ESTIMATING BREEDING SEASON BIRD

DENSITIES FROM TRANSECT COUNTS

JOHN T. EMLEN

ABSTRACT.--In this paper I propose that for each speciesin an area the number of birds
detectedalonga transecttrail can be translatedinto an estimateof absolutedensity(birdsper unit
of area) by countingall the detectable(cue-producing)birds in a trailside strip narrow enoughto
permit detectionof all cuesproduced(the specificcensusstrip), and adjustingthis countfor the
undetectable(silent)birds in the strip by applying a locally and concurrentlyderived index of the
frequencyof cue productionfor the species.Cue frequencyis apparentlyimpossibleto measurein
nonbreedingbirds, but if all cues(sightings,calls, etc.) are used, most of the birds in the strip
presumablywill be detectedwhen the observeradvancesslowly enoughto allow each bird a good
chanceto make its presenceknown. In the nonbreedingseasonthe strip width is setfor eachspecies
at the distancefrom the trail at which total cue detectionstartsto decline.In the breedingseason
cue frequencymay be determinedfor songcues,and strip widths for eachspeciesset at a conve-
nient distancewithin the relatively great distanceat which songdetectionstartsto decline.Values
obtainedby thesetransectproceduresreflectthe densityof eachspeciesat the time the traverseis
run; a seriesof traversesmay be made to provide mean valuesfor selectedperiodsthroughthe
season.

The method is similar to that describedin an earlier paper. New or modified proceduresare
describedfor recordingdetections, establishing
specificstrip widths,bypassing the calculationof
"coefficientsof detectability,"estimatingdistancesin the field, determiningan optimum rate of
progress,and measuringthe frequencyof singingin a representativesampleof the population.
Problemsof convertingadjustedtransectcountsof singingmalesto total populationdensityand
of applyinga combinationof all-cueand song-cueproceduresto mixedpopulationsof breedingand
nonbreedingspeciesare discussed.
A comparisonof transect and plot map censusmethodsis presented.In the transect method
densityestimatesare of birdspresentat a particulartime ratherthan of birdswholly or partially
residentat one time or anotherduring a season.The samplingquadratsof transectcensuses are
elongaterectanglesspanningextensivetractsof habitat rather than truncatedblocksof representa-
tive habitat. The transectmethodis applicableat any seasonwhile the plot map methodcan be
usedonlyduringthe breedingseasonwhenbirdsare singingon territories.Problemsof reliabilityin
the plot map method stemmingfrom individual movementsduring a survey period and from
questionson how to interpret clustersof observationpoints on territory maps and how to evaluate
boundaryline territoriesare replacedin the transectmethodby problemsof accuracyin assuming
completecoveragein the all-cue operationsand in assigningbirds as insideor outsidethe lateral
boundarylinesof the censusstrips.More areacanbe coveredper unit of time in the transectthan in
the mappingmethod.---Department ofZoology,The Universityof Wisconsin,Madison, Wisconsin
53706. Accepted 3 November 1975.

THE PROBLEM AND THE MODEL

THE principal accessto absolute density values (birds per unit of area) for land
birds has been the spot-mapmethod (Williams 1936, Robbins 1970). While reason-
ably satisfactoryfor many purposes,spot-mappingproceduresare applicableonly
duringthe breedingseasonand on fairly smalltracts.Success in estimatingabsolute
densitiesin large areashas recentlybeen achievedby coordinatingabsolutedata
from small mappedplots with relative data from extensivetransectcounts(Enemar
and Sjostrand1967, 1970),and in 19711 describeda methodfor translatingcountsof
bird detectionsalonga transectroute to absolutedensityvaluesby applyingcorrec-
tion factors for each speciesbased on the distribution of detectionpoints laterally
from the transecttrail (Emlen 1971). Jiirvinen and Viiisiinen(1975) also used the
principle of decliningdetectabilitywith distance,applying three theoreticalregres-
455 The Auk 94: 455-468.July 1977
456 JOHNT. EMLEN [Auk,Vol. 94

sion curvesto the extensivenarrow and broad belt censusdata of Merikallio (1946,
1958) and others on Finnish bird populations.
There is no test of accuracyfor my 1971 distanceattenuation method, but results
appear to be reasonablysatisfactoryin the nonbreedingseasonwhen mostbirds close
to the trailside provide sound or visual cues as the observer passes.It is quite
unsatisfactoryduring the breeding season,however, when many individuals, nota-
bly nestingfemales,remain silent and essentiallyundetectableeven at closerange.
This paper is concernedprimarily with breedingseasoncensusproblemsbut reviews
various aspectsof nonbreedingseasontransectprocedures.
The relation between the number of cues tallied along a transect trail and the
absolutedensityof birds that number representsappearsto rest on two variables,
both of which can be controlledor measuredin many situations:(1) cue attenuation
with lateral distancefrom the trail and (2) the frequencywith which birds produce
visual or auditory cues detectable by an observer at any range. This paper
reexaminesand revisesthe proceduresfor attenuation control describedin my 1971
paper, outlinesproceduresfor measuringcue frequency, and discusses methodsfor
coordinatingthe two proceduresin varioussituations.It alsocomparesthe charac-
teristics,applicability, basicassumptions,advantages,and weaknessesof the tran-
sectand the traditionalplot-mapcensusmethods.The appendixpresentssomeresults
of preliminaryfield testsmade in Wisconsinusingthe two methods.
Cue attenuation.--My basic procedurefor controlling the attenuation of sound
cuesand visual cueswith distanceis to determinefor each speciesthe strip width
alongthe trail within which a fully alert observerwill detectessentiallyall cuesthat
are produced. In my 1971 paper I projected the density within this strip to an outer
boundaryline and usedthe ratio of the actual count insidethat boundaryline to the
projected total over many miles of transect as a coefficient of detectability for the
species.The countwithin the strip can be useddirectly as a specificcensusstrip, and
this procedurehas the advantageof beinglocalityand periodspecific.In either case
distancesperpendicularto the trail on either side are estimatedfor each bird as it is
detected,and tallied as dots or other symbolson surveysheetsin columnsrepresent-
ing narrow strips of terrain paralleling the trail. The symbolsin thesecolumnsare
totaledfor eachspeciesat the end of a surveyor seriesof surveys,and the accumu-
lated data are plotted as a regressioncurve with the transecttrail servingas the base.
Curvestypicallyshowfairly level basalplateausout to from 30 to 200 feet (9-60 m),
dependingon the species,beforedecliningrapidly or graduallyto zeroat the limits of
detection.
On the assumptionthat (1) I missrelatively few cuesin the proximal stripsim-
mediately adjacent to the trail, and (2) the plateau form of the curve indicatesthat
there is no appreciableadditional lossin detectionout to the inflection point of the
curve,I adopt the inflectionpoint or someconvenientpoint within it on eithersideof
the trail as marking the lateral boundariesof the specificcensusstrip for the species.
These lateral boundary lines and the ends of the transectroute define the areal base
for the densityfunction as well as delimiting the area in which cuescan be accepted
for densitydeterminations.Specificcensusstrips are thus elongatequadratswithin
which cue detectionapproachescompleteness. They must not be confusedwith the
areasusedin the flushingdistancemethodof King (Leopold 1933, Hayne 1949)in
which the estimateddistancesare alongradii emanatingforward and laterallyfrom
the advancing observer.
In additionto applyingthe specificcensusstrip directlyin place of the derived
July1977] Breeding
Densities
from Transect
Counts 457

