerea and P crucifer Conant andCollins 1998) Likely because both G carolinensis andL sphenocephalus are active during the non-breeding periodin a wide variety of aquatic and terrestrial habitats (Mitchelland Lannoo 2005 Dodd 2013) VES was a more reliable methodof detection Acris crepitans another species with higherdetection probabilities during VES is active and conspicuousduring the non-breeding season Several other anuran speciesmight also be more reliably detected using VES if theirvocalisations are difficult to observe and their non-breedinghabits permit detection (ie little grass frogs (Pseudacrisocularis) or species that exhibit explosive andor short-livedbreeding periods (eg Scaphiopus spp L sylvaticusL capito) In these cases precisely timing MCS during thisbrief window is challenging (Dostine et al 2013) and oftennot ideal because anuran call surveys are less effective duringheavy rain (Weir et al 2005)

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

9Apr

29Apr

19May

8Jun

28Jun

12Apr

1Jun

21Jul

19Sep

18Nov

MCS VES

Est

imat

ed p

roba

bilit

y of

det

ectio

n

Fig 4 Effect of date by survey method (MCS or VES) on conditional detection probability of summer-breeding anuran species along the Broad and Pacolet Rivers South Carolina USA Solid lines represent therelationship between date and detection probability and dashed lines are the 95 CIs for estimates of thecovariate effect

Anuran detection probability Wildlife Research I

We found that MCS represented a more effectivedetection method for H cinerea P feriarum P crucifer andL catesbeianus These species are common vocal and calls can

be heard from a considerable distance (Dorcas et al 2007) Fora few species however both survey methods were equallyeffective (ie for A fowleri L clamitans and L palustris) In

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

14Feb

10Mar

4Apr

24Apr

14May

12Apr

1Jun

21Jul

19Sep

18Nov

Est

imat

ed p

roba

bilit

y of

det

ectio

n

MCS VES

Fig 5 Effect of date by survey method (MCS or VES) on conditional detection probability of winter-breeding anuran species along the Broad and Pacolet Rivers South Carolina USA Solid lines representthe relationship between date and detection probability and dashed lines are the 95 CIs for estimates ofthe covariate effect

J Wildlife Research J C Guzy et al

addition given the marginally larger detection probabilities ofA americanus using VES and of H chrysoscelis using MCSfor these species we also consider both survey methods to besimilarly effective For both A americanus and A fowleriMCS and VES detection probabilities were very similar MCSare effective because males of each species produce loud easilyrecognisable calls (Dorcas et al 2007) but VES are also effectiveperhaps because of their conspicuous tadpoles which school inlarge groups (Lefcort 1998) and the tendency for large numbersof metamorphs to forage for several weeks around the edge ofthe aquatic habitat fromwhich they emerged (Beck and Congdon1999 Dodd 2013) Likewise VES had detection probabilitiessimilar to those of MCS for L clamitans L palustris andH chrysoscelis for L clamitans this may be because whenstartled they emit conspicuous alarm calls as they leap into thewater (Dorcas et al 2007) and L palustrismay call in any monthof the year in the southern United States (Conant and Collins1998) Finally for H chrysoscelis its tendency to callsporadically during the daytime likely accounts for thesimilarity in effectiveness of both survey methods

Datewas an important factor influencing detection probabilityof anurans particularly regarding call surveys For summer-breeding anurans detection probability steadily increasedtoward the end of the summer survey windows and likewisefor winter-breeding anurans detection probability was highesttoward the beginning of the winter survey windows Previousstudies have documented strong patterns in seasonal breedingactivity and breeding windows are well established (Droege andEagle 2005) However for VES the potential influence of dateto increase or decrease the effectiveness of this survey methodis not well known Our results indicated that for VES for mostspecies detection probability was relatively constant throughoutthe sampling window (ie AprilndashNovember each year) althoughfor a few species (A fowleri H chrysoscelis H cinerea andL clamitans) detection probability steadily decreased into thefall months Furthermore although we showed that VES arebetter for some species we also highlighted that VES canprovide adequate occupancy estimation even when conductedoutside the breeding window For example for L clamitans andA fowleri VES conducted in early April resulted in a 60probability of detection and during MCS it was not until lateJune toward the end of the breeding season that detectionprobabilities reached 60 Similarly for H chrysoscelis VESconducted in early April resulted in an 80 probability ofdetection and for A crepitans detection probabilities remained~60 during the VES survey window for both of thesespecies detection probabilities comparable to 80 and 60respectively did not occur until late June Therefore we suggestthat spring and fall VES can detect species that vocalise mostlyduring the summer months

Management implications

The goals of a particular study should dictate the appropriatesurvey method however results of the present study highlightedthe importance of estimating method-specific detectionprobabilities which should prove especially useful in thedesign of multispecies surveys We recommend the use MCSfor anurans with broad calling windows and conspicuous

vocalisations and VES for anurans that either vocalisesporadically or discreetly Furthermore we recommend the useof VES throughout the activity season not just the breedingseason because we found that survey date was a largelyunimportant predictor of detection using VES Our results alsosuggested that VES can provide more flexibility for anuranresearchers because robust estimates of detection andoccupancy can be obtained outside a narrow breeding window

Acknowledgements

This research was supported with funding provided by the Broad RiverMitigation Trust Fund administered through the South CarolinaDepartment of Natural Resources Additional funding was provided by theDukeEndowment through theDavidsonResearch Initiative and theDavidsonCollege Department of Biology We thank E Eskew S Hunt L WitczakC Williams A Domske D Millican C Oldham R Bauer B Abbuhl andC Ruder for assistance in the field and J Micancin for his help in identifyingcricket frog vocal recordings For assistance with logistics related to theprojectwe thankGVaughanSBennettBPerry andVVejdaniFinallywethank D Ramsey and two anonymous reviewers for statistical assistanceand comments which greatly improved this manuscript

References

Adams M J Pearl C A Galvan S andMcCreary B (2011) Non-nativespecies impacts on pond occupancy by an anuran The Journal of WildlifeManagement 75 30ndash35 doi101002jwmg29

Bailey L L Simons T R and Pollock K H (2004) Estimating siteoccupancy and species detection probability parameters for terrestrialsalamandersEcologicalApplications14 692ndash702 doi10189003-5012

BalasC JEulissNH andMushetDM (2012) Influenceof conservationprograms on amphibians using seasonal wetlands in the prairie potholeregion Wetlands 32 333ndash345 doi101007s13157-012-0269-9

Beck CW andCongdon J D (1999) Effects of individual variation in ageand size at metamorphosis on growth and survivorship of southern toad(Bufo terrestris)metamorphsCanadian Journal of Zoology77 944ndash951doi101139cjz-77-6-944

Blair W F (1961) Calling and spawning seasons in a mixed population ofanurans Ecology 42 99ndash110 doi1023071933272

Brander S M Royle J A and Eames M (2007) Evaluation of the statusof anurans on a refuge in suburbanMaryland Journal of Herpetology 4152ndash60 doi1016700022-1511(2007)41[52EOTSOA]20CO2

Brandt B B (1936) The frogs and toads of eastern North Carolina Copeia1936 215ndash223 doi1023071436326

Brown G W Scroggie M P Smith M J and Steane D (2007) Anevaluation of methods for assessing the population status of thethreatened Alpine tree frog Litoria verreauxii alpina in southeasternAustralia Copeia 3 765ndash770 doi1016430045-8511(2007)2007[765AEOMFA]20CO2

BucklandSTBurnhamKP andAugustinNH (1997)Model selectionan integral part of inference Biometrics 53 603ndash618 doi1023072533961

Burbrink F T Phillips C A and Heske E J (1998) A riparian zone insouthern Illinois as a potential dispersal corridor for reptiles andamphibians Biological Conservation 86 107ndash115 doi101016S0006-3207(98)00054-8

