Pacific Science (1992), vol. 46, no. 1: 46-57
© 1992 by University of Hawaii Press. All rights reserved
Origin and Population Growth of the Brown Tree Snake,
Boiga irregularis, on Guam!
GORDON H. RODDA,2 THOMAS H. FRITTs,3 AND PAUL J. CONRy4
ABSTRACT: After the accidental introduction ofthe Brown Tree Snake, Boiga
irregularis, to the island of Guam after World War II, the snake became
exceedingly numerous, and most of Guam's native vertebrates either became
endangered or disappeared from the island. In this paper we summarize what is
known about populations of this snake on Guam and the likely origin of the
Guam population. Scale counts and transportation records suggest that the
Guam population originated in the Admiralty Islands, about 1500 km south of
Guam. It was probably transported to Guam in ships that transported salvaged
war materiel after World War II. For ca. 35 yr after its introduction, the
presence of the snake on Guam was documented only by popular accounts,
occasional photographs, and a few museum specimens, indicating that the
snake's distribution was fairly limited initially, but ultimately a period of sharp
population growth and wide dispersal occurred, with the snake reaching all parts
of the island by the late 1960s. Peak population levels were attained about a
decade or more after each area was colonized. Mark-recapture and removal data
indicate that the capture of50 snakes per ha at one site in northern Guam during
1985 probably represented a population density of around 100 snakes per ha,
but by 1988 this population had declined to around 30% of the 1985 density.
However, this reduction may not be permanent. In central Guam, where the
snake irrupted decades ago, the snake's numbers have continued to fluctuate,
and in some cases it has attained densities in excess of 50 per hectare.
THE INTRODUCTION OF A SPECIES provides an
opportunity to evaluate ecological relation 
ships within a community by monitoring the
trajectories of the component populations as
a new equilibrium is established. This process
is most observable in communities with few
species and major perturbations, conditions
that facilitate interpretation ofcausation. The
1 Funding for this research was provided by the U.S.
Department of the Interior, Technical Assistance Pro 
gram, the U.S. Navy, and the U.S. Air Force. Manuscript
accepted 5 April 1991.
2 Arizona Cooperative Fish and Wildlife Research
Unit, Biological Sciences East Room 210, University of
Arizona, Tucson, Arizona 85721.
3U.S. Fish and Wildlife Service, National Museum of
Natural History, Washington, D.C. 20560.
4Guam Division of Aquatic and Wildlife Resources,
P.O. Box 2950, Agana, Guam 96910. Present address:
Hawaii Division of Forestry and Wildlife, 1151
Punchbowl Street, Honolulu, Hawaii 96813.
accidental introduction of the Brown Tree
Snake, Boiga irregularis, on Guam is believed
responsible for the nearly complete extirpa 
tion of native bird species (Savidge 1987,
Engbring and Fritts 1988), for hundreds of
power outages (Fritts et al. 1987), for loss of
domestic and pet animals (Fritts and McCoid
in press), for envenomation of human babies
(Fritts et al. 1990), and for the probable
extirpation of native bat and lizard popula 
tions (Wiles 1987, Fritts 1988). Thus, Guam's
vertebrate population ecology is of great in 
terest not only to researchers and island biolo 
gists but also to managers responsible for the
control and containment of the snake.
Only recently has any attempt been made
to document the numbers ofthe snake. Popu 
lation size measurements were not undertaken
before 1985, but some information exists on
the date of initial colonization and the chro 
nology of the spread of the snake across the
46
Brown Tree Snake on Guam-RoDDA, FRITTS, AND CONRY 47
island (Savidge 1987). These data are useful
for predicting the possible time course of
Boiga colonizations on other islands and for
evaluating the snake's ability to disperse from
extralimital populations. Minimum estimates
ofthe snake's density (Savidge 1987, Fritts et
al. 1989) based on the number of snakes
removed from two trap grids in 1985 had not
been corrected for the number of snakes that
eluded capture or those that might have been
inactive during the sampling periods. Those
two studies used different trap technologies at
different sites, and the information is not
closely comparable. Thus we did not use these
trap data in our detailed analyses ofhistorical
trends in snake numbers.
A population assessment program is needed
to: (1) assess the predatory threat resulting
from the snake, (2) identify the factors that
influence the snake's ecological success, (3)
predict the snake's future population cycles,
and, (4) provide baseline data for quantifying
the effectiveness of management efforts to
control the snake. This paper is a review
of available information on the origin and
growth of the snake's populations on Guam
before 1985, and subsequent attempts to esti 
mate population trends by visual censuses and
mark-recapture studies.
