10
139
THS

WWNINIHNIHMIIHIHIHHNIWIWHIHHNHW

  

   
     

  

LIBRARY

Michigan State . ‘
University

lllllllllllllllllll ll WM

3 129119720 3261..

...\..\‘...-—--

This is to certify that the
thesis entitled

A Comparison of the Emlen Transect Technique

and a Time-Area Count.

presented by

Dennis P. Fjalkowski

has been accepted towards fulfillment
of the requirements for

Master of Science Fisheries and Wildlife

degree in

 

 

 

Date September 12, 1978

 

0-7639

 

j“
x._ I

 

 

 

 

 

A COMPARISON OF THE EMLEN TRANSECT TECHNIQUE

AND A TIME-AREA COUNT

BY

Dennis Fijalkowski

A THESIS

Submitted to
Michigan State Univeristy
in partial fulfillment of the requirements
for the degree of

MASTER OF SCIENCE
Department of Fisheries and Wildlife
1978

l o
V

ABSTRACT

A COMPARISON OF THE EMLEN TRANSECT TECHNIQUE
AND A TIME-AREA COUNT
by

Dennis Fijalkowski

The ability to determine bird populations in a relatively
short period of time is a worthwhile goal. In this research
the Emlen transect technique (1971) was compared to a time-
area count (Robbins, 1970) to determine the effectiveness of
both methods in accurately representing bird communities.

The methods were tested on two sites in mid-Michigan during
the preparation of reconnaissance environmental analyses con-
ducted by a team of investigators from MSU.

No significant difference was found between the methods
in contacting numbers of species and numbers of individuals.
The time-area count was, however, found to be significantly
more efficient (a = .10) at contacting species/minute and
individuals/minute. '

Low similarity was found between census efforts, and
even between actual replicates, which illustrated how dif-
ficult it was to choose areas representative of the major
habitats from aerial photos and distant observations.

Visibility, the observers ability to first contact

birds by sight, was almost three times higher using the

Emlen technique than the time-area method. Almost half of
the Emlen contacts were visible ones. Visibility data for
the two observers was almost equal in all habitat types.

Sources of error were analyzed and discussed. The
estimation of contact distances was thought to be the largest
source of error using both methods. Distances to audible
contacts were easier to estimate using the time-area count.

Other sources of error were caused by lack of homo-
geneity within areas chosen to be censused, flocking of
birds, observer differences, and double recording.

Observers found the time-area count easier to use under
reconnaissance conditions. The Emlen technique was esti-
mated to be twice as labor intensive as the time-area count.

Recommendations were made for future bird censusing
during environmental analyses. Problems in the reporting
of data were discussed and recommendations made for future

reporting of bird census information.

"Only men of leisure have time to wonder, or wander, which-

ever the case."

A man of leisure - 1976

ii

ACKNOWLEDGMENTS

I would like to thank Dr. Leslie W. Gysel for his
assistance and patience in what must have, at times, been
thought a hopeless cause. He, along with Dr. Donald Beaver,
Dr. Harold Prince and Dr. Stan Zarnoch, served as invaluable
counsel during the preparation of this manuscript.

I also wish to thank the Michigan Agricultural Experi-
ment Station which funded my graduate studies at Michigan
State University.

Thanks to the three guys who helped me collect my data:
Wayne, Dietrich and Jim (wherever you are). I especially
would like to thank Wayne and his partner in crime Pat, who
collectively heckled, cajoled, and hounded me until I
finally finished what should have been done long ago.

Lastly, to you mom and Mary, thank you for your encour-
agement, harrassment, and love. Accomplishments are empty

without someone to share them with.

iii

INTRODUCTION . . . .
STUDY AREAS. . .
Fowler Site.

Type 1:

Type 2:

Type 3:

Type 4:

Type 5:

TABLE OF CONTENTS

Wet Area Vegetation . . . .
Old Field Communities . . .
Shrub-Sampling Communities.
Immature Woods. . . . . . .

Old-Growth Woods. . . . . .

TrUfant Site. 0 O O O I I O O O O O O 0

Type 1:

Type 2:

METHODS. . . . . . .

Early Succession. . . . . .

Dry WOOds O O O O O O O O O

COLLECTION OF CENSUS DATA. . . . .

ANALYSIS OF CENSUS DATA. . . . . .

RESULTS AND DISCUSSION . . . . . . . .
EFFICIENCY OF THE METHODS. . . . .
Time Efficiency. . . . . . . .
Completeness . . . . . . . . .
ISIMILARITY MEASUREMENTS. . . . . .
VISIBILITY . . . . . . . . . . . .
EASE OF SAMPLING . . . . . . . . .

ERRORS IN SAMPLING . . . . . . . .

iv

12

13

17
17
20
20
23
25
29

30

SUMMARY AND CONCLUSIONS

APPENDIX A. . . .

APPENDIX B. . . .

LITERATURE CITED.

40

43

Table

Table

Table

Table

LIST OF TABLES

Mean Number of Species and Individuals
per Census Effort by Habitat Type for
EaCh MethOd O O O O O O O C O O O O 0

Most Frequently Contacted Species,
Their Percent Visibility by Method, and
Total Percent Visibility. . . . . . . .

Total Contacts and Percent Visible
Contacts by Habitat . . . . . . . . . .

Percent Visible Contacts by Observers
Using Emlen Method. . . . . . . . . . .

vi

18

19

27

28

Figure
Figure

Figure

Figure

LIST OF FIGURES

Fowler Study Area . . . .

Trufant Study Area. . . .

Comparison of Species/Minute Data

for the Two Methods . . .

Comparison of Individuals/Minute

Data for the Two Methods.

vii

21

22

INTRODUCTION

Man has long tried to determine the value of a particu-
lar area for a certain species of wildlife. These deter-
minations were for the most part made by scientists for
their own benefit, and seldom was there any practical use
for this information. The exceptions, of course, were for
managing a species of economic or sporting value, or for
preserving an endangered species.