Burnham K P and Anderson D R (2002) lsquoModel Selection andMultimodel Inference a Practical Information-theoretic Approachrsquo2nd edn (Springer-Verlag New York)

Conant R and Collins J T (1998) lsquoA Field Guide to Amphibians andReptiles Eastern andCentralNorthAmericarsquo 3rd edn (HoughtonMifflinCompany Boston MA)

CookR P Tupper TA Paton PWC andTimmBC (2011) Effects oftemperature and temporal factors on anuran detection probabilities at

Anuran detection probability Wildlife Research K

018 for P crucifer (Fig 5) For all winter-breeding frogsdetection probability during the calling-survey time spandecreased with increasing day of year (Fig 5) Specifically

from mid-February to mid-May of each year the estimatedprobability of detection during MCS decreased from 038 to002 for L sphenocephalus from 098 to 001 for P feriarum

10MCS

Est

imat

ed p

roba

bilit

y of

det

ectio

n

VES

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

9Apr

29Apr

19May

8Jun

28Jun

12Apr

1Jun

21Jul

19Sep

18Nov

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

Fig 3 Effect of date by survey method (MCS or VES) on conditional detection probability ofsummer-breeding anuran species along the Broad and Pacolet Rivers South Carolina USA Solid linesrepresent the relationship between date and detection probability and dashed lines are the 95 CIsfor estimates of the covariate effect

H Wildlife Research J C Guzy et al

from 089 to 005 for A americanus and from 098 to 001 forP crucifer (Fig 5)

Discussion

Our study has provided insight for monitoring anurans byimplementing a modelling framework where data from bothMCS and VES are evaluated simultaneously resulting inmethod-specific detection probabilities Our results providedstrong evidence of variation in anuran detection probabilitiesamong two different detection methods Specifically for 10 of13 species the best-supported model indicated that detectionvaried by survey method Consistent with previous studies (egGooch et al 2006 Steelman and Dorcas 2010) calling surveyswere effective at detecting a wide range of anurans Visualencounter surveys however were also effective at detecting awide range of anuran species and several species had a higherdetection probability with VES thanwithMCSVisual encountersurveys are likely to be under-used as a monitoring tool foranurans perhaps because of a lack of information on theirutility Instead VES are often used along with other methodsin surveys for focal or rare species where the intent is gearedtoward complete coverage of all life stages to increase detectionprobability (eg Drost and Fellers 1996 Burbrink et al 1998Lips 1999 Lehtinen and Galatowitsch 2001)

Although both MCS and VES are effective methods forcertain species in our study one method was most effective

Specifically VES represent a more effective method fordetecting A crepitans G carolinensis and L sphenocephaluswhich is perhaps reflective of variation in species life historyFor example G carolinensis and L sphenocephalus like mostanurans can often be observed via vocalisations made duringbreeding however the breeding windows are wide (particularlyfor L sphenocephalus) calling by both species is sporadic andfor G carolinensis calling is usually initiated only by heavyrains (Brandt 1936 Dodd 2013) In addition calls of bothG carolinensis and L sphenocephalus are easily masked bylouder species (eg H cinerea and P crucifer Conant andCollins 1998) Likely because both G carolinensis andL sphenocephalus are active during the non-breeding periodin a wide variety of aquatic and terrestrial habitats (Mitchelland Lannoo 2005 Dodd 2013) VES was a more reliable methodof detection Acris crepitans another species with higherdetection probabilities during VES is active and conspicuousduring the non-breeding season Several other anuran speciesmight also be more reliably detected using VES if theirvocalisations are difficult to observe and their non-breedinghabits permit detection (ie little grass frogs (Pseudacrisocularis) or species that exhibit explosive andor short-livedbreeding periods (eg Scaphiopus spp L sylvaticusL capito) In these cases precisely timing MCS during thisbrief window is challenging (Dostine et al 2013) and oftennot ideal because anuran call surveys are less effective duringheavy rain (Weir et al 2005)

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

9Apr

29Apr

19May

8Jun

28Jun

12Apr

1Jun

21Jul

19Sep

18Nov

MCS VES

Est

imat

ed p

roba

bilit

y of

det

ectio

n

Fig 4 Effect of date by survey method (MCS or VES) on conditional detection probability of summer-breeding anuran species along the Broad and Pacolet Rivers South Carolina USA Solid lines represent therelationship between date and detection probability and dashed lines are the 95 CIs for estimates of thecovariate effect

Anuran detection probability Wildlife Research I

We found that MCS represented a more effectivedetection method for H cinerea P feriarum P crucifer andL catesbeianus These species are common vocal and calls can

be heard from a considerable distance (Dorcas et al 2007) Fora few species however both survey methods were equallyeffective (ie for A fowleri L clamitans and L palustris) In

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

14Feb

10Mar

4Apr

24Apr

14May

12Apr

1Jun

21Jul

19Sep

18Nov

Est

imat

ed p

roba

bilit

y of

det

ectio

n

MCS VES

Fig 5 Effect of date by survey method (MCS or VES) on conditional detection probability of winter-breeding anuran species along the Broad and Pacolet Rivers South Carolina USA Solid lines representthe relationship between date and detection probability and dashed lines are the 95 CIs for estimates ofthe covariate effect

J Wildlife Research J C Guzy et al

addition given the marginally larger detection probabilities ofA americanus using VES and of H chrysoscelis using MCSfor these species we also consider both survey methods to besimilarly effective For both A americanus and A fowleriMCS and VES detection probabilities were very similar MCSare effective because males of each species produce loud easilyrecognisable calls (Dorcas et al 2007) but VES are also effectiveperhaps because of their conspicuous tadpoles which school inlarge groups (Lefcort 1998) and the tendency for large numbersof metamorphs to forage for several weeks around the edge ofthe aquatic habitat fromwhich they emerged (Beck and Congdon1999 Dodd 2013) Likewise VES had detection probabilitiessimilar to those of MCS for L clamitans L palustris andH chrysoscelis for L clamitans this may be because whenstartled they emit conspicuous alarm calls as they leap into thewater (Dorcas et al 2007) and L palustrismay call in any monthof the year in the southern United States (Conant and Collins1998) Finally for H chrysoscelis its tendency to callsporadically during the daytime likely accounts for thesimilarity in effectiveness of both survey methods

Datewas an important factor influencing detection probabilityof anurans particularly regarding call surveys For summer-breeding anurans detection probability steadily increasedtoward the end of the summer survey windows and likewisefor winter-breeding anurans detection probability was highesttoward the beginning of the winter survey windows Previousstudies have documented strong patterns in seasonal breedingactivity and breeding windows are well established (Droege andEagle 2005) However for VES the potential influence of dateto increase or decrease the effectiveness of this survey methodis not well known Our results indicated that for VES for mostspecies detection probability was relatively constant throughoutthe sampling window (ie AprilndashNovember each year) althoughfor a few species (A fowleri H chrysoscelis H cinerea andL clamitans) detection probability steadily decreased into thefall months Furthermore although we showed that VES arebetter for some species we also highlighted that VES canprovide adequate occupancy estimation even when conductedoutside the breeding window For example for L clamitans andA fowleri VES conducted in early April resulted in a 60probability of detection and during MCS it was not until lateJune toward the end of the breeding season that detectionprobabilities reached 60 Similarly for H chrysoscelis VESconducted in early April resulted in an 80 probability ofdetection and for A crepitans detection probabilities remained~60 during the VES survey window for both of thesespecies detection probabilities comparable to 80 and 60respectively did not occur until late June Therefore we suggestthat spring and fall VES can detect species that vocalise mostlyduring the summer months