coefficientof detectabilityI have madeseveralinnovationsor modifications

of pro-
ceduresincepresenting thetransectcensus modelin 1971:(1) To meettheproblemof
obtainingadequatesamplesfor an uncommonspeciesI may arbitrarilygroupthe
available data with those for one or more common speciesdisplaying similar cue-
attenuation characteristicsto derive an approximate value.
(2) Wherehabitatsoccurin narrowlinearshapesasalonga riverbank,a roadway,
or an urban city block, I adopt the natural boundaryof the habitat to definethe
censusstripexceptfor speciesin whichthe specificstripis narrowerthan the habitat
strip.
(3) Where birds are concentratedin flocks it is often difficult to tally each indi-
vidual as a separatedot on the surveysheet.Under theseconditionsI estimatethe
flock sizeand treat it as a unit, apportioningthe lateral distributionpointsaccording
to my bestestimateof the positionand dispersionof the flock at the momentit was
encountered.
(4) Where,asin the breedingseason,the detectabilityof membersof a population
fluctuatesrapidly and irregularly or varies stronglyamong individuals, I focuson
one or a few of the moststablecue types,suchas song,and basemy specificstrip
boundariesand calculationsof densityexclusivelyon these.Data obtainedby this
procedurerequirespecialadjustmentsfor cue frequencyas describedbelow.
Cuefrequency.--Cue attenuationshouldtheoreticallybe completelycontrolledby
the proceduresdescribedin the previous section, the observersimply basing his
density calculationsfor each specieson the count obtained in the relatively narrow
strip within which his tally of detectablecuesapproachescompleteness. But, entirely
aside from cue attenuation, individual birds may still be bypassedbecausethey
produce no detectable cues, either visible or audible, while the observer is within
detectionrange. These momentarilyundetectablebirds cannot be counteddirectly,
but their numbers can be computed if the frequency of cue production (the propor-
tion of observerencountersin which the birds make their presenceknown by emit-
ting detectablecues)is determined for a representativesample of the strip's popula-
tion and interpretedas the proportionof detectedindividualsin the total population
of the strip. Thus, in a hypotheticalcase,if 10 birds of a selectedspeciesare detected
on a transectcount run at a speedthat gives the observer6 min within the detection
range of each bird, and it is independentlydetermined that representativemembers
of the population make themselvesdetectablein 50% of a seriesof 6 min test periods
in which they are continuously within detection range, we can conclude that 10
additional birds were bypassedon the transect count and that the population in the
strip is 10 detectedplus 10 undetectedbirds = 20 birds.
The frequencyof cue productionis difficult to determine under most conditions
becausewhen an observerfollows them persistantly,birds tend to alter their natural
behavior in ways that make them less or more detectable. Total cue frequencies
(usingall visual and soundcues)are, in fact, essentiallyunattainable, but fortunately
during much of the nonbreedingseasonmostof the birds within the narrow specific
strip seemto make their presenceknown if the observeradvancesslowlyand restricts
his countsto favorable early morning conditions(Emlen 1971).
Cue frequenciescan apparently be estimatedwith reasonableaccuracyduring the
breedingseasonif calculationsare basedexclusivelyon song.Representativeterrito-
rial males may be selectedas samplesand watched continuouslyover extended
periods(Enemar 1959). Or, as it is difficult to recognizetruly representativebirds, a
458 JOHNT. EMLEN [Auk, Vol. 94

series of territorial males may be visited repeatedly for shorter periods and their
individual songfrequenciesaveraged (Hickey 1943).
When cue frequenciesare basedexclusivelyon songcues,transectcountsmust, of
course,alsobe restrictedto songs.The tally will thus be smaller, but the much wider
specificcensusstrips that can be employedwhen only loud vocalizationsare used in
determiningattenuationdistances,and the longerbird-observerexposuretimesavail-
able for each territorial singer compensatefor the omissionof numerousnonsong
cues from the calculations.

Density computation.--When cuefrequencycannot be measured,as when all cues

are usedduring the nonbreedingseason,densitymust be computedentirely from cue
attenuation data. When indices on cue frequency are available, as for song in the
breeding season,total density can be computedby multiplying the count of detected
(cue-producing)birds in the censusstrip by the reciprocal of the locally determined
cue frequency for the species.
The density values obtained by the specificstrip censusmethod, with or without
applicationof cue-frequencydata, apply to the populationpresentin the strip at the
time the traverse is run. Individual birds that drift back and forth across the bound-
ary lines are includedif they happento be inside, excludedif they are outsidewhen
the observerpasses.The effectsof suchtransboundarymovementswill presumably
balance out for common specieson long traverses, and fluctuations in a stable
populationshouldbe small in a seriesof standardizedtraversesover the sameroute.
Variations in computeddensity will occur, however, with changesin weather and
variations in field proceduresuchas rate of progressor time of day.
When proceduresare standardized, density estimatesfor a seriesof specificstrip
traversesmay be averagedto reduceerrorscausedby small samplesizes,or statisti-
cally analysedfor information on the completenessof cue detectionunder various
conditions.Palmgren(1930)discussed the averagingof transect-deriveddensityval-
uesfor an area, notingthat in open(no strip boundary)transectsor wherewide fixed
stripsare used,the largestcountfor a speciesin a seriesof traverseswill approachthe
actual populationlevel more closely(be more complete)than the mean for the series.
This principle appliesin situationswhere the populationbeingcountedis assumedto
be definitive and where variations in the count are due to variations in the complete-
nessor efficiencyof the counting;it doesnot apply to specificstrip countswherethe
populationbeingcountedfluctuatesas birds drift back and forth acrossthe boundary
line and where all countsare assumedto be essentiallycompleteor at leastrepresen-
tative of the birdspresentwithin the indicatedboundarieswhen the countwas made.