All that changed with the advent of the National Envir-
onmental Policy Act (NEPA) of 1970. Section 102(2)(C) of
NEPA requires federal agencies to prepare environmental
impact statements on "proposals for legislation and other
major federal actions significantly affecting the quality
of the human environment." One requirement of these envir-
onmental impact statements is an evaluation of the importance
of affected areas to man and wildlife. To meet this provision,
agencies have sought ways to evaluate wildlife resources.
Primary to any evaluation is some estimate of the actual
population of wild animals.

Estimating wildlife numbers is a difficult task even
with intensive study. Ideally, a method for estimating
bird populations should be simple and efficient and should
be adaptable to reconnaissance situations as well as inten-
sive analyses. It should yield species composition and
densities. Lastly, since a census method may have to be

1

2
used in different habitat types, and should be adaptable to
such situations.

In the past, simple transect counts were used to census
birds (Robbins, n,d.; Kendeigh, 1956; Enemar and Sjostrand,
1967). These yielded fairly accurate species composition
but such terms as rare, common, and abundant lack definition
when used to indicate species abundance. Even professional
investigators define these terms differently, and they can
be highly misleading.

In this research project, two methods of censusing
birds are investigated for reconnaissance type surveys that
would appear to yield density values and meet the require-
ments previously stated. Both methods were tested between
June 13 and July 26, 1973 on two separate study areas being
considered by the Consumers Power Company of Michigan as
possible sites for power plant development. The two methods
tested and compared were the Emlen transect technique
(Emlen, 1971) and a time-area count (Robbins, 1970). Since
the Emlen technique, a moving strip census, has been used
successfully to determine density and composition of bird
populations, it was given primary consideration. The time-
area count was used for comparison. In this study contrasts
are made between the two methods and an observer's ability
to contact birds using them. No attempt is made to assess

the accuracy of either method in determining density values.

STUDY AREAS

FOWLER SITE

The first study area, referred to as the Fowler Site
(Figure 1), consists of 11,267 hectares (27,840 acres)
located in Clinton County, Michigan. It is characterized
by level to gently rolling land traversed by several streams
which have been largely altered by dredging. Approximately
85 percent of the land is devoted to agriculture. An aerial
view shows a grid pattern imposed on the landscape as a
result of the early division of the land into square mile
sections. Most sections contain one to several woodlots,
located either in the center or on the more poorly drained
soils (Fijalkowski et 31., 1973). Farming practices are
intensive with little fence row or roadside vegetation.

The natural vegetation of the site was divided into
five types for the purpose of an environmental analysis.
Due to the variety of management practices prevailing over
such a large area, the vegetative communities of any one
type varied considerably. It was also common to find
ecotones and small areas of one type within another type.
These areas were not censused separately and probably
accounted for some of the variability of data within each
type. A description of the five natural vegetative types

is presented below.

  

 

  

l

 

 

_
_
-IIIL lllllllllllllllllllll u
u _
_
n . _
n
_ .n _
_ .w u u
_ . .
. o m _
_ c 0 mil 0 n
r
u M ,«lw “ T
_ .. . .. , .
N . O - _
_ I. - fl ,
- L I - _
_ C . M. rm _
I
— '3 m I —
_ e \ s «I!!— _
— fl 0" —
. w _
_ .m _
L. L

 

Figure 1. Fowler Study Area

Type 1: Wet Area Vegetation

 

Open water marshes and water-course vegetation were
combined under the category of wet area vegetation for the
purposes of censusing birds during the environmental analysis.
Few open water marshes existed on the area, the largest being
less than five acres. Water in temporary marshes persisted
into early summer in years of normal rainfall. Water-course
vegetation is composed of floodplains or lowland communities
within the four remaining types, but usually with denser lower
strata of willows (Salix EBB)! dogwoods (Cornus gpp), or other

brushy vegetation.

Type 2: 01d Field Communities

 

An old field is described as an area which was previously
cleared but has lain fallow long enough for the first stages
of woody vegetation to appear, with up to 40 percent woody
crown cover. This overstory is typically comprised of

hawthorns (Crataequs s22), green ash (Fraxinus pennsylvanica),

 

 

and slippery elm (Ulmus rubra) to heights of 25 feet.

 

Type 3: Shrub-Sapling Communities

 

Shrub-sapling communities are those which are beyond
the old field stages, but not mature enough to be classified
as woods. Type 3 is most commonly composed of dense stands

of hawthorns, prickly ash (Zantholeum americanum), elms,

 

and tree species associated with nearby woodlots, with an

average dbh of 2—4 inches and heights to about 30 feet.

Type 4: Immature Woods

 

Immature woods are those with trees predominately 6-10
inches dbh, and heights generally less than 60 feet. Often
larger trees were found in various densities (probably the
result of selective forestry or high-grading), but not until
this larger stratum reached 50 percent density was it con-
sidered Type 5 (old-growth woods). Due to the large size
of the study area and existing management practices, a
variety of composition, density, and structure is found with-

in this type.

Type 5: Old-Growth Woods

In general, old-growth woods are like Type 4, but are
more mature second growth. Those woodlots with an upper
stratum of 60 feet or more and density of greater than 50
percent, and with an average dbh greater than 10 inches are

classified old-growth woods.

TRUFANT SITE

The second study area consisted of 8,223 hectares
(20,320 acres) located in Kent and Montcalm Counties
(Figure 2). The area is generally level to gently rolling
in nature with low-lying pockets of organic soils. Approxi-
mately 60 percent of the study area is devoted to agricul-
ture of some type. The remaining land is in natural vegeta-
tion, usually on the wetter sites due to the difficulty of
removing timber and preparing the soil for agriculture.

These wetter or organic soils are generally correlated with

 

 

I u...

 

+2

 

o Trutent

 

O-I-o-l-o-J'l-l-' ’ -’ -, -’-’-’-’-I-I-’-l-I-

.EQL‘PL'JESLQ'..-
Kent Co.

 

Figure 2. Trufant Study Area

8
the drainage pattern for the area. The natural vegetation
of the site was divided into four types. The two types

described below were used in all analyses.

Type 1: Early Succession

 

Early succession vegetation is composed of old fields,
shrub communities, shrub-sapling communities, and immature
woods or some combination of these. By definition a wide
variety of seral conditions were lumped as one type. This
was necessary to inventory the bird fauna in a limited time

span.

Type 2: Dry Woods

 

Dry woods were those immature and mature stands on the

more mesic sites.