Management implications

The goals of a particular study should dictate the appropriatesurvey method however results of the present study highlightedthe importance of estimating method-specific detectionprobabilities which should prove especially useful in thedesign of multispecies surveys We recommend the use MCSfor anurans with broad calling windows and conspicuous

vocalisations and VES for anurans that either vocalisesporadically or discreetly Furthermore we recommend the useof VES throughout the activity season not just the breedingseason because we found that survey date was a largelyunimportant predictor of detection using VES Our results alsosuggested that VES can provide more flexibility for anuranresearchers because robust estimates of detection andoccupancy can be obtained outside a narrow breeding window

Acknowledgements

This research was supported with funding provided by the Broad RiverMitigation Trust Fund administered through the South CarolinaDepartment of Natural Resources Additional funding was provided by theDukeEndowment through theDavidsonResearch Initiative and theDavidsonCollege Department of Biology We thank E Eskew S Hunt L WitczakC Williams A Domske D Millican C Oldham R Bauer B Abbuhl andC Ruder for assistance in the field and J Micancin for his help in identifyingcricket frog vocal recordings For assistance with logistics related to theprojectwe thankGVaughanSBennettBPerry andVVejdaniFinallywethank D Ramsey and two anonymous reviewers for statistical assistanceand comments which greatly improved this manuscript

References

Adams M J Pearl C A Galvan S andMcCreary B (2011) Non-nativespecies impacts on pond occupancy by an anuran The Journal of WildlifeManagement 75 30ndash35 doi101002jwmg29

Bailey L L Simons T R and Pollock K H (2004) Estimating siteoccupancy and species detection probability parameters for terrestrialsalamandersEcologicalApplications14 692ndash702 doi10189003-5012

BalasC JEulissNH andMushetDM (2012) Influenceof conservationprograms on amphibians using seasonal wetlands in the prairie potholeregion Wetlands 32 333ndash345 doi101007s13157-012-0269-9

Beck CW andCongdon J D (1999) Effects of individual variation in ageand size at metamorphosis on growth and survivorship of southern toad(Bufo terrestris)metamorphsCanadian Journal of Zoology77 944ndash951doi101139cjz-77-6-944

Blair W F (1961) Calling and spawning seasons in a mixed population ofanurans Ecology 42 99ndash110 doi1023071933272

Brander S M Royle J A and Eames M (2007) Evaluation of the statusof anurans on a refuge in suburbanMaryland Journal of Herpetology 4152ndash60 doi1016700022-1511(2007)41[52EOTSOA]20CO2

Brandt B B (1936) The frogs and toads of eastern North Carolina Copeia1936 215ndash223 doi1023071436326

Brown G W Scroggie M P Smith M J and Steane D (2007) Anevaluation of methods for assessing the population status of thethreatened Alpine tree frog Litoria verreauxii alpina in southeasternAustralia Copeia 3 765ndash770 doi1016430045-8511(2007)2007[765AEOMFA]20CO2

BucklandSTBurnhamKP andAugustinNH (1997)Model selectionan integral part of inference Biometrics 53 603ndash618 doi1023072533961

Burbrink F T Phillips C A and Heske E J (1998) A riparian zone insouthern Illinois as a potential dispersal corridor for reptiles andamphibians Biological Conservation 86 107ndash115 doi101016S0006-3207(98)00054-8

Burnham K P and Anderson D R (2002) lsquoModel Selection andMultimodel Inference a Practical Information-theoretic Approachrsquo2nd edn (Springer-Verlag New York)

Conant R and Collins J T (1998) lsquoA Field Guide to Amphibians andReptiles Eastern andCentralNorthAmericarsquo 3rd edn (HoughtonMifflinCompany Boston MA)

CookR P Tupper TA Paton PWC andTimmBC (2011) Effects oftemperature and temporal factors on anuran detection probabilities at

Anuran detection probability Wildlife Research K

from 089 to 005 for A americanus and from 098 to 001 forP crucifer (Fig 5)

Discussion

Our study has provided insight for monitoring anurans byimplementing a modelling framework where data from bothMCS and VES are evaluated simultaneously resulting inmethod-specific detection probabilities Our results providedstrong evidence of variation in anuran detection probabilitiesamong two different detection methods Specifically for 10 of13 species the best-supported model indicated that detectionvaried by survey method Consistent with previous studies (egGooch et al 2006 Steelman and Dorcas 2010) calling surveyswere effective at detecting a wide range of anurans Visualencounter surveys however were also effective at detecting awide range of anuran species and several species had a higherdetection probability with VES thanwithMCSVisual encountersurveys are likely to be under-used as a monitoring tool foranurans perhaps because of a lack of information on theirutility Instead VES are often used along with other methodsin surveys for focal or rare species where the intent is gearedtoward complete coverage of all life stages to increase detectionprobability (eg Drost and Fellers 1996 Burbrink et al 1998Lips 1999 Lehtinen and Galatowitsch 2001)

Although both MCS and VES are effective methods forcertain species in our study one method was most effective

Specifically VES represent a more effective method fordetecting A crepitans G carolinensis and L sphenocephaluswhich is perhaps reflective of variation in species life historyFor example G carolinensis and L sphenocephalus like mostanurans can often be observed via vocalisations made duringbreeding however the breeding windows are wide (particularlyfor L sphenocephalus) calling by both species is sporadic andfor G carolinensis calling is usually initiated only by heavyrains (Brandt 1936 Dodd 2013) In addition calls of bothG carolinensis and L sphenocephalus are easily masked bylouder species (eg H cinerea and P crucifer Conant andCollins 1998) Likely because both G carolinensis andL sphenocephalus are active during the non-breeding periodin a wide variety of aquatic and terrestrial habitats (Mitchelland Lannoo 2005 Dodd 2013) VES was a more reliable methodof detection Acris crepitans another species with higherdetection probabilities during VES is active and conspicuousduring the non-breeding season Several other anuran speciesmight also be more reliably detected using VES if theirvocalisations are difficult to observe and their non-breedinghabits permit detection (ie little grass frogs (Pseudacrisocularis) or species that exhibit explosive andor short-livedbreeding periods (eg Scaphiopus spp L sylvaticusL capito) In these cases precisely timing MCS during thisbrief window is challenging (Dostine et al 2013) and oftennot ideal because anuran call surveys are less effective duringheavy rain (Weir et al 2005)

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

9Apr

29Apr

19May

8Jun

28Jun

12Apr

1Jun

21Jul

19Sep

18Nov

MCS VES

Est

imat

ed p

roba

bilit

y of

det

ectio

n

Fig 4 Effect of date by survey method (MCS or VES) on conditional detection probability of summer-breeding anuran species along the Broad and Pacolet Rivers South Carolina USA Solid lines represent therelationship between date and detection probability and dashed lines are the 95 CIs for estimates of thecovariate effect

Anuran detection probability Wildlife Research I

We found that MCS represented a more effectivedetection method for H cinerea P feriarum P crucifer andL catesbeianus These species are common vocal and calls can

be heard from a considerable distance (Dorcas et al 2007) Fora few species however both survey methods were equallyeffective (ie for A fowleri L clamitans and L palustris) In

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

14Feb

10Mar

4Apr

24Apr

14May

12Apr

1Jun

21Jul

19Sep

18Nov

Est

imat

ed p

roba

bilit

y of

det

ectio

n

MCS VES

Fig 5 Effect of date by survey method (MCS or VES) on conditional detection probability of winter-breeding anuran species along the Broad and Pacolet Rivers South Carolina USA Solid lines representthe relationship between date and detection probability and dashed lines are the 95 CIs for estimates ofthe covariate effect