FIELD PROCEDURI•S FOR THE NONBREEDING SEASON

To estimate densitiesof nonsinging,nonterritorial populationsalong a transect

route one should use all available cues and follow the proceduresdescribedabove
under cue attenuation. The values will theoretically be complete for all cue-
producing(detectable)birds in the specificcensusstrips. Silent and inactive (unde-
tectable)birdswill inevitablybe bypassed,and no satisfactorytechniquehasyet been
devisedfor estimating them. In the absenceof data on cue frequency, best guess
adjustments(basal detectabilityadjustments)may be made for these undetected
birds where best estimatesare preferable to minimum estimates.
July 1977] BreedingDensities
from TransectCounts 459

FIELD PROCEDURES FOR BREEDING SEASON POPULATIONS

To estimatebreedingseasondensitiesby the specificstrip methodoneshoulduse

only songcuesand then adjust the talliesof singingmalesfor the unrecordednon-
singingmalesand females.This procedurecallsfor two separateoperations in the
fie]d, (1) tallyingall songdetections and their lateraldistancesalongthe route,(2)
determiningindicesof songfrequencyduringthe censusperiod.Althoughthesetwo
operations are functionallydistinct,the data for eachmay be collectedconcurrently
alonga transectroutewithoutprejudiceto eithersetof dataandwithoutlossof time.
Countingsingingmales.--The field procedures for songtransectsare similarto
thoseusedin all-cue transectsas describedin my 1971 paper, but involve a number
of specialconsiderations as discussed below:
(1)Recordall detections.--Althoughdensitycalculationsin this modelare based
entirely on songcues,all detectedcuesshouldbe recorded.Songcuesshouldbe
clearlydifferentiatedfrom the otherson the tally sheetby somedistinctivesymbol
such as a small circle or the letter s.
(2)Songstripboundaries.--Because songsin mostspecies canbeheardat relatively
greatdistances, the basalplateausof lateralregression curvesaremuchbroaderwhen
based exclusivelyon song cues than when soft call notes and sightingsare also
included. Under thesecircumstances the problemsof distanceestimationand density
calculationcan be simplified by setting the boundary lines for the censusstrip at
some convenientarbitrary distancewell inside the limits of song detectionset by
soundattenuation. In the Wisconsintest study (seeAppendix) I selected200 feet on
either side of the trail (a 400-foot strip) for most speciesand 100 feet (200-footstrip)
for a few quiet-voicedspecies.
(3) Distance estimates.--As distances from the transect trail to unseen singing
birds can rarely be measured,a subjectiveapproachis necessary.Elaborateestima-
tion techniquesmust be avoided, however, as they involve distractionsthat can
affect the efficiencyof distant songdetectionadversely(Merikallio 1946). Fortu-
nately if a fixed distancewell within the absolute limit of detectionis set, as advo-
cated in the precedingparagraph, the only critical decisionto be made for each
observationis relatively simple, whether or not the bird is inside or beyond that
prescribedcensusstrip boundary line when first detected. In most casesthe correct
answer is subjectivelyobvious,but there may be a good many borderlinecases.In
any event, every census-takermust face the subjectivityproblem squarelyand work
out a systemfor himself in which he can test his performancelevel objectively at
frequentintervals.I find that with practiceI can almostinvariably predictto within
10 or 15% the number of paces(3 feet) it will take me to reach a selectedfixed object
200 feet away, the distanceto the strip boundary line for most speciesin breeding
seasontransects.This level of accuracyI am obligedto acceptas the bestI can do. In
making theseestimatesfor self-testingand in actual transectcountsituations,I find
it helpful to cultivate and retain mental images of familiar settings with known
dimensions,such as a room in my home, a tennis court baseline, a fallen 100-foot
tree, or a 100-yard race track straightaway.
To apply these acquired skills to an unseen songbird along a transect route one
must first determinethe approximatelocation of the sourceof the soundwith refer-
ence to someconspicuousand fixed object in the habitat such as a distinctivetree
trunk or a tall shrub, and then estimatethe distanceof that object from the trail
when he is approximatelyoppositeit. Both the locatingand the estimatingoperations
46O JOHN T. EMLEN [Auk, Vol. 94
TABLE 1
SOME COMPARISONS OF THE PLoT-MAP AND TRANSECT-STRIP CENSUS METHODS