METHODS

A reconnaissance of each study area was done by auto
and airplane and the natural vegetation divided into seral
stages or types suitable for sampling. It was often diffi-
cult to classify an area by type because its vegetation was
not uniform or because it was borderline between types,
however, a decision was always made as to which type it
closest fit.

Considerations for choosing a plot were threefold:

(l) the area must be large enough for a suitable transect
to run through it (generally at least 1/4 mile long and

1/8 mile wide); (2) it must be relatively uniform in compo-
sition, density, and structure; and (3) sampling of the
area must be consistent with other objectives of the envir-
onmental analysis, which were to collect data on all envir-
onmental parameters of the site.

With each census, the observer(s) was given data
recording sheets (Appendix A) and a panchromatic aerial
photograph (1:14,400) marked with designated transects to
be walked. Distances were marked on the photograph to help
the observer start a transect or find his way to other
transects.

Portable cassette recorders were carried by observers

during all field sampling. When unknown species were

10
encountered, the songs or calls were taped and later checked
against various bird song recordings.

Even though density values were not used in this study,
contact distances were recorded. Rangefinders were used by
observers for a short period of time to develOp proficiency
at estimating distances.

Birds were censused on the study areas to provide
estimates of densities and relative abundance. The two
methods used for all nonflocking land birds were an Emlen
(1971) technique, a moving transect count, and a time-area
count (Robbins, 1970). Using the Emlen technique the
number of individuals seen or heard for each species and
contact distances were used to determine distribution
patterns in relation to observers on the transect. It was
assumed that birds were distributed randomly at the begin-
ning of each census, and that they tended to move away
from approaching observers.

Emlen felt that a minimum of five transect miles were
needed to obtain reliable data for density estimates. On
the reconnaissance surveys done in conjunction with this
study as little as two miles of transect were used for
estimating densities.

For use in the time-area count method the photos also
had stations marked on the transects for observers to listen
from. In woodland plots the stations were 400 feet apart
with the first one located 200 feet from the beginning of

the transect. In field communities the stations were

ll
spaced 800 feet apart. Transects were generally at least
one quarter mile in length and the census plot usually con-
tained at least two transects, spaced far enough apart so
that the observer would not easily contact the same bird
from both lines. The first transect was located 200-300
feet inside the plot and running parallel to the edge so
that species frequenting the edge would be contacted, but
not overrepresented.

Two principle investigators and two recorders did the
censusing, working in teams. A census effort was defined
as one team of observers (a principle investigator and a
recorder) censusing one morning, on one area. The princi-
ple investigators were experienced birders and considered
equal in proficiency during the study.

When encountering more than two individuals of the
same species they were recorded as one contact. This was
found to be necessary to avoid overestimating densities of
flocking species, since flocked birds are not randomly
distributed. After young birds fledged flocking was a
serious problem in some species.

Using the time-area count, the observer stopped at
intervals (stations) along the transect for three minutes,
recording all contacts with birds either by sight or sound
(Robbins and VanVelsen, undated). Density values from
time—area data could be calculated as in the modified Emlen
method.

The basic difference between the methods was birds

were recorded continuously along the line of travel for

12
the Emlen technique, whereas for the time-area count birds
were only recorded at specific intervals along the transect.
When an area was censused using one method, it was

usually censused the next day using the other method. It
was hoped that any error caused by seasonal variations could
be eliminated and results compared if census methods were
tested on consecutive days. In most cases the same investi-
gator used both census methods on an area so valid compari-

sons could be made.

COLLECTION OF CENSUS DATA

Censuses were conducted in the early morning, starting
about 1/2 hour after local sunrise (Emlen, 1971) and
generally not continuing more than two hours. One or two
observers walked transects which were picked to cover areas
representative of the major nonagricultural vegetative
types, and laid out so that observers contacted birds on
the edges as well as the interiors of areas censused. When
two observers worked together, one would collect data and
give it to the other verbally, who would record the data on
field sheets.

Data collected using either method was the same:
species, sex, mode of contact (sight or sound), and estimated
distance from the observer. Contact distances were estimated
in lO-foot intervals to 100 feet, then single intervals of
100-200 feet, 200-400 feet, and 400 feet plus.

In accordance with the Emlen method, the observer(s)

walked the specified route at a slow pace pausing briefly

13
to look and listen. Long stOps were avoided to reduce the
danger of double recording, and birds ahead of the observer
were not recorded until they were within 100 feet. Speeds
averaged about one mile per hour. The lateral distance
from the transect to the initial point of detection of the
bird was estimated. Squeaking and pishing sounds were made
by observers to lure birds from hiding for identification
purposes.

Time-area count census efforts were conducted in the
same manner except that birds were recorded at specific
intervals (stations) along the transect rather than contin-
uously. For the time-area method, distance was estimated
from the bird to the observer.

When three or more birds of a single species were con-
tacted together, they were plotted as a single contact in
order to reduce the possibility of overestimating species
density. Immature birds were recorded in the same way as
adults, but nestlings were not recorded.

Time was kept by the observers for computation of
efficiency values. Time began when the first transect was
started and ending when the last transect was finished.

No attempt was made to standardize time on census efforts

or make it equal using the two methods.

ANALYSIS OF CENSUS DATA
Since this research compared two census methods, the
beginning hypothesis to be tested was Ho: Modified Emlen =

time-area count. The alternate hypothesis was H1:

14
Modified Emlen # time-area count. If Ho was rejected, the
question was which method was the better. Several methods
were used to test Ho' The a level was set at .10 for all
statistical tests due to the small sample size and the
imprecise nature of the census methodologies. The higher
a (.10) resulted in lower Type II error.

Efficiency is an important consideration in choosing
any census method, especially on a reconnaissance survey.
Efficiency in this study was measured in terms of time,
and accuracy or completeness. The primary concern of how
much time was required to census an area was evaluated
using the parameters of species and individuals contacted
per minute.

Species per minute was defined as the total number of
taxa contacted during a given census effort divided by the
number of minutes spent censusing. Individuals per minute
was the total number of individual contacts during a census
effort divided by the number of minutes spent censusing.