J Wildlife Research J C Guzy et al

addition given the marginally larger detection probabilities ofA americanus using VES and of H chrysoscelis using MCSfor these species we also consider both survey methods to besimilarly effective For both A americanus and A fowleriMCS and VES detection probabilities were very similar MCSare effective because males of each species produce loud easilyrecognisable calls (Dorcas et al 2007) but VES are also effectiveperhaps because of their conspicuous tadpoles which school inlarge groups (Lefcort 1998) and the tendency for large numbersof metamorphs to forage for several weeks around the edge ofthe aquatic habitat fromwhich they emerged (Beck and Congdon1999 Dodd 2013) Likewise VES had detection probabilitiessimilar to those of MCS for L clamitans L palustris andH chrysoscelis for L clamitans this may be because whenstartled they emit conspicuous alarm calls as they leap into thewater (Dorcas et al 2007) and L palustrismay call in any monthof the year in the southern United States (Conant and Collins1998) Finally for H chrysoscelis its tendency to callsporadically during the daytime likely accounts for thesimilarity in effectiveness of both survey methods

Datewas an important factor influencing detection probabilityof anurans particularly regarding call surveys For summer-breeding anurans detection probability steadily increasedtoward the end of the summer survey windows and likewisefor winter-breeding anurans detection probability was highesttoward the beginning of the winter survey windows Previousstudies have documented strong patterns in seasonal breedingactivity and breeding windows are well established (Droege andEagle 2005) However for VES the potential influence of dateto increase or decrease the effectiveness of this survey methodis not well known Our results indicated that for VES for mostspecies detection probability was relatively constant throughoutthe sampling window (ie AprilndashNovember each year) althoughfor a few species (A fowleri H chrysoscelis H cinerea andL clamitans) detection probability steadily decreased into thefall months Furthermore although we showed that VES arebetter for some species we also highlighted that VES canprovide adequate occupancy estimation even when conductedoutside the breeding window For example for L clamitans andA fowleri VES conducted in early April resulted in a 60probability of detection and during MCS it was not until lateJune toward the end of the breeding season that detectionprobabilities reached 60 Similarly for H chrysoscelis VESconducted in early April resulted in an 80 probability ofdetection and for A crepitans detection probabilities remained~60 during the VES survey window for both of thesespecies detection probabilities comparable to 80 and 60respectively did not occur until late June Therefore we suggestthat spring and fall VES can detect species that vocalise mostlyduring the summer months

Management implications

The goals of a particular study should dictate the appropriatesurvey method however results of the present study highlightedthe importance of estimating method-specific detectionprobabilities which should prove especially useful in thedesign of multispecies surveys We recommend the use MCSfor anurans with broad calling windows and conspicuous

vocalisations and VES for anurans that either vocalisesporadically or discreetly Furthermore we recommend the useof VES throughout the activity season not just the breedingseason because we found that survey date was a largelyunimportant predictor of detection using VES Our results alsosuggested that VES can provide more flexibility for anuranresearchers because robust estimates of detection andoccupancy can be obtained outside a narrow breeding window

Acknowledgements

This research was supported with funding provided by the Broad RiverMitigation Trust Fund administered through the South CarolinaDepartment of Natural Resources Additional funding was provided by theDukeEndowment through theDavidsonResearch Initiative and theDavidsonCollege Department of Biology We thank E Eskew S Hunt L WitczakC Williams A Domske D Millican C Oldham R Bauer B Abbuhl andC Ruder for assistance in the field and J Micancin for his help in identifyingcricket frog vocal recordings For assistance with logistics related to theprojectwe thankGVaughanSBennettBPerry andVVejdaniFinallywethank D Ramsey and two anonymous reviewers for statistical assistanceand comments which greatly improved this manuscript

References

Adams M J Pearl C A Galvan S andMcCreary B (2011) Non-nativespecies impacts on pond occupancy by an anuran The Journal of WildlifeManagement 75 30ndash35 doi101002jwmg29

Bailey L L Simons T R and Pollock K H (2004) Estimating siteoccupancy and species detection probability parameters for terrestrialsalamandersEcologicalApplications14 692ndash702 doi10189003-5012

BalasC JEulissNH andMushetDM (2012) Influenceof conservationprograms on amphibians using seasonal wetlands in the prairie potholeregion Wetlands 32 333ndash345 doi101007s13157-012-0269-9

Beck CW andCongdon J D (1999) Effects of individual variation in ageand size at metamorphosis on growth and survivorship of southern toad(Bufo terrestris)metamorphsCanadian Journal of Zoology77 944ndash951doi101139cjz-77-6-944

Blair W F (1961) Calling and spawning seasons in a mixed population ofanurans Ecology 42 99ndash110 doi1023071933272

Brander S M Royle J A and Eames M (2007) Evaluation of the statusof anurans on a refuge in suburbanMaryland Journal of Herpetology 4152ndash60 doi1016700022-1511(2007)41[52EOTSOA]20CO2

Brandt B B (1936) The frogs and toads of eastern North Carolina Copeia1936 215ndash223 doi1023071436326

Brown G W Scroggie M P Smith M J and Steane D (2007) Anevaluation of methods for assessing the population status of thethreatened Alpine tree frog Litoria verreauxii alpina in southeasternAustralia Copeia 3 765ndash770 doi1016430045-8511(2007)2007[765AEOMFA]20CO2

BucklandSTBurnhamKP andAugustinNH (1997)Model selectionan integral part of inference Biometrics 53 603ndash618 doi1023072533961

Burbrink F T Phillips C A and Heske E J (1998) A riparian zone insouthern Illinois as a potential dispersal corridor for reptiles andamphibians Biological Conservation 86 107ndash115 doi101016S0006-3207(98)00054-8

Burnham K P and Anderson D R (2002) lsquoModel Selection andMultimodel Inference a Practical Information-theoretic Approachrsquo2nd edn (Springer-Verlag New York)

Conant R and Collins J T (1998) lsquoA Field Guide to Amphibians andReptiles Eastern andCentralNorthAmericarsquo 3rd edn (HoughtonMifflinCompany Boston MA)

CookR P Tupper TA Paton PWC andTimmBC (2011) Effects oftemperature and temporal factors on anuran detection probabilities at

Anuran detection probability Wildlife Research K

We found that MCS represented a more effectivedetection method for H cinerea P feriarum P crucifer andL catesbeianus These species are common vocal and calls can

be heard from a considerable distance (Dorcas et al 2007) Fora few species however both survey methods were equallyeffective (ie for A fowleri L clamitans and L palustris) In

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

10

08

06

04

02

0

14Feb

10Mar

4Apr

24Apr

14May

12Apr

1Jun

21Jul

19Sep

18Nov

Est

imat

ed p

roba

bilit

y of

det

ectio

n

MCS VES

Fig 5 Effect of date by survey method (MCS or VES) on conditional detection probability of winter-breeding anuran species along the Broad and Pacolet Rivers South Carolina USA Solid lines representthe relationship between date and detection probability and dashed lines are the 95 CIs for estimates ofthe covariate effect

J Wildlife Research J C Guzy et al

addition given the marginally larger detection probabilities ofA americanus using VES and of H chrysoscelis using MCSfor these species we also consider both survey methods to besimilarly effective For both A americanus and A fowleriMCS and VES detection probabilities were very similar MCSare effective because males of each species produce loud easilyrecognisable calls (Dorcas et al 2007) but VES are also effectiveperhaps because of their conspicuous tadpoles which school inlarge groups (Lefcort 1998) and the tendency for large numbersof metamorphs to forage for several weeks around the edge ofthe aquatic habitat fromwhich they emerged (Beck and Congdon1999 Dodd 2013) Likewise VES had detection probabilitiessimilar to those of MCS for L clamitans L palustris andH chrysoscelis for L clamitans this may be because whenstartled they emit conspicuous alarm calls as they leap into thewater (Dorcas et al 2007) and L palustrismay call in any monthof the year in the southern United States (Conant and Collins1998) Finally for H chrysoscelis its tendency to callsporadically during the daytime likely accounts for thesimilarity in effectiveness of both survey methods