MATERIALS AND METHODS
Study Sites
The study sites were located in central and
northern Guam (Figure 1). Censuses at two of
the sites involved counts of snakes on chain 
link fences (Andersen Fence and Airport
Fence); the other transects were in secondary
limestone forest (Stone 1970) with a roadside
border of the tree Leucaena leucocephala.
Details of their locations and plant species
composition are given in the following sec 
tion. Transect sampling durations are given
in Table 1.
Transect Descriptions
The habitat descriptions for the sites stud 
ied in 1988 are based on an average of 337
point samples ofthe vegetation per site. These
point samples of vegetation were collected
during routine snake censuses by noting the
species of plant being viewed at the instant
when an unpredictable timer sounded.
The Airport Road site (13° 29.64' N, 144°
47.58' E), along the entrance road leading to
Guam International Airport (Guam Route
lOA), extended for about 0.7 km southeast
from the airport and along the edge of a
forest consisting ofabout 71%Leucaena leuco 
cephala. No other tree species was abundant.
This isolated patch offorest in an urban area
is at the bottom of the cliff on which the
Airport Fence census line was located. These
two lines were about 100 m apart at their
nearest point.
The Northwest Forest transect made a
triangle, with a perimeter length of about
2.4 km, to the southeast of 13° 38.10' N,
144° 50.76' E ca. 2.4 km southwest ofRitidian
Point on the northern end of Guam. Much
of this transect was along a narrow path
through a second-growth forest, including
three common trees: Leucaena leucocephala
(28%), Scaevola taccata (32%), and Casuarina
equisetifolia (18%).
The Communications Site transect was
near a cliff edge at 13° 34.8' N, 144° 49.8' E
immediately inland from Haputo Beach on
the Naval Communication Area Master
Station-Western Pacific. This was the most
diverse site botanically, with 16 genera of
common forest trees that were abundant and
only 20% Leucaena.
The Weapons Storage transect, in the Con 
ventional Weapons Storage Area ofAndersen
Air Force Base, was 2.9 km northeast ofPott's
Junction at 13° 36.06' N, 144° 53.22' E. The
transect ran along 1.5 km of roadside forest
edge similar to that found at Northwest Road.
The Andersen Fence transect (13° 35.1' N,
144° 56.52' E) was a l-km-long chain-link
fence at the eastern end of the main flightline
of Andersen Air Force Base. The fence ran
along the edge of cleared areas at 4 to 30 m
distance from typical limestone forest.
Airport Fence was a 2.3-m-high chain-link
fence extending along the cliff line above the
Airport Road site (13° 29.64' N, 144° 47.58'
E), near the Guam International Airport.
Compared to Andersen Fence, it was longer
48 PACIFIC SCIENCE, Volume 46, January 1992
N
t
APRA HARBOR
OROTE S:;
AIRPORT ROAD •
•
AIRPORT
FENCE
STORAGE
ANDERSEN
FENCE
GUAM
10KM
FIGURE I. Outline map of Guam and the locations ofstudy areas.
TABLE I
NUMBER OF HOURS OF CENSUSING CONDUCTED AT EACH TRANSECT
TRANSECT 1985 1986 1987 1988 1989 1990
Airport Fence 7.0 10.2 14.4 13.7 5.0
Airport Road 10.6 2.5 1.3 1.9
Andersen Fence 20.4 22.7 25.9 31.5 31.5
Northwest Forest 9.0 49.2
Northwest Road 66.0 55.5 9.1 31.0 16.5 5.9
Orote 4.0 1.5 20.4 17.1 45.5
Weapons Storage 34.0 3.5 14.5
Brown Tree Snake on Guam-RoDDA, FRITTS, AND CONRY 49
(3.3 km) and usually closer to the adjacent
vegetation (2-4 m).
The Orote transect was a relatively short
transect (0.7 km) through an almost pure
(95%) stand ofLeucaena located on the tip of
Orote Peninsula (13
0
26.58' N, 144
0
37.32' E)
adjacent to Apra Harbor, where the snake
first became established on Guam. Unlike
the other transects, which had about 0.5%
vine cover, Orote had extensive viney areas
(10%).
The Northwest Road site was near the
Northwest Forest site 2.5 km southwest of
Ritidian Point. Itwas usually searched for 2.0
km to the southwest from 13
0
38.10' N, 144
0
50.76' E. Like most sites it was dominated by
Leucaena (72%), with no other species of
tree being conspicuous.