Completeness, the most direct measure of accuracy
this research yielded, was simply a comparison of total
species and total individuals values using each method on
the same area. The method contacting the greatest amount
of species or individuals was more complete.

Time efficiency and completeness data were analyzed
using the paired Wilcoxon Signed Rank test. When there
were more than one census of the same area using the same

method, mean values were used.

 

 

 

 

-'|r a. my .-Ini'.'l si.‘

 

 

 

15

Similarity, a measure of variability, was used to gauge
the effectiveness of both methods in representing the avian
communities of the area censused. Although Murdoch (1973)
developed the index to measure the similarity between two
vegetative communities, in this study the index was used to
measure the similarity of two avian communities. The simi—
larity index (I) is defined as:

s

I = l-0.5(§|ai-bi|),
where ai is the proportion of the total individuals in
sample A that belongs to species i and bi is the proportion
in sample B belonging to species i, and there is S species.
Complete similarity will yield I = l, and complete dis-
similarity gives I = 0. Since the I value does not portray
any bias of the two methods, it should not be considered a
measurement of accuracy. Methods tested can be considered
accurate only in the sense that replication of the same
area should yield high similarity if the census method is
efficient at contacting a representative sample of the avian
community. If two or more censused of the same area are
very similar, then we can assume the method used was fairly
repeatable.

Visibility, or the chance of an observer contacting
individual birds by sight, was compared using the two
methods, by species and overall.

Observers were compared by the use of efficiency and

visibility measurements from the two methods.

16

Three computer programs were used in the analysis of
the data. They were written in FORTRAN IV for the CDC 6500
system at Michigan State University. The first program
(SINDEX) was designed to compare the census efforts with
each other and determine the similarity index.

A second computer program (TOTALS) summed the number
of visible and audible contacts, and computed the percent
of visible contacts for each of the species contacted for
each census effort and overall.

The third program (PRINCE) totaled the contacts by
group (a combination of method and habitat type) yielding
total visible, total audible, percent visible, and percent
audible for each species.

Tests were run of selected samples to check the accuracy

of all computer analyses.

RESULTS AND DISCUSSION

Forty census efforts were conducted in seven habitat
types on the two study areas (Table 1) during this research.
A total of 115 species (Appendix B) were contacted 3,175
times using the two methods. Of the 115 species contacted,
84 species were nonflocking land birds used in this study
to test the methods.

Of the 3,175 contacts, 1,317 or 41.48 percent were
visible contacts. The mean number of contacts per census
effort was 79.38 of which 32.93 were visible contacts.

The most frequently contacted species during the study
was the song sparrow which was contacted 425 times (Table 2).
Four other species were contacted greater than 200 times:
the redwing blackbird, grackle, robin, and cowbird, with
300, 275, 269, and 219 contacts respectively. Twenty-two
species were contacted greater than 50 times during the

study (Table 2).

EFFICIENCY OF THE METHODS

Since this study was intended to evaluate two census
methods under reconnaissance conditions, efficiency of the
methods was analyzed. In this study efficiency was seen as
having two components: time and accuracy.

Time efficiency was analyzed in terms of species per
minute contacted and individuals per minute contacted. Com-
pleteness was evaluated as an indirect measure of accuracy.

17

18

Table 1. Mean Number of Species and Individuals per Census
Effort by Habitat Type for Each Method

 

Habitat Type Emlen Time-Area

 

 

Census Indivi- Census Indivi-
Efforts Speciesa dualsa Efforts Species duals

 

Fowler
1 4 20 76 2 18 50
2 4 19 82 2 17 70
3 2 27 96 2 20 56
4 4 23 74 3 24 68
5 2 20 55 2 24 79
Trufant
1 7 26 111 l 28 156
2 3 20 44 2 28 83

 

a
Mean

l9

 

 

 

mm em so mm umxomaeooz ememmm emm mu
8 mm mm mm emanumz zoaam» am
0 mm NH Hm Menopausam emumouo on
o Hm ea Hm mmnzoe ma
o um ma 3 30.3 ma
m e e we mwzmm RH
Ha ma «H mm gene mmsom ma
m ma NH as manucsm ma
4 we mm Ne mun msam «a
m mm ma eh conundm camem ma
om em mm «m m>oo maecusoz «H
e Hm mm mm umxoeam Ha
m ma R HHH omue> swam emm OH
ma an em mad euflnumu m
5 ma ea mma Hmaeeumo m
ea me He mma nonemeaoo 5
me mm we oeH mcwaumum 8
mm mm as mam omensoo m
em om we mom canon e
mm an ax mum maxomno m
em mm me com onesemm m
s mm mm mmv zonumdm meow H
mafiaflneme> Speaflbflme> mafiaebflmfl> muomucoo mweomdm scam
«re cmflsm Hmuoa Hmuoe
Mafiawnflmfi> unmouwm Hmuoa
cam .cocuwz ma >uwaabama> unmoumm Hausa .mmflommm cmuomucoo aaucmnkum umoz .m manna

20

Time Efficiency

 

The Wilcoxon Signed Rank Test was used to compare the
time efficiency of both methods. Twelve sample pairs of
species per minute data from seven habitat types were used
in the test. When more than one census existed on the
same area using the same method, the mean was used. The
results at the a = .10 level were found to be significantly
different (a* = .084)1. The time-area count method showed
a markedly higher species per minute rate than the Emlen
method (Figure 3).

Twelve sample pairs of data were used in a Wilcoxon
Signed Rank Test to compare individuals per minute data
from seven habitat types using both methods. When more
than one census was done on the same area using the same
method, the mean value was used in the test. The time-area
count method did show a significantly higher individual per

minute count at a = .10 (d* = .092) (Figure 4).

Completeness

 

Number of species and individuals contacted using both
methods were compared. When there was more than one census
of the same area using the same method, mean values were
used. Thirteen pairs of values representing seven differ-
ent habitat types were listed and compared using the

Wilcoxon Signed Rank Test. The census methods were not

 

1(01*) is the smallest level at which the hypothesis
(Ho) would be rejected.