Datewas an important factor influencing detection probabilityof anurans particularly regarding call surveys For summer-breeding anurans detection probability steadily increasedtoward the end of the summer survey windows and likewisefor winter-breeding anurans detection probability was highesttoward the beginning of the winter survey windows Previousstudies have documented strong patterns in seasonal breedingactivity and breeding windows are well established (Droege andEagle 2005) However for VES the potential influence of dateto increase or decrease the effectiveness of this survey methodis not well known Our results indicated that for VES for mostspecies detection probability was relatively constant throughoutthe sampling window (ie AprilndashNovember each year) althoughfor a few species (A fowleri H chrysoscelis H cinerea andL clamitans) detection probability steadily decreased into thefall months Furthermore although we showed that VES arebetter for some species we also highlighted that VES canprovide adequate occupancy estimation even when conductedoutside the breeding window For example for L clamitans andA fowleri VES conducted in early April resulted in a 60probability of detection and during MCS it was not until lateJune toward the end of the breeding season that detectionprobabilities reached 60 Similarly for H chrysoscelis VESconducted in early April resulted in an 80 probability ofdetection and for A crepitans detection probabilities remained~60 during the VES survey window for both of thesespecies detection probabilities comparable to 80 and 60respectively did not occur until late June Therefore we suggestthat spring and fall VES can detect species that vocalise mostlyduring the summer months

Management implications

The goals of a particular study should dictate the appropriatesurvey method however results of the present study highlightedthe importance of estimating method-specific detectionprobabilities which should prove especially useful in thedesign of multispecies surveys We recommend the use MCSfor anurans with broad calling windows and conspicuous

vocalisations and VES for anurans that either vocalisesporadically or discreetly Furthermore we recommend the useof VES throughout the activity season not just the breedingseason because we found that survey date was a largelyunimportant predictor of detection using VES Our results alsosuggested that VES can provide more flexibility for anuranresearchers because robust estimates of detection andoccupancy can be obtained outside a narrow breeding window

Acknowledgements

This research was supported with funding provided by the Broad RiverMitigation Trust Fund administered through the South CarolinaDepartment of Natural Resources Additional funding was provided by theDukeEndowment through theDavidsonResearch Initiative and theDavidsonCollege Department of Biology We thank E Eskew S Hunt L WitczakC Williams A Domske D Millican C Oldham R Bauer B Abbuhl andC Ruder for assistance in the field and J Micancin for his help in identifyingcricket frog vocal recordings For assistance with logistics related to theprojectwe thankGVaughanSBennettBPerry andVVejdaniFinallywethank D Ramsey and two anonymous reviewers for statistical assistanceand comments which greatly improved this manuscript

References

Adams M J Pearl C A Galvan S andMcCreary B (2011) Non-nativespecies impacts on pond occupancy by an anuran The Journal of WildlifeManagement 75 30ndash35 doi101002jwmg29

Bailey L L Simons T R and Pollock K H (2004) Estimating siteoccupancy and species detection probability parameters for terrestrialsalamandersEcologicalApplications14 692ndash702 doi10189003-5012

BalasC JEulissNH andMushetDM (2012) Influenceof conservationprograms on amphibians using seasonal wetlands in the prairie potholeregion Wetlands 32 333ndash345 doi101007s13157-012-0269-9

Beck CW andCongdon J D (1999) Effects of individual variation in ageand size at metamorphosis on growth and survivorship of southern toad(Bufo terrestris)metamorphsCanadian Journal of Zoology77 944ndash951doi101139cjz-77-6-944

Blair W F (1961) Calling and spawning seasons in a mixed population ofanurans Ecology 42 99ndash110 doi1023071933272

Brander S M Royle J A and Eames M (2007) Evaluation of the statusof anurans on a refuge in suburbanMaryland Journal of Herpetology 4152ndash60 doi1016700022-1511(2007)41[52EOTSOA]20CO2

Brandt B B (1936) The frogs and toads of eastern North Carolina Copeia1936 215ndash223 doi1023071436326

Brown G W Scroggie M P Smith M J and Steane D (2007) Anevaluation of methods for assessing the population status of thethreatened Alpine tree frog Litoria verreauxii alpina in southeasternAustralia Copeia 3 765ndash770 doi1016430045-8511(2007)2007[765AEOMFA]20CO2

BucklandSTBurnhamKP andAugustinNH (1997)Model selectionan integral part of inference Biometrics 53 603ndash618 doi1023072533961

Burbrink F T Phillips C A and Heske E J (1998) A riparian zone insouthern Illinois as a potential dispersal corridor for reptiles andamphibians Biological Conservation 86 107ndash115 doi101016S0006-3207(98)00054-8

Burnham K P and Anderson D R (2002) lsquoModel Selection andMultimodel Inference a Practical Information-theoretic Approachrsquo2nd edn (Springer-Verlag New York)

Conant R and Collins J T (1998) lsquoA Field Guide to Amphibians andReptiles Eastern andCentralNorthAmericarsquo 3rd edn (HoughtonMifflinCompany Boston MA)

CookR P Tupper TA Paton PWC andTimmBC (2011) Effects oftemperature and temporal factors on anuran detection probabilities at

Anuran detection probability Wildlife Research K

addition given the marginally larger detection probabilities ofA americanus using VES and of H chrysoscelis using MCSfor these species we also consider both survey methods to besimilarly effective For both A americanus and A fowleriMCS and VES detection probabilities were very similar MCSare effective because males of each species produce loud easilyrecognisable calls (Dorcas et al 2007) but VES are also effectiveperhaps because of their conspicuous tadpoles which school inlarge groups (Lefcort 1998) and the tendency for large numbersof metamorphs to forage for several weeks around the edge ofthe aquatic habitat fromwhich they emerged (Beck and Congdon1999 Dodd 2013) Likewise VES had detection probabilitiessimilar to those of MCS for L clamitans L palustris andH chrysoscelis for L clamitans this may be because whenstartled they emit conspicuous alarm calls as they leap into thewater (Dorcas et al 2007) and L palustrismay call in any monthof the year in the southern United States (Conant and Collins1998) Finally for H chrysoscelis its tendency to callsporadically during the daytime likely accounts for thesimilarity in effectiveness of both survey methods

Datewas an important factor influencing detection probabilityof anurans particularly regarding call surveys For summer-breeding anurans detection probability steadily increasedtoward the end of the summer survey windows and likewisefor winter-breeding anurans detection probability was highesttoward the beginning of the winter survey windows Previousstudies have documented strong patterns in seasonal breedingactivity and breeding windows are well established (Droege andEagle 2005) However for VES the potential influence of dateto increase or decrease the effectiveness of this survey methodis not well known Our results indicated that for VES for mostspecies detection probability was relatively constant throughoutthe sampling window (ie AprilndashNovember each year) althoughfor a few species (A fowleri H chrysoscelis H cinerea andL clamitans) detection probability steadily decreased into thefall months Furthermore although we showed that VES arebetter for some species we also highlighted that VES canprovide adequate occupancy estimation even when conductedoutside the breeding window For example for L clamitans andA fowleri VES conducted in early April resulted in a 60probability of detection and during MCS it was not until lateJune toward the end of the breeding season that detectionprobabilities reached 60 Similarly for H chrysoscelis VESconducted in early April resulted in an 80 probability ofdetection and for A crepitans detection probabilities remained~60 during the VES survey window for both of thesespecies detection probabilities comparable to 80 and 60respectively did not occur until late June Therefore we suggestthat spring and fall VES can detect species that vocalise mostlyduring the summer months

Management implications

The goals of a particular study should dictate the appropriatesurvey method however results of the present study highlightedthe importance of estimating method-specific detectionprobabilities which should prove especially useful in thedesign of multispecies surveys We recommend the use MCSfor anurans with broad calling windows and conspicuous

vocalisations and VES for anurans that either vocalisesporadically or discreetly Furthermore we recommend the useof VES throughout the activity season not just the breedingseason because we found that survey date was a largelyunimportant predictor of detection using VES Our results alsosuggested that VES can provide more flexibility for anuranresearchers because robust estimates of detection andoccupancy can be obtained outside a narrow breeding window