Plot-map method Transect-stripmethod

Objectives
To estimate the number of birds resident To estimate the number of birds present during
during the breeding season. a single censusoperation.
Nature of the data
Data units are the territories lying within or Data units are the individual detections of birds
partially within (fractions)the plot boundaries. as the observer moves along the route.
Each visit to the plot contributes data to a Each traverse of the route provides the com-
single censusestimate for the season. plete record for a definitive estimate; the results
of repeated traverses can be averaged.
Plots are usually truncated in shape. Plots are elongate in shape (strips).
Applicability
Limited to the season when birds are on ter- Applicable at any season.
ritories.
Plots must be replicated when objective is to A long transect plot provides a representative
characterizea regionor vegetationtype. samplingof a region or vegetationtype.
Problemsof procedureand interpretation
(Repeated traverses over the same area will In the absence of cue frequency data (non-
reduce errors of omission.) breeding season) an unknown number of silent
(no cue) birds within detection range (the specific
strip) will be bypassedin the single traverse that
constitutes a definitive transect census.
(No distanceestimatesare required.) Lateral distance measurements to detection
points are only rough subjective estimates.
Double recording of individuals is difficult to (Double recording is rarely a problem when fol-
control unlessneighboringmales sing concur- lowing a straight transect course at more than
rently. 0.70 mph.)
Determination of territory boundarieson base (Precise boundary determinations are not re-
maps may be difficult especially where ter- quired.)
ritories are contiguous.
Determination of the fraction of boundary (No territory evaluations are required.)
line territories lying within a plot requires
knowledgeof the transboundaryextensions.
Individual territorial birds may enter, leave (Intervisit changes are minimal when intervals
or shift within a plot betweenspacedvisits. are short.)
Efficiency (hypothetical)in hours
For a 24-acre plot (birds residentthrough the For a 24-acre (0.5 mix 400 feet) strip segment
breeding season) (birds present during three 5-day periods)
Staking and mapping -- 10 hr. Mapping -- 2 hr.
Vegetation survey -- 4 hr. Vegetation survey -- 4 hr.
8 surveys x 2 hr. -- 16 hr. 15 traverses x 0.6 hr. -- 9 hr.
Total -- 30 hr. Total -- 15 hr.

are best accomplishedby moving along the trail and sightingtowards the object or
the soundsourcefrom severalspacedpoints.A clearly visible and reasonablystraight
trail is important as a referencebasefor this operation.Major landmarks previously
plotted to scaleon a strip map of the route (seenext subsection)greatly facilitate all
distance estimates in the tract.
(4) Rate of advance.--The rate of progressalong the trail and the distanceahead
and to the rear within which songsshould be recorded are critical insofar as they
determinethe length of time the observeris exposedto eachbird on the censusstrip.
A net walking speedof about 0.75 mph combined with a 200-foot limit for recording
birds, fore and aft, allows 6 min for each bird. Where 200 feet is also used as the
lateral distanceto the strip boundary(seeconsideration 2 above),the observeris, in
effect, concentratinghis attention on the birds in a slowly advancing400-foot square
area in which he is centered.
July 1977] BreedingDensitiesfrom TransectCounts 461

TABLE 2
CENSUS DATA FOR A 48-ACRE STAND OF MIXED WOODLAND IN WISCONSIN •

Total Song Terri-

Total count in count in tories in
count per Width of specific 200-foot Song 200-foot
mile2 specificstrip3 strip4 strip5 frequenc? strip?
Mourning Dove
(Zenaida rnacroura) 2.0 40 + 40 0.5 0.64 0.40 2.0
Yellow-billed Cuckoo
(Coccyzusamericanus) 0.1 200 + 200 0.2s 0.09 v8 --
Common Flicker
(Colapresauratus) 1.9 60 + 60 0.9 0.45 0.15 3.1
Red-bellied Woodpecker
(Melanerpes carolinus) 0.4 100 + 100 0.2 -- -- 0.5
Downy Woodpecker
(Picoidespubescens) 2.2 100 + 100 1.3 -- -- 4.0
Great CrestedFlycatcher
(Myiarchus crinitus) 0.6 200 + 200 0.2 -- -- 0.4
Blue Jay
(Cyanocittacristata) 13.0 100 + 100 8.4 -- -- 8.2
Black-cappedChickadee
(Parus atricapillus) 16.2 60 + 60 10.4 -- -- 8.8
White-breasted Nuthatch
(Sitta carolinensis) 0.2 200 + 200 0.1 -- -- 0.2
Red-breasted Nuthatch
(S. canadensis) 1.1 100 + 100 0.9 -- -- 1.0
House Wren
(Troglodytesaedon) 3.1 200 + 200 3.8s 1.91 0.66 3.0
Gray Catbird
(Durnetella carolinensis) 5.2 40 + 40 1.6 1.68 0.44 7.9
Brown Thrasher
(Toxostornaruffurn) 4.3 60 + 60 2.9 0.91 0.19 6.0
American Robin
(Turdusrnigratorius) 6.3 40 + 40 3.3 0.91 0.13 8.1
Wood Thrush
(Hylocichla rnustelina) 3.0 200 + 200 3.5s 1.77 0.52 3.6
Cedar Waxwing
(Bornbycillacedrorurn) 0.6 50 + 50 0.6 -- -- 1.0
European Starling
(Sturnusvulgaris) 0.3 100 + 100 0.2 -- v --
Common Yellowthroat
(Geothlypistrichas) 0.6 200 + 200 0.6s 0.32 0.27 1.0
Red-winged Blackbird
(Agelaiusphoeniceus) 0.4 200 + 200 0.5s 0.23 0.60 --
Northern Oriole
(Icterus galbula) 0.3 100 + 100 0.3 0.09 0.10 0.8
Common Grackle
(Quiscalusquiscula) 7.7 40 + 40 4.2 -- v --
Brown-headed Cowbird
(Molothrus ater) 4.7 50 + 50 2.2 1.95 0.51 3.5
Cardinal
(Cardinalis cardinalis) 6.4 60 + 60 3.0 2.91 0.59 7.0
Rose-breasted Grosbeak
(Pheucticusludovicianus) 2.4 50 + 50 0.8 0.86 0.28 4.2
Indigo Bunting
(Passerinacyanea) 0.8 200 + 200 0.8s 0.41 0.53 0.9

• Twentytraverses alonga 1.10-miletransectroutethrougha 48-acrestandin Madison,18 Juneto 7 July 1974.