21

 

 

 

 

 

 

 

 

 

 

         

 

 

.—.50
Z _ _
2 —.4o
:3 _
.“J
o —.30
“J
% —
E —.20
2 — 10
1 2 3 4 1 2 '—
FOWLER TYPES TRUFANT TYPES
LEGEND:
fill/£521.79” Iflfifilfiéfim
2—Old field 2—Dry {mods T'ME'AREA

3—Shrub-sapllng
4—lmmature woods EMLEN
s—Mature woods

FIGURE 3. Comparison of species/minute
data for the two methods.

22

— 2.6
—2.4
—2.2 7

 

 

MEAN INDIVIDUALS/MIN. .
1.
.5

 

 

 

 

 

 

 

 

 

       

 

 

— .0
— .4
— .2
2 3 4 1 2
FOWLER TYPES TRUFANT TYPES
LEGEND:
FOWLER TYPES TRUFANT TYPES
1—Wet area 1—Early succession
2—Old field 2—Dry woods
3-Shrub-sapling
4—Immature woods
5—Mature woods
TIME-AREA

mi]

FIGURE 4. Comparison of individuals/minute
data for the two methods.

 

23
found to be significantly different at a = .10 (a* = .396)
when species contacted data for the two methods were com-
pared.
A Wilcoxon Signed Rank Test was used to compare total
individuals contacted data for the same thirteen sample
pairs. The methods were not found to be significantly dif-

ferent in total contacts at a = .10 (a* = .420).

SIMILARITY MEASUREMENTS

A measurement of variability used in this study was a
similarity index (I) (Murdoch gt 21., 1973). The similarity
index was used to evaluate the efficiency of an investigator
in choosing representative communities in an area to be
studied. This point can be especially critical on recon-
naissance surveys since time is limited. The observers'
ability to quickly reconnoiter a study site and choose
areas representative of the dominant vegetative communities
for analysis can be the most important task of the investi-
gator. For this reason, a similarity index program (SINDEX)
compared the 40 census efforts with each other to determine
I values. It must be remembered that this index is a rela-
tive value. An I value of 1.0 would be complete similarity,
an almost impossible occurance in natural communities. An
I value of 0.0 would mean complete dissimilarity between
two communities.

Mean similarity indices of all censuses on areas of

the same type were compared using both methods. The mean I

24
(computed by averaging all similarity indices between
censuses on the same type using the same method) for the
Emlen census was .476 while the mean I for the time-area
count was .550. A Wilcoxon Ranked Sum Test was used to
compare these mean I values. The time-area count efforts
were significantly more similar at a = .10 than Emlen
efforts which meant simply a time-area effort was more
repeatable.

The mean I values appeared to be low, therefore, I
values for actual replicates of the same area, by the same
observer, using the same method were investigated. The
mean value for actual replicates using the Emlen method
was .659, and .679 for the time-area count. This compari-
son of replicates in part explains why more similarity was
not seen in avian communities on areas picked as represen-
tative of the same habitat type.

To compare the two census methods used on the same
area two mean values were calculated. A mean I of .566
irrespective of observer was about equal to the value of
.560 for both methods on the same area by one observer.
The mean I value for comparisons between observers on the
same area using the Emlen method was computed to be .539.
These data indicate there was as much variability between
methods as there was between observers, and no conclusions
can be drawn. Similarity measurements illustrated the
difficulty in choosing similar communities to sample from

aerial photos and distant observations.

25

VISIBILITY

One factor influencing the effectiveness of any census
is visibility. Although a very complex phenomenon, in this
study visibility was considered as having three components
affecting the observer's ability to visibly contact indivi-
dual birds: (1) habitat characteristics; (2) conspicuous-
ness of the species (size, coloration, and behavioral
characteristics); and (3) observation conditions such as
weather, time of day, etc. The last factor was an indepen-
dent variable and controlled in the study.

Visibility in this study was defined as the observers
ability to first contact birds visibly. The bird may also
have been heard, but was sighted first. Visibility data
in this study may not accurately reflect a species true
visibility (an observer's ability to contact a bird by
sight). The factor most influencing whether a contact
would be audible or visual was the individual species
behavioral characteristics. Species that actively forage,
react quickly to an observers presence, or are reluctant
to vocalize tend to be contacted visibly rather than
audibly. Therefore, any measure of visibility here must
not be taken out of context or used as a comparison for
visibility from other studies. It is merely intended as a
mode of comparison between the two census methods, observers,
habitat types and individual species. In the analysis
visibility is expressed in terms of the percentage of

visible contacts of the total.

26

Program TOTALS computed contacts and visibility by
census effort (Table 3). Mean percent visible contacts
averaged almost three times higher using the Emlen method
compared to the time-area count. The greatest difference
occurred in Type 5 from the Fowler area where an observer
using the Emlen method was almost five times more likely to
contact birds visibly than the time-area method. The Dry
Woods on the Trufant area showed that the Emlen method was
less than twice as efficient at contacting birds visibly as
the time-area.

With almost half of the contacts being visible using
the Emlen method, and considering how many of the audible
contacts are also observed, it may be possible to compute
visibility ratios for each species so that an observer not
proficient in identifying calls could conduct a census, and
density figures could be computed using visibility ratios.
An extreme case would be to have a deaf observer do the
censusing and carry a tape recorder so that a competent
person could review the recording for completeness of the
data afterwards. Relying solely on visible contacts, how-
ever, may tend to omit or underrepresent seldom-seen species
such as the ovenbird or pewee.

A comparison of percent visible contacts using repli-
cates of the Emlen method by different observers was done
(Table 4). The percent visibility was nearly equal in all

comparisons between the two observers.