Origin ofthe Guam Population
Our information on the colonization of
the Brown Tree Snake on Guam was based
on newspaper reports, photographs, museum
specimens, sight records from herpetologists
visiting Guam, and postwar books and maga 
zine articles describing the fauna ofGuam. To
identify the most probable source ofthe colo 
nizers, we compared standard scale counts of
specimens from Guam with those of speci 
mens from various parts of the snake's native
range. We also reviewed records of human
activities and interviewed naval and civilian
personnel who monitored or handled ship 
ments originating in the snake's native range.
We studied preserved material at the National
Museum of Natural History (USNM), Califor 
nia Academy ofSciences (CAS), University of
Kansas (KU), Los Angeles County Museum
(LACM), Carnegie Museum (CM), and the
American Museum of Natural History
(AMNH).
Relative Abundance
To estimate the relative abundance of this
nocturnal arboreal snake, we conducted
timed visual searches. Population trends were
inferred from changes in the annual averages
ofsnakes seen per hour. Most visual censuses
were conducted while walking slowly (mean
walk rates ranged from 0.8 to 1.1 km/h) along
the edge of a road at night and looking for
snakes on fences or in the vegetation from a
distance of 2-4 m. Most of the forests cen 
sused were "typhoon climax" assemblages
of second growth (Fosberg 1960), rarely
exceeding 10 m in height.
Two transects, Airport Fence and Andersen
Fence, were along chain-link fences, instead
of being along forested roadsides. The fences
were usually observed from a slow-moving ve 
hicle (e.g., Airport Fence mean = 4.8 km/hr).
Snakes are relatively easy to spot against
the regular symmetry of a chain-link fence,
allowing higher censusing speeds. However,
the earlier censuses of Airport Fence were
usually done on foot, and car and foot cen 
suses are not equatable on aper~hourbasis.
Thus, annual comparisons of sightings at
fence sites were based on the mean number of
snakes seen per census instead of snakes seen
per hour.
Different searchers have different profi 
ciencies in seeing snakes in the forest. This
factor was statistically controlled by entering
"searcher" as a separate main effect in the
analysis ofvariance (ANOVA) that was used
to compare years. Time-of-year effects were
not significant in any of 12 preliminary
ANOVAs conducted to identify important
censusing variables; therefore, data from a
wide variety of months were pooled for the
annual comparisons. The data to be analyzed
were log transformed to eliminate a weak
correlation between means and variances, but
results are reported in untransformed values.
All statistical procedures except absolute pop 
ulation estimates (see following section) were
carried out using the SAS (Statistical Analysis
System) computer package (SAS Institute Inc.
1988). The SAS General Linear Model pro 
gram was used to fit least squares means to
each site-year combination, to estimate stan 
dard errors for these, and to test for differ 
ences between adjacent years at each site with
a t test.
Absolute Population Estimates
We made three absolute population esti 
mates for sites that were sampled concurrently
50
for relative abundance: Orote, 1988; North 
west Road, 1985; and Northwest Road, 1989.
In addition, we reanalyzed the removal trap 
ping data of Savidge (1986) to provide an
unaccompanied absolute population estimate
for Communications Site, 1985. The com 
puter program CAPTURE (White et al. 1982)
was used for maximum likelihood estimates
of absolute population size at Communica 
tions Site, 1985; Orote, 1988; and Northwest
Road, 1989, but the Northwest Road, 1985
data, collected for another purpose (Fritts et
al. 1989), were not suitable for analysis with
CAPTURE.
We used two methods to extrapolate
absolute population estimates for Northwest
Road, 1985. In the first, the ratio of number 
captured to total-estimated-population in the
comparable mark-recapture study at Orote,
1988 provided a visibility correction multiplier
to extrapolate from the number of captures
for Northwest Road, 1985 to a total estimated
population for Northwest Road, 1985. In the
second method, we estimated the relative
population decline at Northwest Road, 1985 
1988 from visual censuses and extrapolated
back in time from the absolute population
estimate of Northwest Road, 1989.
To temporarily mark snakes, we painted
small numbers on their heads using paint
pens. For the mark-recapture effort at North 
west Road in 1989, we also clipped the edges
PACIFIC SCIENCE, Volume 46, January 1992
oftwo unique combinations ofventral scales.
On four occasions we captured active snakes
that had been painted earlier in the same
evening; from these captures we concluded
that paint marking did not cause the snakes
to hide. We painted 12 snakes that were
maintained in captivity for the duration ofthe
1989 marking study to evaluate the marking
techniques. None of these lost either their
paint markings or scale clips during the mark 
recapture time interval. Both marks were
visible on all marked animals recaptured.