2 All birdsdetected
by sightings,
callsor songs---per
mile.
3 Distancebetweeninflectionpointsfor thespecieson eachsideof trall•in feet.
4 All birdsdetected
bysightings,
callsor songs
withinthespecific
strip,or songs
detected
(indicated
bys) x 2, whichever
islarger.
s All malesdetectedby songwithin200 feet(100feetfor Catbird)--permile.
s Proportionof 6 min. territorycrossings
in whichtheresidentbirdsang.
? Sumof wholeor fractionalterritories as determinedby clustersof pointsrepresenting
songsiteswithin200 feet of trail on date
when the speciespopulationwas at maximum.
s v = nonresidentvagrants.
462 JOHN T. EMLEN [Auk, Vol. 94

TABLE 2•Continued

Total Song Terri-

Total count in count in tories in
count per Width of specific 200-foot Song 200-foot
mile2 specificstrip3 strip4 strips frequency
6 strip?
American Goldfinch
(Spinus tristis) 1.0 50 + 50 0.5 0.18 0.25 0.9
Rufous-sided Towhee
(Pipilo erythrophthalmus) 3.1 50 + 50 1.0 1.73 0.72 0.8
Field Sparrow
(Spizella pusilla) 2.6 200 + 200 2.7 1.36 0.60 2.8
Song Sparrow
(Melospiza melodia) 0.4 200 + 200 0.4s 0.14 0.75 0.3

(5) Number oftraverses.--Because the calculated values obtained in thesetransect

censusesapply to the number of birds in the strip at the moment of counting,
traversesover a routemay be repeatedand averaged.To avoid complicationsrelated
to seasonalor breeding cycle changes,such traverse replicationsshould be made
within a limited period, ideally on successive
days. For reasonsexplainedin the next
subsection,the number of replicationsin a seriesis limited to about 5 when song
frequency measurementsare involved. Double (10) or triple (15) seriesmay, of
course,be used.Where a compositerecordfor an entire breedingseasonis desired,as
in traditional spot-mappingcensuses,severalseriesof countswill be needed,perhaps
one 5-day seriesevery 2 weeks.
Determining songfrequencies.--The conversion of transect counts of singing
males to population densitiesrequiresmeasuresof mean songfrequencyfor the area
and seasonin which the counts were made. A series of observation periods at ter-
ritories located along the censusroute can provide a record of the mean incidenceof
singing by their resident occupants.This operation can be coordinatedwith the
transect count operation to provide song incidence recordsefficiently for the same
population and periods as thosecovered by the count.
A simple strip map of the route (scaleabout 1:2400)showingprominent landmarks
to 200 feet laterally shouldbe prepared at the start. I carry a set of such strip maps,
one for each residentspecies,on my clipboard beneaththe censustally sheetfor the
day. On thesemapsI plot the positionof eachdetectionpoint as a coloredsymbol
and draw lines to indicate a bird's movementsfrom perch to perch. I use a different
coloron the mapsfor eachtraverse,and asI can clearlydiscriminateonlyfive colors
of small dots on a map, I set five as the number of traversesin a series.Each map
thus providesthe completerecordfor a speciesover a seriesof 5 traverses.Obvious
clustersof differentlycoloredsymbolsand linesdelineatethe territoriesof localized
males on thesemaps, and songfrequenciescan be read directly as the number of
color-distinctsongsymbolsin a cluster(from 0 to 5) divided by the number of visits
or checksof the territory in a series•always five. In this systema specieswith six
recognizable territoriesalonga stripprovidesa samplesizeof 6 x 5 = 30 songchecks.
Mean frequencyvaluesobtainedin this way will be too high if the countof territories
is incompletebecauseof the presenceof nonsingingmales(zerofrequency)that fail to
reveal themselvesby any cue during the five visits to their territory, or that are
detectedby nonsongcuesonly onceor twice and classedas nonterritorialbirds.
The additional time neededfor the double entry of detectionsin this coordinated
transect-songcheck procedureis negligible, and the attention neededfor carefully
July1977] Breeding
Densities
fromTransect
Counts 463

TABLE 3

THREE DENSITY ESTIMATES (BIRDS PER 100 ACRES)BASED ON THE CENSUS DATA PRESENTED IN
TABLE 2 •

Transectmethod

Based on song cues

Based on all and adjusted for
detectedcues2 songfrequency
3 Plot-mapmethod
4
Mourning Dove 4.8 6.6 7.5
Yellow-billed Cuckoo 0.4 v v
Common Flicker 6.0 12.4 11.7
Red-bellied Woodpecker 0.9 -- 1.9
Downy Woodpecker 5.2 -- 15.0
Great Crested Flycatcher 0.4 -- 1.5
Blue Jay 17.2s -- 30.1
Black-cappedChickadee 25.1' -- 33.1
White-breasted Nuthatch 0.2 -- 0.8
Red-breasted Nuthatch 3.4 -- 3.8
House Wren 7.8 11.9 11.7
Gray Catbird 15.7 31.4 29.7
Brown Thrasher 20.1 19.7 22.8
American Robin 32.2 28.8 30.8
Wood Thrush 7.4 14.0 13.7
Cedar Waxwing 4.9 -- 4.1
European Starling 0.9 v v
Common Yellowthroat 1.3 4.9 3.8
Red-winged Blackbird 1.0 1.6 1.3
Northern Oriole 1.1 3.7 3.4
Common Grackle 41.0 v v
Brown-headed Cowbird 18.4 15.77 20.77
Cardinal 21.0 20.3 26.3
Rose-breasted Grosbeak 6.4 12.6 15.8
Indigo Bunting 1.7 3.2 3.4
American Goldfinch 3.7 3.0 3.8
Rufous-sided Towhee 7.5 6.08 4.98
Field Sparrow 5.7 9.4 10.5
Song Sparrow 0.7 0.8 1.1
No valuesare givenin column2 for nonsinging
species
and in columns2 and 3 for species
that wererepresented
onlyby vagrants(v).
Calculated for 100 acres from column 3 in Table 2.
Calculatedfrom column4 in Table 2 and adjustedfor songfrequency(column5) and for undetectedfemales.
Calculatedfrom column6 in Table 2 and adjustedfor undetectedfemales.
Fledgedjuveniles(est. 2 per averagefamily flock of 4) have beensubtracted.
Fledgedjuveniles(est.4 per averagefamilyflockof 6) havebeensubstracted.
Observedsexratiosof cowbirdssuggest an averageof abouttwo femmesper mme.Male territorieswereve• largeanddifficultto plot.
Only oneof the two singingmaleson the tract waspaired.