27

 

 

 

«m.mH mm.wv z¢m£
fl .mImJImIH. 3.3 8.3 3388. N
o>.hm mm.aw oo.mma oo.HHH udmmsue H
mm.o vm.am oo.mh om.vm HdaBOh m
mm.m o>.mm no.hm m>.mh Hoa3om v
NH.om o¢.oo oo.mm oo.om Hoa3om m
~m.mm vm.mm om.mw oo.~m Hoasom N
m~.mH em.ho om.m¢ mh.mh Hoa30m H
moudloefle cmHEm mmn<|ofifle doHEm mend monum uouflnmm
wuflawnflmfi> w com: mucoucoo cow:
.umuflnom an mucoucoo manfimfl> unwound can mucoucoo Houoa .m canoe

28

Table 4. Percent Visible Contacts by Observers Using Emlen

 

 

Method.
Observer
Habitat Type Site A B
l Fowler 58.75 54.35
73.17 84.27
65.96a 69.31a
2 Fowler 67.90 70.00
72.97 68.49
70.44a 69.25a
l Trufant 36.67 52.94
70.00
53.34a 52.94
2 Trufant 22.22

 

 

 

a
Means

29

In order to provide more exact data the percent visi-
bility by census effort and total visibility during the
study were determined for each species. Total percent
visibility of the top 22 species ranged from 4.35 for the
pewee to 80.71 for the grackle (Table 2). The second and
third most visible species were the cowbird and starling
with 78.85 and 74.04 percent visibility, respectively. The
second least visible species was the red-eyed vireo with

only 7.26 percent of the 111 contacts being visible ones.

EASE OF SAMPLING

One point which must be considered when choosing any
sample method is the ease with which the observer, espec-
ially an inexperienced one, can use the method. The time-
area count method was found to be the easiest of the two
methods to use. The observers found it easier to walk to
the next station than to follow an unmarked north/south or
east/west transect line and felt more comfortable censusing
while stationary than while walking.

Since transects were marked on photos to be used with
both methods, and stations were chosen for the time-area
method by merely telling the observer how far to walk
between stations, both methods required an equal amount of
preparatory time before censusing. However, without a
second observer, or with one of questionable proficiency,
it would have been difficult to conduct an Emlen count in
very dense vegetation while maintaining a compass direction,

warding off biting insects, and making sure to record every

30
contact accurately. Therefore, if two observers are
required to insure reasonable accuracy using the Emlen
method, it must be considered about twice as labor inten-
sive as the time-area count.

Estimating distances to audible contacts is difficult,
and in this study observers felt more confident in their
estimates using the time-area method than the Emlen method.
However, since estimating visual distances is easier than
audible distances, the Emlen method had the advantage because
using it observers were three times as likely to contact

birds visibly.

ERRORS IN SAMPLING

When doing any research it is important to test the
hypothesis under ideal conditions. In this study data was
collected during an actual reconnaissance survey where cer-
tain variables were not controllable (i.e., observer differ-
ences, habitat variability). Much of the variability or
error in the data is certainly a result of the less than
ideal conditions under which information was collected.
Probably the most important source of error resulted from
estimating distances to contacts.

Enemar and Sjostrand (1967) determined that estimation
of distances by eye was impossible and, therefore, data
collected on a strip survey should not be used to calculate
density values. Emlen, on the other hand, found no problem

estimating distances, especially those less than 100 feet

31

which are the most critical, after practicing with a range-
finder and setting mental references.

Estimating distances was at times difficult, especially
in dense vegetation. Observers practiced estimating
distances and then checking their estimates with range—
finders and by pacing. Since the methods being tested were
so dependent upon an observer's ability to accurately
record contact distances, this factor could be a consider-
able source of error in some circumstances, especially when
trying to identify the location of an audible contact that
can not be seen. As was already discussed, the error in
estimating distances to audible contacts was thought to be
higher using the Emlen method, but using the Emlen method
an observer was much less likely to contact birds audibly.
No measurement of this source of error was attempted in the
study.

Because more than one investigator was used in this
study, there was a theoretical error introduced by differ-
ences in proficiency between observers. Kendeigh (1944)
found that when censusing birds the degree of accuracy
varied widely between observers and times of year. Enemar
and Sjostrand (1967) stated only that the differences
between census takers was considerable. Using any census
method differences exist between the color and hearing per-
ception of observers, as well as their ability to recognize
song and call notes. Often times a contact has to be

identified by silhouette or be recorded as an unknown.

32

When comparing observers using data from this study, there
did not appear to be a significant difference. The third
observer in this study, who recorded data for both of the
primary observers, thought there might have been a differ-
ence between observers in their willingness, or reluctance,
to record a questionable contact. Naturally, because of
differences in proficiency and confidence of the observers,
differences in willingness to record a contact certainly
did exist. However, it was felt that these differences.
were more a function of the observer than of the census
methods used and were discounted since they could not be
tested.

Important to the question of observer error is the use
of tape recorders for recording questionable contacts.

Many times a recorder was used in the study to record an
unknown, and was later identified after comparison with

known sound recordings. In the case of call notes, the

bird was not always identified by sound and since birds

were not chased far from the transect they were recorded
as unknowns.

Since birds were censused in small tracts of natural
vegetation on areas largely altered by man, a question of
whether these areas were large enough and homogeneous
enough for accurate estimates arises. Williamson (1967)
felt that because of edge effect, areas smaller than 100
acres should not be considered for density figures as they

would tend to inflate estimates due to the large number of

33
birds in the edge. Kendeigh (1944) felt that population
densities for forest-edge and forest-interior birds should
be separated because densities of forest-edge species
during the breeding season equal more than one-third of the
population. This, of course, could not be true in all
cases because of the variability in size, configuration,
and structure of the edge and also the corresponding size
of the interior. Nonetheless, Kirkland (1969) in his
research of southeast Michigan woodlots censused interior
and edge separately.

Emlen did not consider edge effect a problem but did
try to pick census areas of at least 50 acres in size. On
smaller tracts, he crisscrossed routes and repeated tra-
verses to obtain an adequate sample size. In this study,

a concerted effort was made to choose the largest, most
homogeneous sites, and most often they were about 50 acres.
In this study the lack of homogeneity in the areas censused
was probably the second most important source of error.

Adverse weather can affect any census method. Although
wet conditions were encountered many times, only once did
it rain hard enough to affect census results in the investi-
gators Opinion. That census was not used in the analysis
of data for this study. At other times; light precipita-
tion did not appear to affect census data. Many mornings
investigators encountered heavy dew without noticeable
effect. No attempt to correlate barometric pressure with

census data was made.

34

The Emlen transect count, because of the method of
computing density values, is supposed to be unaffected by
seasonal variation of pOpulations and individual species
conspicuousness. However, as the season progresses and
young birds fledge, flocking of family units becomes preva-
lent in some species. For this reason, when three or more
birds of a single species were contacted together they were
counted as one to avoid overestimating the true population.
Either way, flocking probably leads to a source of error.
This is the reason Emlen does not attempt to use his method
on flocking species.