Acknowledgements

This research was supported with funding provided by the Broad RiverMitigation Trust Fund administered through the South CarolinaDepartment of Natural Resources Additional funding was provided by theDukeEndowment through theDavidsonResearch Initiative and theDavidsonCollege Department of Biology We thank E Eskew S Hunt L WitczakC Williams A Domske D Millican C Oldham R Bauer B Abbuhl andC Ruder for assistance in the field and J Micancin for his help in identifyingcricket frog vocal recordings For assistance with logistics related to theprojectwe thankGVaughanSBennettBPerry andVVejdaniFinallywethank D Ramsey and two anonymous reviewers for statistical assistanceand comments which greatly improved this manuscript

References

Adams M J Pearl C A Galvan S andMcCreary B (2011) Non-nativespecies impacts on pond occupancy by an anuran The Journal of WildlifeManagement 75 30ndash35 doi101002jwmg29

Bailey L L Simons T R and Pollock K H (2004) Estimating siteoccupancy and species detection probability parameters for terrestrialsalamandersEcologicalApplications14 692ndash702 doi10189003-5012

BalasC JEulissNH andMushetDM (2012) Influenceof conservationprograms on amphibians using seasonal wetlands in the prairie potholeregion Wetlands 32 333ndash345 doi101007s13157-012-0269-9

Beck CW andCongdon J D (1999) Effects of individual variation in ageand size at metamorphosis on growth and survivorship of southern toad(Bufo terrestris)metamorphsCanadian Journal of Zoology77 944ndash951doi101139cjz-77-6-944

Blair W F (1961) Calling and spawning seasons in a mixed population ofanurans Ecology 42 99ndash110 doi1023071933272

Brander S M Royle J A and Eames M (2007) Evaluation of the statusof anurans on a refuge in suburbanMaryland Journal of Herpetology 4152ndash60 doi1016700022-1511(2007)41[52EOTSOA]20CO2

Brandt B B (1936) The frogs and toads of eastern North Carolina Copeia1936 215ndash223 doi1023071436326

Brown G W Scroggie M P Smith M J and Steane D (2007) Anevaluation of methods for assessing the population status of thethreatened Alpine tree frog Litoria verreauxii alpina in southeasternAustralia Copeia 3 765ndash770 doi1016430045-8511(2007)2007[765AEOMFA]20CO2

BucklandSTBurnhamKP andAugustinNH (1997)Model selectionan integral part of inference Biometrics 53 603ndash618 doi1023072533961

Burbrink F T Phillips C A and Heske E J (1998) A riparian zone insouthern Illinois as a potential dispersal corridor for reptiles andamphibians Biological Conservation 86 107ndash115 doi101016S0006-3207(98)00054-8

Burnham K P and Anderson D R (2002) lsquoModel Selection andMultimodel Inference a Practical Information-theoretic Approachrsquo2nd edn (Springer-Verlag New York)

Conant R and Collins J T (1998) lsquoA Field Guide to Amphibians andReptiles Eastern andCentralNorthAmericarsquo 3rd edn (HoughtonMifflinCompany Boston MA)

CookR P Tupper TA Paton PWC andTimmBC (2011) Effects oftemperature and temporal factors on anuran detection probabilities at

Anuran detection probability Wildlife Research K

Cape Cod National Seashore Massachusetts USA implications forlong-term monitoring Herpetological Conservation and Biology 625ndash39

Crump M L and Scott N J Jr (1994) Visual encounter surveys InlsquoMeasuring and Monitoring Biological Diversity Standard Methods forAmphibiansrsquo (Eds W R Heyer M A Donnelly R W McDiarmidL C Hayek and M S Foster) pp 84ndash92 (Smithsonian InstitutionWashington DC)

Dahl C Novotny V Moravec J and Richards S J (2009) Beta diversityof frogs in the forests of New Guinea Amazonia and Europe contrastingtropical and temperate communities Journal of Biogeography 36896ndash904 doi101111j1365-2699200802042x

Doan T M (2003) Which methods are most effective for surveying rainforest herpetofauna Journal of Herpetology 37 72ndash81 doi1016700022-1511(2003)037[0072WMAMEF]20CO2

Dodd C K (2013) lsquoFrogs of the United States and Canadarsquo (JohnsHopkins University Press Baltimore MD)

DorcasME andGibbons JW (2008) lsquoFrogs andToads of the Southeastrsquo(The University of Georgia Press Athens GA)

DorcasME Price S JBeane JC andCross S S (2007) lsquoTheFrogs andToads of North Carolinarsquo (North Carolina Wildlife ResourcesCommission Raleigh NC)

Dorcas M E Price S J Walls S C and Barichivich W J (2010)Auditory monitoring of anuran populations In lsquoConservation andEcology of Amphibiansrsquo (Ed C K Dodd) pp 281ndash298 (OxfordUniversity Press Oxford UK)

Dostine P L Reynolds S J Griffiths A D and Gillespie G R (2013)Factors influencing detection probabilities of frogs in the monsoonaltropics of northern Australia implications for the design of monitoringstudies Wildlife Research 40 393ndash402

Droege S and Eagle P (2005) Evaluating calling surveys In lsquoAmphibianDeclines the Conservation Status of United States Speciesrsquo (EdM J Lannoo) pp 314ndash319 (University of California Press BerkeleyCA)

Drost C A and Fellers G M (1996) Collapse of regional frog fauna in theYosemite area of the California Sierra Nevada Conservation Biology 10414ndash425 doi101046j1523-1739199610020414x

Eskew E A Price S J and Dorcas M E (2012) Effects of river-flowregulation on anuran occupancy and abundance in riparian zonesConservation Biology 26 504ndash512 doi101111j1523-1739201201842x

Farmer A L Smith L L Castleberry S B and Gibbons J W (2009)A comparison of techniques for sampling amphibians in isolatedwetlands in Georgia USA Applied Herpetology 6 327ndash341 doi101163157075309X12470350858433

Fry J Xian G Jin S Dewitz J Homer C Yang L Barnes C HeroldN and Wickham J (2011) Completion of the 2006 National LandCover Database for the Conterminous United States PEampRS 77858ndash864

Gooch M M Heupel A M Price S J and Dorcas M E (2006) Theeffects of survey protocol on detection probabilities and site occupancyestimates of summer breeding anurans Applied Herpetology 3 129ndash142doi101163157075406776984211

Grover M C (2006) Comparative effectiveness of nighttime visualencounter surveys and cover object searches in detecting salamandersHerpetological Conservation and Biology 1 93ndash99

Heard GW Robertson P and ScroggieM P (2006) Assessing detectionprobabilities for the endangered growling grass frog (Litoria raniformis)in southern Victoria Wildlife Research 33 557ndash564 doi101071WR04080

Hines J E (2006) lsquoPRESENCE (Version 57) Software to Estimate PatchOccupancy and Related Parameters USGS-PWRCrsquo Available at httpwwwmbr-pwrcusgsgovsoftwarepresencehtml [verified 1 August2014]

Hunt S D Guzy J C Price S J Halstead B J Eskew EA andDorcasM E (2013) Responses of riparian reptile communities to dammingand urbanization Biological Conservation 157 277ndash284 doi101016jbiocon201208035

Knutson M G Richardson W B Reineke D M Gray B R ParmeleeJ R and Weick S E (2004) Agricultural ponds supportamphibian populations Ecological Applications 14 669ndash684 doi10189002-5305

Lefcort H (1998) Chemically mediated fright response in southerntoad (Bufo terrestris) tadpoles Copeia 1998 445ndash450 doi1023071447439

Lehtinen R M and Galatowitsch S M (2001) Colonization ofrestored wetlands by amphibians in Minnesota American MidlandNaturalist 145 388ndash396 doi1016740003-0031(2001)145[0388CORWBA]20CO2