placingthe symbolon the map is complementaryto, rather than competitivewith

that neededfor assigninga lateral distancevalue on the tally sheet.
Conversionto absolutedensity estimates.--The adjustment or conversionfactor
for a countof singingmalesin a song-cuetransectstrip is the reciprocalof the song
incidencefor that populationand period. This holds for both high and low song
frequencies,a low frequencysimplyindicatingthe needfor a largeadjustment.The
method is thus applicableover wide rangesof singingactivity.
Densitiescomputedfor a song-cue stripshouldbe convertedto standardunitssuch
as birds per 100 acresor squarekilometers.Where the strip width for a specieshas
been set as 200 feet on either side of the trail, one mile of strip will cover 48.5 acres.
Conversionto birds per 100 acresin this caseis accomplished
by multiplyingthe
densityin the strip by 100/48.5 = 2.06.
To obtaintotal population
densityfor a species,
the valueobtainedfor male
density must be adjusted for the uncountedfemales. For monogamoussinging
speciesin the Wisconsintest area I appliedthe imprecisebut not unreasonable
assumptions that the songtalliesreflectedbothresidentand vagrantmales,and that
464 JOHN T. EMLEN [Auk, Vol. 94

TABLE 4
SINGING INCIDENCE OF TERRITORIALXV[ALES
AT XV[ADISON,
WISCONSIN,DURINGFIVE PERIODSBE-
TWEEN 18 JUNE AND 17 AUGUST t974. •

21 July-
18-29 June 1-7 July 14-19 July 5 August 12-17 August
Inc. N Inc. N Inc. N Inc. N Inc. N

MourningDove 0.37 (30) 0.43 (40) 0.43 (30) 0.27 (30) 0.33 (30)
CommonFlicker 0.30 (30) 0.15 (40) 0.18 (40) + X + X
House Wren 0.70 (50) 0.64 (80) 0.50 (60) 0.60 (20) -- X
Gray Catbird 0.50 (tOO) 0.38 (90) 0.31 (tt0) -- (90) -- (120)
BrownThrasher 0.18 (80) 0.20 (60) 0.03 (80) -- (60) -- (70)
AmericanRobin 0.06 (80) 0.19 (90) 0.23 (90) 0.40 (t0) -- X
Wood Thrush 0.58 (80) 0.43 (60) 0.43 (80) 0.23 (60) 0.03 (30)
CommonYellowthroat 0.35 (20) 0.20 (20) 0.40 (20) 0.20 (t0) -- X
Brown-headedCowbird 0.60 (50) 0.37 (30) -- X -- X -- X
Cardinal 0.64 (80) 0.54 (80) 0.41 (80) 0.35 (60) 0.40 (50)
Rose-breastedGrosbeak 0.45 (40) 0. t0 (50) 0.04 (50) -- X -- X
Indigo Bunting 0.65 (20) 0.35 (20) 0.30 (t0) -- X -- X
AmericanGoldfinch 0. t0 (20) 0.15 (20) 0.35 (20) 0.13 (30) 0.10 (30)
Rufous-sidedTowhee 0.80 (20) 0.65 (20) 0.80 (20) 0.70 (t0) 0.50 (20)
Field Sparrow 0.70 (30) 0.50 (30) 0.37 (30) 0.55 (20) 0.35 (20)
SongSparrow 0.80 (t0) 0.70 (t0) 0.20 (10) 0. t0 (t0) -- X
• Valuesare the proportions
of 4-6-min.earlymorningvisitsto (crossings through)territories
duringwhichthe residentbird sang.
Numbersin parenthesesgivethesamplesizefor eachvalue(territoriesx visits).

the overall sex ratio in the populationswas roughly equal. On this basisI simply
multiplied the computedmale densityby two.
Procedures for nonsingingspecies.--A number of speciesin a breedingcommunity
such as the woodpeckersand jays may have nothing equivalent to the loud and
frequent advertisement songsof most song birds, yet remain localized as pairs or
small flocks for at least part of the breedingseason.Such speciescan be treated as
nonterritorial birds by recording all cues and assumingnearly complete cue fre-
quency within the specificcensusstrip of the species(the nonbreedingseasonproce-
dure) or, when thesebirds are foragingin flocksor pairs on delimited home ranges,
they can be sampledfor total cuefrequencyin the samemannerthat singingspecies
are sampledfor songfrequency.The frequencyvalue can then be appliedto the tally
of pairs or flocks within the censusstrip of the speciesto provide a densityestimate
for pair or flock units. Adjustment of this estimateto total adult densityfor the
speciesmay then be accomplishedby multiplying the number of flocks by an inde-
pendently derived value of mean flock size. This procedurealleviatesthe practical
problem of counting the individuals in each flock when encounteredin the field.
FIELD PROCEDURESFOR MIXED AND TRANSITION POPULATIONS

The seasonaltransition from breedingto nonbreedingconditionand back is

gradualin populationsof any givenspecies,and an avian communitycharacteristi-
cally containsboth breedingand nonbreedingspeciesthroughmuch of the year.
Thus a censustaker will oftenbe confrontedwith a mixtureof species,someneeding
the all-cuemethodwithoutfrequencyadjustmentsand otherseligiblefor the song-
cuemethodincorporating measurements of songfrequency.The choicewill generally
be determined by theuniformityandfrequency of cueproductionby thebirdsat the
time,andtheopportunities availablefor recognizing
andkeepingtabsonindividual
birdsas requiredby cue-frequency measurement procedures.With field operations
standardizedand restrictedto optimumweatherconditions,reasonablyhighunifor-
mity andfrequencyof cueproductioncanbe assumed for mostspeciesthroughmuch
of the nonbreeding seasonand, for some,throughoutthe year. Opportunities for
July 1977] Breeding Densitiesfrom TransectCounts 465

individual recognitionare provided when individuals isolate themselveson distinct

and exclusiveterritorieswhere they can be visited and checkedperiodically, and this
occursfor many speciesin the breeding seasonand for a few throughout the year.
As communities often contain representativesof both categoriessimultaneously
and as a speciesmay change from one category to the other rather rapidly, field
proceduresshouldbe designedto coverthe data requirementsfor each. This raisesno
seriousproblems, and a tally sheetcan be planned that provides spacefor recording
all the pertinent information for eachprocedure. Strip maps are, of course,required
for song-frequency measurements and should be included with the tally sheet
whenever the use of this procedure for one or more speciesseems indicated. The
choicebetween simple cue attenuation (all cues)and song-frequencyprocedurescan
then be made after the fieldwork is completedwith full data in hand. Where density
values are obtained for a speciesby both methods simultaneously, a selectionbe-
tween the two can be made on the basis of size and clarity of the data samples
supporting each and on considerationsof the basic reliability of the two procedures
(seeAppendix).