A last source of error was the possibility of contact-
ing the same bird twice. Using the Emlen method it did not
matter if the same bird was contacted from different tran-
sects. However, if a bird was contacted twice on the same
transect, it was a source of error. Emlen did not discuss
this source of error using his technique. Although a
conscious effort was made to avoid this source of doubling,
it must have occurred. The problem of doubling on the
time-area count (contacting the same bird from different
stations) was not considered to be a serious source of
error because observers attempted to mentally note location
of contacts and would not count them again at consecutive
stations. Also, distance between stations was chosen so as

to minimize the error for most species.

SUMMARY AND CONCLUSIONS

The Emlen transect technique and a time-area count
were tested and compared on two reconnaissance surveys to
determine density values of bird populations. The study
was done during the preparation of two environmental impact
statements on sites in the mid-Michigan area. The two
methods were not shown to be significantly different
(a = .10) in terms of number of species and individuals
contacted. The time—area count was, however, significantly
more efficient (a = .10) at contacting species per unit
time and individuals per unit time than the Emlen method.

A similarity index was used as an indirect measurement
of accuracy of the two methods. The time-area count censuses
were significantly more similar at a = .10. Similarity of
bird communities in areas classified as the same vegetative
type was low, which indicated that although vegetation
looked similar in structure and age class there may have
been substantial differences between them. For this reason
on larger study areas grouping vegetation into types for
sampling purposes is difficult and was probably the second
most important source of error.

Probably the most important source of error was the
difficulty in estimating contact distances, especially
audible ones. Lesser sources of error included flocking of
birds, observer differences and double recording birds.

35

36

An observer using the Emlen method had almost three
times the chance of contacting birds visibly than using the
time-area count, which might make it easier for a less pro-
ficient observer to use. However, the Emlen technique
appeared to be more disruptive, which might explain the
significantly lower species per minute and individual per
minute rate at contacting birds. This disruption in the
community might also be responsible for the higher visibil-
ity rates for the Emlen technique.

Overall visibility was nearly equal for two observers
using the Emlen method. The grackle was the most visible
of the twenty-two most contacted species, and the pewee the
least. The song sparrow was the most frequently contacted
species followed by the redwing blackbird, grackle, robin,
and cowbird.

The investigators in this study preferred the time-
area count for ease of sampling. They were more comfortable
censusing at stations than along the entire transect. With-
out a second observer on the census or with one of question-
able proficiency, it was felt the Emlen method would have
lacked accuracy. This might have been different if the
transects were marked or had followed trails.

Observer error on identifying aural contacts was mini-
mized by having investigators carry tape recorders and.
comparing questionable contacts with known recordings.

Although the Emlen transect count is to be used for

censusing nonflocking land birds, many species not

37
considered to flock are contacted in groups. This phenome-
non can be a problem during the latter part of the breeding
season when birds are fledging and family units are contacted.
Robins are an example of this phenomenon.

Approximately one-third the total hours of the two
reconnaissance surveys was spent either censusing, compiling,
or calculating density values of avian populations. Since
two methods were being used, approximately one-fourth the
total hours would be required to use one method. This may
represent too much time for one facet of a reconnaissance
environmental analysis. However, as the intensity of the
study increases, collection of density information would
represent a smaller fraction of the total effort. During
this study the Emlen technique was found to be about twice
as labor intensive as the time-area count.

In conclusion, I am not sure that density values are
important enough in a reconnaissance analysis to warrant
the time required. In the five years since this research
began, the state of the art has not progressed to where
density values are reported even for intensive surveys.
Although it may be desirable to report density values, it
is questionable whether it will ever be the practice in
reconnaissance surveys. More important is the reporting of
less common species, and an estimation of the sites impor-
tance to those species. Since the time-area count was more
time efficient at contacting numbers of species and indivi-

duals, I recommend this method be used.

38

A further recommendation would be to test a combination
of the simple transect walk and the time-area count. An
investigator could record contacts while walking a transect
and stopping at specified locations along the route to
record for a specified period of time. Various station times
could be easily tested for efficiency. This combination may
be more efficient at contacting birds than either of the
methods tested in this research.

Since lack of standardization of reporting results has
always been a problem, I recommend that species per minute
and individuals per minute be reported for each habitat type
followed by the number of transect miles in parentheses.
Although data reported in this manner would not indicate
densities, it would serve as an index. Species per minute
and contacts per minute can be easily calculated from simple
transect or station counts and are a measure of relative
abundance. Using data expressed in this manner along with
a species list, a diversity index (MacArthur and MacArthur,
1961), and a report of less common species (i.e., threatened
and endangered) comparisons can be made between communities
on the same site or between sites and would be adaptable to

any habitat.

APPENDIX A

APPENDIX A:

Date:

Time Start:

Temp:

Species

39
Bird Census Data

Observer(s): Site:

Area:
Time Finish: Photo #:
Wind: Cloud Cover: Precipitation:
CONTACTS
Sex Distanct From Transect angle in

 

 

M F Visual Song-Call 0-100 100-200 zoo-Foo h00+

degrees

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

APPENDIX B

APPENDIX B

SPECIES CONTACTED DURING THE STUDY

Common Name

 

Pied-billed grebe
Great blue heron
Green heron
American bittern
Mallard
Green-winged teal
Blue-winged teal
Wood duck

Turkey vulture
Sharp-shinned hawk
Cooper's hawk
Red-tailed hawk
Red-shouldered hawk

Marsh hawk
Kestrel
Ruffed grouse

Bobwhite

Ring-necked pheasant

Virginia rail

Sora

American coot

Killdeer

Woodcock

Common snipe

Spotted sandpiper

Solitary sandpiper

Rock dove

Mourning dove

Yellow-billed cuckoo

Black-billed cuckoo

Screech owl

Great horned owl

Whip-poor-will

Common nighthawk

Chimney swift

Ruby-throated humming-
bird

Belted kingfisher

Yellow-shafted flicker

 

Scientific Namea

 

Podilymbus podiceps
Ardea herodias
Buterides virescens
Botaurus lentiginosus
Anas platyrhymchos
Anas carolinensis
Anas discors

Aix sponsa
Cathartes aura
Accipiter striatus
Accipiter cooperii
Buteo jamaicensis
Buteo lineatus

Circus cyaneus
Faleo tinnunculus
Bonasa umbellus

Colinus virginianus
Phasianus colchicus
Rallus limicola
Porzana carolina
Fulica americana
Charadrius vociferus
Philahela minor
Capella gallinago
Actitus macularia
Tringa solitaria
Columba livia
Zenaidura macroura
Coccyzus americanus
Coccyzus erythropthalmus
Otus asio

Bubo virginianus
Caprimulgus vociferus
Chordeiles minor
Chaetura pelagica

Archilochus colubris
Megaceryle alcyon
Colaptes auratus

aAfter A.O.U. check-list (1957).