RESULTS
Origin ofthe Guam Population
As denoted by its name, B. irregularis is an
extremely variable species in scutellation and
coloration. Few specimens exist from selected
parts ofits range, and no complete analysis of
the geographic variation in the species is
available. However, our preliminary compari 
sons suggest that the snakes on Guam are
more similar to snakes from the Admiralty
Islands north of New Guinea than to those
from elsewhere in the range. The number of
dorsal scales at three positions on the body
show significant (t test, P < .05) differences
between snakes on the Admiralty Islands and
those of New Guinea (Table 2). In each
TABLE 2
VARIAnON IN NUMBERS OF DORSAL SCALE Rows COUNTED AROUND THE BODY FOR THREE SAMPLES OF
BROWN TREE SNAKES
SAMPLE SITE
Neck
Midbody
Posterior body
NO. OF
SCALE ROWS NEW GUINEA ADMIRALTY IS. GUAM
21 15 (37) 1 (3)
23 26 (63) 1 (20) 9 (29)
25 3 (60) 21 (58)
26 1 (20)
19 2 (5)
21 37 (90) 1 (3)
23 2 (5) 4 (80) 27 (88)
24 1 (3)
25 1 (20) 2 (6)
15 41 (100)
17 5 (100) 31 (100)
NOTE: Values are number ofspecimens exhibiting the scale counts indicated (percentages ofsample in parentheses). The New Guinea
sample includes small islands immediately offshore, but excludes the Bismarck Archipelago and Admiralty Islands.
Brown Tree Snake on Guam-RoDDA, FRITTS, AND CONRY 51
of three scale counts (Table 2), the Guam
sample is not statistically distinguishable from
the Admiralty Island sample, but does differ
significantly from the mainland New Guinea
sample (P < .05). The number of scale rows
tends to be higher in specimens from Guam
and the Admiralty Islands at all three sites on
the body, but the presence of 17 scale rows
around the posterior body at a longitudinal
position 10 scale rows anterior to the vent is
the most conspicuous single character distin 
guishing the Guam and Admiralty Island
populations from snakes elsewhere. There are
obvious differences in coloration among spec 
imens from Guam, and the Admiralty Islands,
and many localities on New Guinea, where
the species is quite variable in coloration. A
more thorough analysis of geographic varia 
tion by one of us (THF) is under way.
Initial Colonization
The date of initial colonization cannot be
determined unequivocally. No evidence exists
for the presence of a snake (other than the
worm snake Ramphotyphlops braminus) on
Guam before the Japanese invasion in 1941
and the American invasion in 1944. Although
record keeping was impaired during the Japa 
nese occupation, several American soldiers
made extensive collections of reptiles on
Guam following the occupation and the only
snakes encountered were worm snakes (col 
lections at USNM, CAS, KU; Smith 1949, Dryden
and Taylor 1969). The earliest written or
photographic evidence for large snakes poten 
tially identifiable as B. irregularis on Guam is
a cryptic statement by Stevens (1953) that
"there have been persistent rumors during
recent years ofsome [snakes] in the vicinity of
Apra Harbor. Allegedly they were brought
inadvertently to the island from a ship berthed
in Apra Harbor. There are no native snakes,
and if any do exist they are merely strays
brought in from the outside." That this state 
ment refers to the Brown Tree Snake is con 
firmed by a statement published two years
later (Guam, Office of the Governor 1955):
"During the past four years eight snakes have
been captured in the area adjoining Apra
Harbor. They ranged from four to eight feet
in length and were all of one specie [sic].
The specie [sic] has been identified as Boiga
irregularis, a rear-fang snake .,.believed
to have entered the Territory from visiting
ships."
We found no historical evidence indicating
that the initial colonization near Apra Harbor
spread quickly. The Guam newspapers
published numerous photographs of large
B. irregularis from the Santa Rita area near
the harbor during 1958-1962. Although
B. irregularis does not occur in the Philip 
pines, the Guam snake was erroneously
known as the Philippine Rat Snake and was
assumed to have arrived on Guam from the
Philippine Islands. Captions indicate that the
early newspaper photos were published to
disprove the popular notion that Guam had
no snakes. Several articles indicated that
snakes were known only from the Santa Rita/
Naval Magazine area of Guam. A U.S. Navy
publication in 1962 (Anonymous 1962) was
unusual for this era in that it correctly identi 
fied both the snake and its mild venom:
"There are a couple of species of snakes, the
rear-fanged Borja Irregularis [sic] ofthe Santa
Rita area being slightly poisonous." This
misspelling of Boiga may have been due to
a subliminal slip by a herpetologically naive
typesetter; "Borja" is a common surname
on Guam.
Evidence for the presence of snakes in 
creased rapidly after 1960. In 1960 the first
museum specimen was collected (USNM).