DISCUSSION

The transectmethoddiffers from the familiar territory mappingmethod(seeRob-

bins 1970) in the nature of the density values obtained, in aspectsof reliability and
accuracy, and in overall efficiency. The summary and comparison of the two
methodspresentedin Table 1 may be usefulin selectingthe bestapproachfor various
types of ecological and behavioral studies. Some of the major considerationsare
discussedin greater detail below.
Density values.--Transect censusesprovide data on the number of birds of each
specieson a transectplot (strip) at the time the traverse is made. Repeated traverses
along the sameroute within the span of a week or two provide replicate samplesof
the same population suitable for averaging and other statisticaltreatments. By con-
trast, the territory mapping method provides a compositerecord of the number of
individualsof eachspeciesresidenton the selectedplot at one time or anotherduring
a breeding season.Repeatedvisits to the plot increasethe completenessof the record
but do not constitute replications and cannot be averaged. In the transect method
time-bracketedseriesmay be repeatedon the sameplot at spacedintervalsto provide
data for an overall record for the season,while in the territory mapping method,
provided adequate data are collected,the record may be broken down by periodsto
provide information on direct speciesassociationsor on changesin community struc-
ture and distribution as the season advances.
Where the objectiveof a censusoperationis to determinemean densityvaluesover
a large area or an extensivehabitat type, a long rectangularstrip transectingthe
area, as provided by the transect method, will produce a better sampling than a
compact, truncated plot (of the same size). Where the objective is to measure the
populations on a small island or an isolated block of distinctive habitat too small to
accommodatean elongatetransect strip,.the mapping method is preferable.
Where a record of seasonalchangesthrough a full year is desired the territory
mapping method is inapplicable. The specific-striptransectmethod as describedin
this report may be applied, althoughthe necessarychangesof procedurebetweenthe
breedingand nonbreedingseasonsmay give rise to errors.
Reliability and accuracy.--Territories that overlap censusplot boundary lines
present problems for evaluating densitiesby the territory mapping method. This
466 JOHNT. EMLEN [Auk, Vol. 94

problem assumesmajor proportionswhen species'territories are large in relation to

the censusplot. Unlessinformation is obtained on the boundariesof suchterritories
outside the plot, the fraction inside, and hence the number of birds represented,
cannotbe reliably evaluated. The problemof boundaryline territoriesis bypassedin
the transectstrip methodas the censusunitsare simplybirdspresentinsidethe strip
at the momentof counting.For songfrequencydeterminationsin the breedingseason
the samplescan be restrictedto residentbirds whoseterritorieslie acrossor near the
trail and who can thus be assumedto remain continuouslywithin hearing range.
A secondsourceof error in the territory mapping method is in the interpretation of
clustersof observationpointsas territories.This becomesparticularly difficult where
a speciesis abundant and territories are contiguous. Supplementary notations of
concurrent singing by neighboring territory holders and of behavioral interactions
betweensuchneighborsare very usefulin locatingboundarylinesin thesecases,but
interpretationsof the same set of data by several experiencedobserversmay still
differ considerably(Svensson1974, Best 1975, Mannes and Alpers 1975). Transect
countsdo not require any interpretation of territory boundaries, and the samplesfor
song-frequencymeasurementsmay be selectedjudiciously to avoid territories in
confusing situations.
While the transect method escapesthe hazards of misinterpreting boundary line
overlaps and point clusters on censusmaps, a fair comparisonof the two systems
must balance threats to reliability against the threats to accuracy posed by the
investigator'sinability to verify the two basicassumptionsof the variable-striptran-
sectmethod, completeness of countcloseto the trail, and even distributionlaterally
from the trail. Theoretically the former can be covered by adjustmentsfor cue
frequencyduring the breedingseason,but remainsa seriousfactor of unknown and
variable magnitudeat other times;the latter must be controlledas far as possibleby
selectingcensustracts with broad stretchesof essentiallyuniform habitat structure.
Efficiency.--The relative efficiencyof the two methodsin terms of time and effort
is difficult to judge becausea singlecompositedensityvalue for a seasoncannotbe
equatedreadily with a seriesof time-bracketedvalues distributedthrough the same
season. Using hypothetical values, however, I estimate in Table 1 that to obtain a
single compositedensity value by the mapping method requires roughly twice as
many hours as a seriesof three time-bracketed mean density values based on five
transectseach. This apparently greater efficiency of the transect method must, of
course,be weighedagainstconsiderationsof the nature of the valuesdesiredfor any
particular study.

ACKNOWLEDGMENTS

Much of the fieldwork underlyingthe developmentof this methodwas donewith supportfrom grants
from the National ScienceFoundationand The Frank M. ChapmanMemorial Fund. In preparingthis
report I benefitted much from critical commentsand suggestionsby JosephJ. Hickey, Margaret B.
Hickey, Olli J•irvinen,and ChandlerS. Robbins.This paperis publicationNo. 79 of the Universityof
Wisconsin Arboretum.