Lips K R (1999) Mass mortality and population declines of anurans at anupland site in western Panama Conservation Biology 13 117ndash125doi101046j1523-1739199997185x

MacKenzie D I and Bailey L L (2004) Assessing fit of site occupancymodels Journal of Agricultural Biological and Ecological Statistics 9300ndash318

MacKenzie D L Nichols J D Lachman G B Droege S Royle J Aand Langtimm C A (2002) Estimating site occupancy rates whendetection probabilities are less than one Ecology 83 2248ndash2255doi1018900012-9658(2002)083[2248ESORWD]20CO2

MacKenzieD INichols JDRoyle JA PollockKHBaileyLL andHines J E (2006) lsquoOccupancy Estimation and Modeling InferringPatterns and Dynamics of Species Occurrencersquo (Academic Press SanDiego CA)

MacKenzie D I Nichols J D SeamansM E andGutieacuterrez R J (2009)Modeling species occurrence dynamics with multiple states andimperfect detection Ecology 90 823ndash835 doi10189008-01411

Mattfeldt S D andGrant E H C (2007) Are twomethods better than oneArea constrained transects and leaf litterbags for sampling streamsalamanders Herpetological Review 38 43ndash45

Miller D A Nichols J DMcClintock B T Grant E H C Bailey L Land Weir L A (2011) Improving occupancy estimation when twotypes of observational errors occur non-detection and speciesmisidentification Ecology 92 1422ndash1428 doi10189010-13961

MillerDAWeirLAMcClintockBTGrantEHCBaileyLL andSimons T R (2012) Experimental investigation of false positive errorsin auditory species occurrence surveys Ecological Applications 221665ndash1674 doi10189011-21291

Mitchell J C and Lannoo M J (2005) Gastrophryne carolinensis(Holbrook 1836) Eastern narrow-mouthed toad In lsquoAmphibiandeclines conservation status of United States speciesrsquo (EdM J Lannoo) pp 501ndash503 (University of California Press BerkeleyCA)

Nichols J D Bailey L L OrsquoConnell A F Jr Talancy N W CampbellE H Grant E H C Gilbert A T Annand E M Husband T P andHines J E (2008) Multiscale occupancy estimation andmodeling usingmultiple detection methods Journal of Applied Ecology 45 1321ndash1329doi101111j1365-2664200801509x

Parris KMNorton TW andCunninghamRB (1999) A comparison oftechniques for sampling amphibians in the forests of south-eastQueensland Australia Herpetologica 55 271ndash283

Pellet J and Schmidt B R (2005) Monitoring distributions using callsurveys estimating site occupancy detection probabilities and inferringabsence Biological Conservation 123 27ndash35 doi101016jbiocon200410005

Price S JMarksDR HoweRWHanowski J andNiemiG J (2005)The importance of spatial scale for conservation and assessment ofanuran populations in coastal wetlands of the western Great LakesLandscape Ecology 20 441ndash454 doi101007s10980-004-3167-6

L Wildlife Research J C Guzy et al

Price S J Cecala KK Browne RA andDorcasM E (2011) Effects ofurbanization on occupancy of stream salamandersConservation Biology25 547ndash555 doi101111j1523-1739201001627x

Roumldel M O and Ernst R (2004) Measuring and monitoring amphibiandiversity in tropical forests I An evaluation of methods withrecommendations for standardization Ecotropica 10 1ndash14

Semlitsch R D and Bodie J R (2003) Biological criteria for buffer zonesaround wetlands and riparian habitats for amphibians and reptilesConservation Biology 17 1219ndash1228 doi101046j1523-1739200302177x

Steelman C K and Dorcas M E (2010) Anuran calling surveyoptimization developing and testing predictive models of anurancalling activity Journal of Herpetology 44 61ndash68 doi10167008-3291

Ver Hoef J M (2012) Who invented the delta method The AmericanStatistician 66 124ndash127 doi101080000313052012687494

Vonesh J R Mitchell J C Howell K and Crawford A J (2010) Rapidassessments of amphibian diversity In lsquoAmphibian Ecology andConservation a Handbook of Techniquesrsquo (Ed C K Dodd Jr)pp 263ndash280 (Oxford University Press Oxford UK)

Weir L A Royle J A Nanjappa P and Jung R E (2005) Modelinganuran detection and site occupancy on North American AmphibianMonitoring Program (NAAMP) routes in Maryland Journal ofHerpetology 39 627ndash639

Woolbright L L (1985) Patterns of nocturnal movement and calling bythe tropical frog Eleutherodactylus coqui Herpetologica 41 1ndash9

Anuran detection probability Wildlife Research M

Appendix 1 Complete candidate model results for detection of anuran species along the Broad and Pacolet Rivers South Carolina USAThe same eight candidate models were constructed for each species from uncorrelated covariates Best supported models are in bold AIC Akaike informationcriteria DQAICc difference in QAICc relative to the top model w model weight c overdispersion parameter estimated by dividing model deviance of theobservedmodel bymean bootstrapmodel deviance K number of parameters in the model Day of year day of year each survey was conducted Days since rain

number of days since rain for a given survey Method anuran survey method (VES or MCS)

Species Model AICc DQAICc w K ndash2 log-likelihood c

A americanus Y()Aq()Ap(Date)B 32364 000 057 4 31314Y()q()p(Date Rain) 32554 190 022 6 30754

Y()q()p(Method Date) 32599 235 018 5 31199Y()q()p(Method Date Rain) 32943 579 003 7 30681 139

Y()q()p(Method) 34712 2348 000 3 33971Y()q()p(Method Rain) 34905 2541 000 5 33505

Y()q()p() 35629 3265 000 2 35162Y()q()p(Rain) 35885 3521 000 4 34835

A fowleri Y()q()p(Date Rain) 64863 000 073 6 63063Y()q()p(Method Date Rain) 65058 195 027 7 62796 063

Y()q()p(Date) 66723 1860 000 4 65673Y()q()p(Method Date) 66780 1917 000 5 65380Y()q()p(Method Rain) 67185 2322 000 5 65785

Y()q()p(Rain) 67755 2892 000 4 66705Y()q()p(Method) 68956 4093 000 3 68215

Y()q()p() 69620 4757 000 2 69153

A terrestris Y()q()p() 5747 000 044 2 5280Y()q()p(Method) 5798 051 034 3 5057

Y()q()p(Method Rain) 6125 378 007 5 4725Y()q()p(Date Rain) 6149 402 006 6 4349

Y()q()p(Rain) 6250 503 004 4 5200Y()q()p(Date) 6293 546 003 4 5243

Y()q()p(Method Date) 6413 666 002 5 5013Y()q()p(Method Date Rain) 6560 813 001 7 4298 110

A crepitans Y()q()p(Method Date) 53038 000 055 5 51638Y()q()p(Method) 53149 111 031 3 52408

Y()q()p() 53435 397 008 2 52968Y()q()p(Method Rain) 53655 617 003 5 52255

Y()q()p(Method Date Rain) 53734 696 002 7 51472 278Y()q()p(Date) 53795 757 001 4 52745Y()q()p(Rain) 53850 812 001 4 52800

Y()q()p(Date Rain) 54365 1327 000 6 52565

H cinerea Y()q()p(Method Date) 35484 000 054 5 34084Y()q()p(Date) 35521 037 045 4 34471

Y()q()p(Date Rain) 36264 780 001 6 34464Y()q()p(Method Date Rain) 36334 850 001 7 34072 461

Y()q()p(Method) 40425 4941 000 3 39684Y()q()p(Method Rain) 41030 5546 000 5 39630

Y()q()p() 43567 8083 000 2 43100Y()q()p(Rain) 44122 8638 000 4 43072

H chrysoscelis Y()q()p(Date) 52682 000 058 4 51632Y()q()p(Method Date) 52900 218 019 5 51100Y()q()p(Date Rain) 52902 220 019 6 51502