Savidge (1987) reported that many people in
the southern and central parts of the island
had seen a large snake by the end ofthe 1960s.
Based on reported sightings in 1968 and 1970,
snakes soon reached the northern end of the
island. Savidge (1987) compiled the evidence
indicating a progressive outbreak ofthe snake
in the northern part ofthe island in the 1970s
and 1980s.
Northern Guam, 1985-1989
Snake population trajectories are available
for four sites in northern Guam for 1985 
1989 (Figure 2). Relative population estimates
for the nonfence sites are least squares means
for each site-year combination from an
52
13
CD
(f)
C
ctl
L 
......
L-
CD
Q.
(f)
CD
~
ctl
C
Cf)
3
2
o
PACIFIC SCIENCE, Volume 46, January 1992
NORTHERN GUAM
Andersen Fence
2 Northwest Forest
CD
......
ctl
L-
en
0
c
~
4
.c
en
-+--+--+
(f)
3
Weapons Storage
>-
L-
2
:::J
0
.c
1J
CD
......
(f)
4
:::J
+ __+__,~ort+Road
=a-
3
~
2
---+
0
85 86 87 88 89 90
Year
FIGURE 2. Annual mean relative population counts (± I SE) for sites in northern Guam. Temporally
adjacent samples are connected by solid lines ifthe annual means differed significantly at P = .05 level. Nonfence sites
are hourly sighting rates adjusted for differences in searcher proficiency.
ANOVA based on the effects of searcher
(F=4.34, df=7,194, P=.0002) and the
site*year interaction (F = 3.57, df = 19,194,
P < .0001). No significant differences have
been detected among searchers censusing fence
lines (F = 0.54, df = 6,83, P = .8). Therefore
the fence transects have been analyzed sepa 
rately, considering only the effect of year
(F = 7.85, df = 4,84, P < .0001).
The Weapons Storage site was sampled
only for the period 1985-1987, and no differ 
ences in counts were detected. The three other
sites registered significant reductions in snake
sightings from 1985 to 1988 (Figure 2).
Three absolute population estimates are
available for the northern end of Guam.
Savidge's 1985 data (Savidge 1986) from
Communications Site reanalyzed through
Brown Tree Snake on Guam-RoDDA, FRITTS, AND CONRY 53
CAPTURE indicated a population density of
about 20 snakes per ha (95% CI, 16.5-24.0),
based on 35 removals from a 1.88-ha area.
Our 1989 mark-recapture data from North 
west Road gave an estimate of 35 snakes per
ha (95% CI, 7-62), based on four recaptures
out of 20 animals marked in a 1.5-ha plot
(CAPTURE detected no violations of the
mark-recapture assumptions in this estima 
tion; for evidence of geographic closure, see
Orote results that follow). In 1985, Fritts et al.
(1989) removed 50 snakes from a 1.0-ha plot
at Northwest Road. Using the 0.42 ratio of
captures to estimated population total from
an extremely similar trapping effort at Orote
in 1988 (similar traps, bait, length of effort),
we tentatively extrapolated this figure to a
total population size ofaround 119 snakes per
ha for Northwest Road, 1985. Using the
3.2-fold decline from 1985 to 1988 in relative
population counts at Northwest Road and the
above estimate of 35 per ha at Northwest
Road in 1989, we can tentatively calculate
another absolute population estimate for
Northwest Road, 1985 of around 112 snakes
per ha.
Central Guam, 1985-1989
Relative population estimates in central
Guam are available for only three sites: Air 
port Road, Orote, and Airport Fence (Figure
3). The site near Apra Harbor (Orote) showed
a significant increase in 1988 over the pooled
samples from 1985 to 1987, a significant
decrease for 1988-1989, and an increase for
1989-1990. The two sites along the cliff near
the commercial airport produced contrary
results. Populations at Airport Fence general 
ly increased throughout the period 1985 
1989, whereas those at Airport Road declined
significantly from 1987 to 1988.
An absolute population estimate calculated
for Orote in October 1988 was based on a
12-day period during which we captured 39
snakes, including six animals that had been
previously marked within the 12-day interval.
These animals were all captured within a 2-ha
plot and recaptured within the same area.
Additional evidence for geographic closure
comes from a total of 16 recaptures that were
obtained by visually censusing the 2-ha plot
and an additional 2 ha surrounding the plot
over a 40-day period. None of the recaptured
snakes moved from outside ofthe 2-ha plot to
inside the plot or vice versa, nor did any move
out of the I-ha subplot in which they were
initially captured. The average detected total
movement distance of these snakes was 38 m.