LITERATURE CITED

BEST, L.B. 1975. Interpretationalerrorsin the "mappingmethod"as a censustechnique.Auk 92:

452-460.
EMLEN, J.T. 1971. Population densitiesof birds derived from transectcounts.Auk 88: 323-342.
ENEMAR,A. 1959. On the determinationof the sizeand compositionof a passefinebird population
duringthe breedingseason.V•r F•gelv•rld, Suppl. 2: 1-114.
July1977] Breeding
Densities
fromTransect
Counts 46 7

--, ANDB. SJOSTRAND.1967. The strip surveyas a complementto studyarea investigations

in
bird censuswork. V• Fftgelv•irld26: 111-130.
1970. Bird speciesdensitiesderivedfrom studyarea investigations
and land transects.Bull.
Ecol. Res. Comm. No. 9, Lund.
HAYNE, D.W. 1949. An examinationof the strip censusmethodfor estimatinganimal populations.J.
Wildl. Mgmt. 13: 145-157.
HICKEY, J.J. 1943. A guide to bird watching. London, Oxford Univ. Press.
JJ•RVINEN,O. AND R. VJ•IS•NEN. 1975. Estimating relative densitiesof breeding birds by the line
transect method. Oikos 26: in press.
LEOPOLD,A. 1933. Game management.New York, Charles Scribner'sSons.
M•NNES,P. ANDR. ALPERS.1975. 0ber FehlergriSssen
bei Siedlungsdichte-Untersuchungen
an
hiShlenbriitenden
SingviSgeln
nach der Kartierungsmethode.J. fiir Ornithol. 116: 308-314.
MERIKALLIO,
E. 1946. 0berregionaleVerbreitungundAnzahlderLandviSgel in Siid-und
Mittelfinn-
land, besonders
in deren0stlichenTeilen, im Lichte yon quantitativenUntersuchungen. I.
AllgemeinerTeil. Ann. Zool. Soc.Vanarno,Tom 12(1):1-140.
1958. Finnish birds. Their distribution and numbers. Fauna Fennica 5: 1-181.
PALMGREN,
P. 1930. QuantitativeUntersuchungen
Ober die Vogelfaunain den W'•ildernSiidfinn-
lands. Acta Zool. Fennica 7: 1-218.
ROBBINS,C.S. 1970. Recommendations
for an internationalstandardfor a mappingmethodin bird
census work. Audubon Field Notes 24: 723-726.
SVENSSON,S. 1974. Interpersonalvariation in speciesmap evaluation in bird censuswork with the
mapping method. Acta Ornithol. 14(23): 322-338.
WILLIAMS, A.B. 1936. The compositionand dynamics of a beech-mapleclimax community. Ecol.
Monogr. 6: 317-408.

APPENDIX

As a test for the breedingseasontransectmethod describedin this paper I conducteda field study
combiningtransectand plot map methodson a tract of mixed woodland in the University of Wisconsin
arboretum at Madison, Wisconsinin the summer of 1974. I ran 20 traversesalong a 1.10 mile transect
routethroughthe tract between18Juneand 7 July, and added30 traversesbetween8 July and 17August.
The censusdata collected on the first 20 traversesare presentedin Table 2, and density estimates
calculatedfrom them by the total-cue method, the adjusted song-cuemethod, and the plot map method
are presented in Table 3.
All traverses(1 or 2 per day) were made during the first 3 h of daylight while walking at an average
speed of 0.70 to 0.80 mph along a well-marked trail that looped through the tract. All detectionsof
movements,call notes,and songsfor eachspecieswere tallied on preparedsheetsin columnsrepresenting
10-footstripsto 100feet, thena 100-footand a 200-footstrip to 400 feet. All detectionswerealsorecorded
as coloredsymbolson strip maps of the route to provide the basisfor song-frequencymeasurements.A
separatemap was usedfor eachfive traversesfor eachspecies,and the symbolsfor each traversewere re-
cordedin differentcolors.The songfrequencyfor a species(column5 in Table 2) wascalculatedfrom these
maps by multiplyingthe number of different coloredsongsymbolsin each selectedterritory (territory
crossings during which the bird was singing)by the numberof selectedterritorieson the map. The total
number of territories(column6 in Table 2) was my best estimateof the sum of whole and fractional ter-
ritories lying within the strip.
The values derived by the adjusted song-cueand plot-map methods(columns2 and 3 respectivelyin
Table 3) correspondcloselyfor most species.As the two are basedon different setsof data this correspon-
dencegivescredenceto the possibilitythat both reflectthe actual densityduringthe censusperiodquite
well. It also suggeststhat the length of the censusperiod, 20 days, was not long enough to reveal any
appreciabledifferencesbetween the maximum density level during the period as measuredby the plot-
map methodand the mean densitylevel for the period as measuredby the transectmethod. Which of the
two setsof resultsis more accuratein terms of the objectivesof its respectivemethodrestson the error
sourcesinherent in the two methods as consideredin the discussionsection of this paper.
Values obtainedby the total cuetransectmethod(not adjustedfor cuefrequency)presentedin column 1
of Table 3 are, with two exceptions,lower than thoseobtainedby the other methods,and in a few cases
substantiallylower. This, of course,is to be expectedduring the breedingseasonwhen residentbirds are
highly irregular in cue production.In somespeciesit may be attributed in large part to the low detectabil-
ity of female during the breedingseason,but in at leastthe sevenspecieswhere songdetections(multi-
pliedby 2 to coverfemales)wereusedbecausetheygavehighervaluesthan the unadjustedvaluesbasedon
468 JOHNT. EMLEN [Auk, Vol. 94

all cues(seefootnote3 in Table 2) it clearlyinvolved nonsingingresidentmales. When adjustedfor song

frequency(column5 in Table 2) thesevaluesare equatedwith the song-cuetransectvalues. The total-cue
and song-cuemethods,of course,cover nonterritorialbirds deliberatelyomitted in the plot-map method.
The song-frequency indicesused for transectcensuses in this paper (occurrencein 4-6 rain periods)
presumablyreflect species-characteristic behavioral traits that will vary in more or lesspredictablepat-
ternsfor eachspecieswith time of day and stageof the nestingcycle.Individual variationswill inevitably
occurbut, roughlystandardizedfor time of day when frequenciesare not changingrapidly, meanvalues
for the populationson a censustract may be expectedto show predictableprogressivechangesas the
breeding seasonadvances.If this prediction can be verified with the accumulationof data, it may be
possibleto apply valuesfor a specifiedsegmentof the seasonto songcountsalonga censusroutewithout
recordinglocal songfrequenciesfor every operation.
Song frequenciesfor 16 specieson the Madison censustract are presentedin Table 4 for five periods
between 18 June and 17 August of 1974. Ten traverseswere run in each period to provide samplesizesof
10x the number of sample territories for each species.Incidence values (frequencies)declinedfor most
speciesas the seasonadvanced.Recordsfor early June would doubtlessreveal higherfrequenciesfor these
species.Irregular fluctuationspresumablyreflect the smallnessof the samplesizes.