40

Common Name

 

Red-bellied woodpecker

Red-headed woodpecker

Yellow-bellied sapsucker

Hairy woodpecker

Downy woodpecker

Eastern kingbird

Great crested flycatcher

Eastern phoebe

Acadian flycatcher

Willow flycatcher

Least flycatcher

Eastern wood pewee

Horned lark

Tree swallow

Rough-winged swallow

Barn swallow

Cliff swallow

Purple martin

Blue jay

Common crow

Black-capped Chickadee

Tufted titmouse

White—breasted nuthatch

House wren

Winter wren

Catbird

Brown thrasher

Robin

Wood thrust

Swainson's thrust

Veery

Eastern bluebird

Blue-gray gnatcatcher

Golden-crowned kinglet

Cedar waxwing

Starling

Yellow-throated vireo

Red-eyed vireo

Philadelphia vireo

Warbling vireo

Black-and-white warbler

Golden-winged warbler

Tennessee warbler

Yellow warbler

Magnolia warbler

Myrtle warbler

Black-throat green
warbler

Blackbur

Chesnut-sided warbler

Bay-breasted warbler

41

Scientific Name

 

Centurus carolinus
Melanerpes erythrocephalus
Sphyrapicus varius
Dendrocopus villosus
Dendroc0pus pubescens
Tyrannua tyrannus

Myiachus crinitus

Sayornis phoebe

Empidonax virescens

Empidonax minimus
Contopus virens
EremOphila alpestris
Iridoprocne bicolor
Stelgidopteryx ruficollis
Hirundo rustica
Petrochelidon pyrrhonota
Progne subis
Cyanocitta cristata
Corvus brachyrhynchos
Parus atricapillus
Parus bicolor

Sitta cardinensis
Troglodytes aedon
Troglodytes troglodytes
Dumetella carolinensis
Toxostoma rufum

Turdus migratorius
Hylocichla mustelina
Hylocichla ustulata
Hylocichla fuscescens
Sialis sialis
Polioptila cacrulea
Regulus satrapa
Bombycilla cedrorum
Sturnus vulgaris

Vireo flavifrons

Vireo olivaceus

Vireo philadelphicus
Vireo giluus

Mniotilta varia
Vermivora chrysoptera
Vermivora peregrine
Dendroica petechia
Dendroica magnolia
Dendroica coronata

Dendroica virens
Dendroica fusca
Dendroica pensylvanica
Dendroica castanea

Common Name

 

Ovenbird

Yellowthroat

American redstart
House sparrow
Bobolink

Eastern meadowlark
Red-winged blackbird
Orchard oriole
Northern oriole
Common grackle
Brown-headed cowbird
Scarlet tanager
Cardinal
Rose-breasted grosbeak
Indigo bunting
Dickcissel

American goldfinch
Rufous-sided towhee
Savannah sparrow
Henslow's sparrow
Vesper sparrow
Chipping sparrow
Field sparrow
White-crowned sparrow
White-throated sparrow
Swamp sparrow

Song sparrow

42

Scientific Name

 

Seiurus aurocapillus
Geothlypus trichas
Setophaga ruticilla
Passer domesticus
Dolichonyx oryzivorus
Sturnella magna
Agelaius phoeniceus
Icterus spurius

Icterus gallenla
Quiscalus quiscula
Molathrus ater

Piranga olivacea
Richmondena cardinalis
Pheucticus ludovicianus
Passerina cyanea

Spiza americana

Spinus trestes

Pipilo erythrophthalmus
Passerculus sandwichensis
Passerherbulus henslowii
Pooecetes gramineus
Spizella passerina
Spizella pusilla
Zonotrichea leucophrys
Zonotrichea albicollis
Melospiza georgiana
Melospiza melodia

LITERATURE CITED

American Ornithologists' Union. 1957. Check-list of North
American birds, 5th edition. American Ornithologists'
Union, Ithaca. 691 pp.

Emlen, J.T. 1971. Population densities of birds derived
from transect counts. Auk 88(2): 223-242.

Enemar, A. and B. Sjostrand. 1967. The strip survey as a
complement to study area investigations in bird census
work. Var Fagelvarld 26: 111-130.

Fijalkowski, D.P., J.A. Kesel, D.C. Schaaf, and W.A. Schmidt.
1973. An ecological analysis of the Fowler site.
Mimeo report for Consumers Power Company, Jackson, MI.
65 pp.

. 1973. An ecological analysis of the Trufant

 

site. Mimeo report for Consumers Power Company,
Jackson, MI. 75 pp.

Kendeigh, S.C. 1944. Measurement of bird pOpulations.
Ecol. Monogr., 14: 67-106.

. 1956. A trail census of birds at Itasca

 

State Park, Minnesota. Flicker, 28: 90-104.
MacArthur, R.H. and J.H. MacArthur. 1961. On bird species
diversity. Ecol. 42: 594-598.
Murdoch, W.W., F.C. Evans, and C.H. Peterson. 1973.
Diversity and pattern in plants and insects. Ecol.

53. 819-827.

43

44

The National Environmental Policy Act of 1969, Public Law
91-190. Jan. 1, 1970. (42 U.S.C. 4321-4347).

Robbins, C.S. and W.T. van Velzen. 1970. Progress report
on the North American breeding bird survey. Bird
Census Work and Environmental Monitoring/Symposium/

Ammarnas, 27-29 June, 1969: 22-30.

"Illlllillllllllllllls