The Schumacher technique estimated a popu 
lation density of 63 per ha for the 2-ha study
area (95% CI, 45-108); the best available
model under program CAPTURE gave a
population estimate of 58 per ha (95% CI,
18-97). Program CAPTURE tests all the
assumptions used in mark-recapture popula 
tion estimation. For the Orote data set, all
assumptions were violated. However, on a
scale of0 to I, where I is the perfect estimator,
the null model that CAPTURE used for the
above estimate rated 0.96. Therefore, the
violations were minor ones.
DISCUSSION
Establishment ofthe Snake on Guam
The Brown Tree Snake probably arrived on
Guam as a passive stowaway in vehicles or
other material transported to Guam from the
Admiralty Islands in the years immediately
after World War II. From 1946 to 1950
the U.S. military endeavored to salvage many
shiploads of damaged war materiel that
had been abandoned in Melanesia (Cdr. M.
Dodge, pers. comm.). The vast naval base at
Manus in the Admiralty Islands was used for
staging this materiel to Guam (see Mead
1956, Lightbody and Wheeler 1985). The
correspondence of scale counts between the
snakes of Guam and snakes from the Admi 
ralty Islands is therefore consistent with the
most likely means of introduction. The pas 
sage quoted from Stevens (1953), indicating
a stowaway arrival of a snake at Apra
Harbor on Guam several years before 1953,
is also consistent with this chronology. The
attraction of the snake to small, dark places
(Pendleton 1947) leaves little doubt that some
snakes were sequestered in the large number
of objects brought to Guam from the
Admiralties.
54
+-'
()
CD
10
U)
C
CO 8
~
+-'
~
6
CD
Q.. 4
en
2
CD
~
CO 0
C
(j)
PACIFIC SCIENCE, Volume 46, January 1992
CENTRAL GUAM
4
Q)
3
Airport Road
+-'
CO
~
2
+-+
en
c
+-'
.c
0
en
U)
-1
>-
~
:::J
0
.c
"'0
3
CD
+-'
U)
2
Orote
:::J
:0-
+----
«
0
85 86 87 88 89 90
Year
FIGURE 3. Annual mean relative population counts (±I SE) for sites in central Guam. We pooled the 1985
and 1987 samples at Orote to assemble a sufficient sample size for comparison to 1988 samples. All other symbols as
in Figure 2.
The plausibility of accidental transport of
snakes in cargo is also suggested by the
discovery of a snake on Wake Island in
1946. This snake, discovered in a tree, was
incorrectly assumed by Downs (1949) to have
been a Brown Tree Snake and presumed to
have arrived in material brought from the
Admiralty or Solomon islands. The snake was
later identified as a North American snake
(Pituophis) by R. Crombie and one of us
(THF), but the fact that Downs called it a
Brown Tree Snake and explained its presence
as a stowaway in military cargo from the
South Pacific suggests that the Brown Tree
Snake had been accidently translocated on
other occasions known to Downs.
Brown Tree Snake on Guam-RoDDA, FRITTS, AND CONRY 55
The initial colonization of the snake on
Guam received little attention, in part because
of the displacement of the island's human
residents from the port area after the Ameri 
can invasion in 1944. For two decades after
the war, the entire Apra Harbor area was a
Navy base off-limits to civilians at night, when
the snakes are active. In contrast to the
resident civilians, many soldiers had grown up
and traveled extensively in places where the
presence of snakes was not noteworthy. The
snakes may already have been established in
military areas for nearly a decade before
their appearance in the civilian areas around
Santa Rita stimulated newspaper photos in
the late 1950s. The snake did not spread far
from Apra Harbor-Santa Rita until the 1960s
(e.g., Anonymous 1962).
Dispersal and Irruption ofthe Snake
throughout Guam
The rapid spread of the snake after 1960 is
unexplained. It is plausible that some people
might have intentionally spread the snake to
suppress rat populations, which were very
high on Guam before establishment of the
snake (Beardsley 1964, Savidge 1986).
The rapid spread of the snake in the 1960s
probably resulted in the snake occurring
islandwide by 1970. The pattern ofbird extir 
pations suggests, however, that numbers of
snakes in northern Guam did not peak until
a decade or more later (Savidge 1987). The
gravid snakes that have been found on Guam
usually have clutches of only three or four
eggs (M. J. McCoid, pers. comm.). Even if
production of two or more clutches in a
season is possible, this species does not have
an especially high reproductive rate (see also
Zwinnenberg 1978). In northern Guam, the
first population increase of snakes does not
seem to have peaked until the mid 1980s.