Y()q()p(Method Date Rain) 53227 545 004 7 50965 630Y()q()p() 55047 2365 000 2 54580

Y()q()p(Rain) 55120 2438 000 4 54070Y()q()p(Method) 55130 2448 000 3 54389

Y()q()p(Method Rain) 55277 2595 000 5 53877

P feriarum Y()q()p(Method Date) 30614 000 097 5 29214Y()q()p(Method Rain Date) 31307 693 003 7 29045 621

Y()q()p(Date) 31939 1325 000 4 30889Y()q()p(Rain Date) 32575 1961 000 6 30775Y()q()p(Method) 38465 7851 000 3 37724

(continued next page)

N Wildlife Research J C Guzy et al

Appendix 1 (continued )

Species Model AICc DQAICc w K ndash2 log-likelihood c

Y()q()p(Method Rain) 39063 8449 000 5 37663Y()q()p() 39597 8983 000 2 39130

Y()q()p(Rain) 40113 9499 000 4 39063

P crucifer Y()q()p(Method Date) 41265 000 085 5 39865Y()q()p(Date) 41641 376 013 4 40591

Y()q()p(Method Date Rain) 42088 823 001 7 39826 446Y()q()p(Date Rain) 42332 1067 000 6 40532Y()q()p(Method) 43135 1870 000 3 42394

Y()q()p(Method Rain) 43770 2505 000 5 42370Y()q()p() 53728 12463 000 2 53261

Y()q()p(Rain) 54289 13024 000 4 53239

G carolinensis Y()q()p(Method) 15168 000 093 5 13768Y()q()p(Method Rain) 15810 642 006 7 13548

Y()q()p() 16065 897 001 3 15324Y()q()p(Rain) 16602 1434 000 5 15202Y()q()p(Date) 17007 1839 000 2 16540

Y()q()p(Rain Date) 17578 2410 000 4 16528Y()q()p(Method Date) 17580 2412 000 4 16530

Y()q()p(Method Rain Date) 18319 3151 000 6 16519 206

l catesbeianus Y()q()p(Date) 43926 000 044 4 42876Y()q()p(Method) 44010 084 029 3 43269

Y()q()p(Method Date) 44118 192 017 5 42718Y()q()p() 44305 379 007 2 43838

Y()q()p(Method Rain) 44578 652 002 5 43178Y()q()p(Date Rain) 44580 654 002 6 42780

Y()q()p(Rain) 44807 881 001 4 43757Y()q()p(Method Date Rain) 44877 951 000 7 42615 534

L clamitans Y()q()p(Date) 44153 000 058 4 43103Y()q()p(Date Rain) 44376 223 019 6 42576

Y()q()p(Method Date) 44474 321 012 5 43074Y()q()p() 44671 518 004 2 44204

Y()q()p(Rain) 44778 625 003 4 43728Y()q()p(Method Date Rain) 44821 668 002 7 42559 498

Y()q()p(Method) 44889 736 001 3 44148Y()q()p(Method Rain) 45054 901 001 5 43654

L palustris Y()q()p() 20071 000 065 2 19604Y()q()p(Method) 20341 270 017 3 19600Y()q()p(Date) 20515 444 007 4 19465Y()q()p(Rain) 20532 461 006 4 19482

Y()q()p(Method Date) 20676 605 003 5 19276Y()q()p(Method Rain) 20880 809 001 5 19480Y()q()p(Date Rain) 21137 1066 000 6 19337

Y()q()p(Method Date Rain) 21434 1363 000 7 19172 322

L sphenocephalus Y()q()p(Method Date) 48414 000 078 5 47014Y()q()p(Method) 48804 390 011 3 48063

Y()q()p(Method Rain Date) 48862 448 008 7 46600 661Y()q()p(Method Rain) 49111 697 002 5 47711

Y()q()p() 51439 3025 000 2 50972Y()q()p(Date) 51653 3239 000 4 50603Y()q()p(Rain) 51756 3342 000 4 50706

Y()q()p(Rain Date) 52138 3724 000 6 50338

AConstant probability of occupancy (ie Y() q ())BProbability of detection which varies by model for each species with or without inclusion of covariates that improve detection estimates

Anuran detection probability Wildlife Research O

wwwpublishcsiroaujournalswr

Appendix 1 (continued )

Species Model AICc DQAICc w K ndash2 log-likelihood c

Y()q()p(Method Rain) 39063 8449 000 5 37663Y()q()p() 39597 8983 000 2 39130

Y()q()p(Rain) 40113 9499 000 4 39063

P crucifer Y()q()p(Method Date) 41265 000 085 5 39865Y()q()p(Date) 41641 376 013 4 40591

Y()q()p(Method Date Rain) 42088 823 001 7 39826 446Y()q()p(Date Rain) 42332 1067 000 6 40532Y()q()p(Method) 43135 1870 000 3 42394

Y()q()p(Method Rain) 43770 2505 000 5 42370Y()q()p() 53728 12463 000 2 53261

Y()q()p(Rain) 54289 13024 000 4 53239

G carolinensis Y()q()p(Method) 15168 000 093 5 13768Y()q()p(Method Rain) 15810 642 006 7 13548

Y()q()p() 16065 897 001 3 15324Y()q()p(Rain) 16602 1434 000 5 15202Y()q()p(Date) 17007 1839 000 2 16540

Y()q()p(Rain Date) 17578 2410 000 4 16528Y()q()p(Method Date) 17580 2412 000 4 16530

Y()q()p(Method Rain Date) 18319 3151 000 6 16519 206

l catesbeianus Y()q()p(Date) 43926 000 044 4 42876Y()q()p(Method) 44010 084 029 3 43269

Y()q()p(Method Date) 44118 192 017 5 42718Y()q()p() 44305 379 007 2 43838

Y()q()p(Method Rain) 44578 652 002 5 43178Y()q()p(Date Rain) 44580 654 002 6 42780

Y()q()p(Rain) 44807 881 001 4 43757Y()q()p(Method Date Rain) 44877 951 000 7 42615 534

L clamitans Y()q()p(Date) 44153 000 058 4 43103Y()q()p(Date Rain) 44376 223 019 6 42576

Y()q()p(Method Date) 44474 321 012 5 43074Y()q()p() 44671 518 004 2 44204

Y()q()p(Rain) 44778 625 003 4 43728Y()q()p(Method Date Rain) 44821 668 002 7 42559 498

Y()q()p(Method) 44889 736 001 3 44148Y()q()p(Method Rain) 45054 901 001 5 43654

L palustris Y()q()p() 20071 000 065 2 19604Y()q()p(Method) 20341 270 017 3 19600Y()q()p(Date) 20515 444 007 4 19465Y()q()p(Rain) 20532 461 006 4 19482

Y()q()p(Method Date) 20676 605 003 5 19276Y()q()p(Method Rain) 20880 809 001 5 19480Y()q()p(Date Rain) 21137 1066 000 6 19337

Y()q()p(Method Date Rain) 21434 1363 000 7 19172 322

L sphenocephalus Y()q()p(Method Date) 48414 000 078 5 47014Y()q()p(Method) 48804 390 011 3 48063

Y()q()p(Method Rain Date) 48862 448 008 7 46600 661Y()q()p(Method Rain) 49111 697 002 5 47711

Y()q()p() 51439 3025 000 2 50972Y()q()p(Date) 51653 3239 000 4 50603Y()q()p(Rain) 51756 3342 000 4 50706

Y()q()p(Rain Date) 52138 3724 000 6 50338

AConstant probability of occupancy (ie Y() q ())BProbability of detection which varies by model for each species with or without inclusion of covariates that improve detection estimates

Anuran detection probability Wildlife Research O

wwwpublishcsiroaujournalswr