Peak densities of Brown Tree Snakes in
northern Guam probably approached 100
per ha around 1985. The 50 removed from a
single hectare by Fritts et al. (1989) probably
included a few individuals that had strayed
into the study plot during the course of the
removal. However, there is no evidence for the
kind of territorial pressure that promotes
rapid immigration to a depleted area. Specifi 
cally, home ranges of Boiga overlap enor 
mously (recapture locations from this study
and radiotelemetry work by Manuel Santana,
pers. comm.). Boiga use communal refugia
(Pendleton 1947), and snakes apparently
often encounter each other without animosity
(unpubl. data). The extrapolations we have
made from the 50 removed in 1985 (Fritts
et al. 1989) are rough estimates. However, it
is noteworthy that the two methods gave
comparable results (112 versus 119). We can 
not put confidence intervals on the value
used for the ratio of captured snakes to
estimated total population that was used for
the extrapolation, but 0.42 seems reasonable
in comparison to the 0.38 ratio that applies to
the 1989 Northwest Road trapping program.'
Less capture effort was applied in 1989 at
Northwest Road, thus it was expected that the
1989 ratio would be slightly lower.
At the time that measurements were begun
in 1985, the population density of snakes in
northern Guam was high, approaching 100
per ha at Northwest Road. However, the
population may already have been declining
when the first measurements were taken.
Therefore, the estimate of maximum density
must be considered a lower bound. Although
the peak density of 100 snakes per ha seems
reasonable in light of the relative population
estimates made since that time, it is much
higher than has been recorded for other large
colubrids that are not clustered around water
or hibernacula (Parker and Plummer 1987).
Such a high density is expected to result in the
local depletion of food sources, which is
consistent with the observed concurrent extir 
pations of birds, bats, and lizards (Savidge
1987, Wiles 1987, Fritts 1988).
The differences in snake abundances among
sites did not seem to reflect habitat differences
as much as prey availability as influenced
by the length of occupancy by snakes. The
density of 20 snakes per ha at Communica 
tions Site in 1985 probably reflects a longer
occupancy by the snake and therefore greater
prey depletion at that site than at Northwest
Road. Both sites are in northern Guam, but
Communications Site is 6.5 km closer to the
site of initial colonization and must have lost
56
its bird population earlier than the more
northern site. Densities of snakes at Commu 
nications Site probably peaked before 1985.
The trajectories of snake populations in
central Guam are more complicated than
those in the north. The initial wave of snakes
probably passed through the central Guam
study areas one or two decades before 1985.
Therefore, the 1985-1989 measurements in
central Guam may be indicative ofconditions
after the initial irruption had passed. During
our study, native birds and bats were absent
from the central Guam study areas, as were
most native lizards. However, introduced ver 
tebrates were present at those sites. In par 
ticular, the introduced gecko Hemidactylus
frenatus reached very high densities, especially
.on man-made objects such as road signs,
utility poles, and buildings where lizards were
inaccessible to the snake. In these areas the
snake has demonstrated an ability to reattain
high densities, at least temporarily.
In 1988 the Orote study area had the highest
density of forest lizards (mostly the intro 
duced species H. frenatus and Carlia fusca)
known to us on Guam. The snake's peak
density of over 50 per ha in 1988 may have
resulted in the depletion of prey resources,
for the numbers of both predator and prey
declined in 1989.
The opposing trajectories exhibited by
snakes along the Airport Road/Airport Fence
cliff transects may reflect long-term shifts of
snakes between the forest (Airport Road) and
the fence (Airport Fence). Snakes may have
moved to prey on the high density of H.
frenatus living within the security of man 
made structures along the fence transect (Air 
port Fence).
The general conclusions that we draw from
these data are that: (1) snake populations
on Guam have expanded irregularly over a
period of more than 35 years, with increases
in numbers and area ofdistribution; (2) snake
densities probably approached 100 per ha in
some areas when the snake was reaching its
peak for the first time; and (3) snakes may
again reach high densities in areas that have
already experienced loss of most native prey
and sharp declines in snake numbers.
PACIFIC SCIENCE, Volume 46, January 1992
ACKNOWLEDGMENTS
Among the many people who assisted with
the fieldwork were J. A. Savidge, H. Muna,
J. P. Guerrero, R. J. Rondeau, E. W. Camp 
bell III, S. W. Gotte, M. J. McCoid, G. J.
Wiles, B. E. Smith, D. E. Wilson, G. R.
Grimm, M. J. Altenbach, G. Neubauer, and
N. J. Scott. Guam's Division of Aquatic and
Wildlife Resources has generously provided
support, unpublished data, and facilities for
researchers on Guam. O. E. Maughan and
J. L. Oldemeyer provided administrative sup 
port. The military services have provided
logistical support as well as ready access to
their bases. E. P. Collins and R. J. Rondeau
prepared the figures. The encouragement of
C. Imamura and E. Kosaka has been impor 
tant to our continued study of this snake.
M. J. McCoid, O. E. Maughan, R. J.
Rondeau, and E. A. Kay suggested improve 
ments to the manuscript.
LITERATURE CITED
ANONYMOUS, 1962. Glimpses of Guam 1962.
U.S. Navy Relief Fund Drive, Agana,
Guam.
BEARDSLEY, C. 1964. Guam past and present.
C. E. Tuttle, Tokyo.
DOWNS, B. H., Jr. 1949. Snakes in paradise!
One found in Wake. Honolulu Advertiser,
6 March 1949.
DRYDEN, G. L., and E. H. TAYLOR. 1969.
Reptiles from the Mariana and Caroline
islands. Univ. Kans. Sci. Bull. 48: 269-279.
ENGBRING, J., and T. H. FRITTS. 1988. Demise
of an insular avifauna: The Brown Tree
Snake on Guam. Trans. West. Sect. Wildi.
Soc. 24:31-37.
FOSBERG, F. R. 1960. The vegetation of Mi 
cronesia. Bull. Am. Mus. Nat. Hist. 119: 1 
75.
FRITTS, T. H. 1988. The Brown Tree Snake,
Boiga irregularis, a threat to Pacific islands.
U.S. Fish Wildi. Serv., BioI. Rep. 88(31).
FRITTS, T. H., and M. J. MCCOID. In press.
Predation by the brown tree snake on poul 
try and other domesticated animals in
Guam. Snake.
Brown Tree Snake on Guam-RoDDA, FRITTS, AND CONRY 57
FRITTS, T. H., M. J. MCCOID, and R. L.
HADDOCK. 1990. Risks to infants from the
bites ofthe Brown Tree Snake (Boiga irreg 
ularis) on Guam. Am. J. Trop. Med. Hyg.
42(6): 607-611.
FRITTS, T. H., N. J. SCOTT, Jr., and J. A.
SAVIDGE. 1987. Activity of the arboreal
Brown Tree Snake (Boiga irregularis) on
Guam as determined by electrical outages.
Snake 19: 51-58.
FRITTS, T. H., N. J. SCOTT, Jr., and B. E.
SMITH. 1989. Trapping Boiga irregularis on
Guam using bird odors. J. Herpetol. 23:
189-192.
GUAM, OFFICE OF THE GOVERNOR. 1955. An 
nual report of the Governor of Guam.
Office of the Governor, Agana, Guam.
LIGHTBODY, M., and T. WHEELER. 1985. Pa 
pua New Guinea, a travel survival kit.
Lonely Planet, South Yarra, Victoria,
Australia.
MEAD, M. 1956. New lives for old: Cultural
transformation-Manus, 1928-1953. Wm.
Morrow, New York.
PARKER, W. S., and M. V. PLUMMER. 1987.
Population ecology. Pages 253-301 in R. A.
Siegel, J. T. Collins, and S. S. Novak, eds.
Snakes: Ecology and evolutionary biology.
Macmillan, New York.
PENDLETON, R. C. 1947. A snake "den" tree
on Guada1canal Island. Herpetologica
3: 189-190.
SAS INSTITUTE INC. 1988. SAS/STATTM
user's guide, release 6.03 ed. SAS Institute
Inc., Cary, North Carolina.
SAVIDGE, J. A. 1986. The role of disease
and predation in the decline of Guam's
avifauna. Ph.D. diss., University ofIllinois,
Urbana-Champaign.
---. 1987. Extinction of an island forest
avifauna by an introduced snake. Ecology
68: 660-668.
SMITH, A. G. 1949. Notes on the herpetology
of Guam, Marianas Islands. Nat. Hist.
Misc. (37): 1-2.
STEVENS, R. L. 1953. Guam, U.S.A., birth
of a territory. Tongg Publishing Co.,
Honolulu.
STONE, B. C. 1970. The flora ofGuam. Micro 
nesica 6: 1-659.
WHITE, G. c., D. R. ANDERSON, K. P.
BURNHAM, and D. L. OTIS. 1982. Capture 
recapture and removal methods for sam 
pling closed populations. Los Alamos
National Laboratory, Los Alamos, New
Mexico.
WILES, G. J. 1987. Current research and future
management of Marianas Fruit Bats
(Chiroptera: Pteropodidae) on Guam.
Aust. Mammal. 10: 93-95.
ZWINNENBERG, A. J. 1978. Die Braune
Nachtbaumnatter Boiga irregularis. Aquar.
Terrar. Z. 31: 177-179.