Population Densities of Poison Dart Frogs in a Regenerating Tropical Forest as Measured by the Hayne Estimator A Thesis Presented by Jennifer Rose Bunnell Miller To the Joint Science Department Of The Claremont Colleges In partial fulfillment of The degree of Bachelor of Arts Senior Thesis in Organismal Biology April 2007
69
Embed
Population Densities of Poison Dart Frogs in a ...costarica.jsd.claremont.edu/pdf/Thesis - Jennie Miller - 2007.pdfdistribution, many poison dart frog species have a range that spans
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Population Densities of Poison Dart Frogs in a Regenerating Tropical Forest
With amphibian populations declining throughout the world, there is an increasing
demand for effective tools to measure species responses to environmental change. This study
investigates the effectiveness of the Hayne Estimator in evaluating the densities of two
species of poison dart frogs in three Costa Rican lowland forest habitats with varying degrees
of recovery from deforestation (selectively-logged riparian forest, post-pasture secondary
forest and non-native bamboo plantation forest). Population densities of Dendrobates
granuliferus and Dendrobates auratus were significantly highest in riparian forest,
substantially lower in bamboo, and very low in secondary forest. This trend corresponds to
previous research on species recolonization after deforestation and subsequent regrowth and
indicates that the Hayne Estimator is well suited for the evaluation of poison dart frogs.
Abiotic factors such as proximity to water, rainfall, temperature and time of day were found
to have some effect on frog sighting frequency. Individuals of both species tended to
aggregate near water, but the proportional distribution of transects according to all habitat
water presence likely negated this effect. Rainfall was unrelated to the sighting frequency of
D. auratus but correlated with the sighting frequency of D. granuliferus. Air temperature did
not impact sighting frequency. Time of day, however, was found to influence the sighting
frequencies of both species, with peaks occurring in the early morning and late afternoon.
The robustness of the Hayne Estimator when used to monitor poison dart frogs suggests that
the technique may be a valuable tool for future conservation research.
4
INTRODUCTION
Since scientists gathered at the First World Congress of Herpetology in 1989 to
address the worldwide decline of amphibian populations, concern for these creatures has
increased at an accelerating rate (Phillips, 1990; Stuart et al., 2004). Now, nearly 2 decades
later, as populations continue to decrease in size and weaken in stability, scientists are calling
for the unprecedented cooperation of all to prevent further loss of amphibian diversity. In
2006, 49 accomplished herpetologists co-authored a forum in Science that announced the
disappearance of 122 species and identified 32.5% of known amphibians as threatened
(Mendelson III et al.). The group asserted that only through the union of “individuals,
governments, foundations, and the wider conservation community” would the escalating rate
of extinctions slow. Their recommendations echo the suggestions of other scientists and
necessitate the implementation of monitoring, surveys, habitat protection and breeding
research colonies in an international effort to ensure the continued existence of amphibians.
Long-term monitoring programs serve as a foundation of restoration ecology planning
because they reflect the responses of at-risk species to environmental change. Knowledge
about the population health of sensitive species in recovering habitats is invaluable to the
conservation community. These studies not only increase the likelihood of successful land
management of protected areas, but they also guide policy towards accurately prioritizing the
protection of land in regions and habitats that are greatly valued for preserving biodiversity.
Without a doubt, Latin America has been more severely impacted by amphibian
declines than any other region in the world (International Union for Conservation of Nature
and Natural Resources [IUCN], 2006). In Central and South America, over 30 genera, 9
families and 1,157 species of amphibians have declined or gone extinct (Young et al., 2001;
5
6
IUCN, 2006; Pounds et al., 2006). The tropical regions of this area have supported a high
diversity of amphibian species for millennia, leading the region to be classified as a
biodiversity hotspot (Figure 1; Myers et al., 2000; Brooks et al., 2002). When one considers
the increasing rate of new amphibian species discoveries (Donoghue and Alverson, 2000),
the potential number of amphibians that may have been harmed by human activities is
astounding.
Because of the extensive body of tropical ecological research conducted in Costa
Rica, this country has become a paradigm for understanding patterns of species decline in
other tropical locations. Although Costa Rica covers only 0.03% of the planet’s surface, it is
home to 4% of the world’s biodiversity and 3% of the world’s threatened amphibians (IUCN,
2006; World Resources Institute, 2006; National Biodiversity Institute, 2007). To date, one
amphibian species has gone extinct in the country (Bufo periglenes, the golden toad) and 64
species are considered threatened (IUCN, 2006). Although more than one-fifth of Costa
Rican land is protected, it is clear that further action must be taken in order to raise, or at
least sustain, the current level of biodiversity (World Resources Institute, 2006).
Like many other countries throughout the world, Costa Rica has been the site of
rampant deforestation over the past few centuries. However, human habitation has followed a
unique trend within the last several decades. With increasing job opportunities in the urban-
based tourism and textile industries, workers have begun to migrate away from agricultural
areas (Aide and Grau, 2004). Since 1960, the rural population in Latin America and the
Caribbean has dropped by 30%, a trend driven in part by a 20 million person decrease in the
population whose livelihood is based on agriculture, hunting, fishing or
7
Figure 1. The hotspots of the world. Costa Rica is classified as part of the Mesoamerica hotspot, which extends from southern
Mexico to Panama (image from Myers et al., 2000).
forestry since 1980 (Food and Agriculture Organization of the United Nations, 2004). While
much of the vacant land has been sold for the expansion of other farms, a substantial amount
has been abandoned, providing an opportunity for regrowth, recolonization and the
reestablishment of natural ecosystems.
While ecologists have extensively explored species’ responses to the degradation of
native habitat, less work has been done on recovering habitats. With the current trend in
Costa Rica favoring natural regrowth, and with the increasing public awareness about the
causes of global warming, a large movement to restore farmed and developed land across the
planet could occur within the next century. Species monitoring programs must be
implemented in order to predict, prepare and assist the associated changes in biodiversity.
In 2001, Pitzer College acquired the Firestone Center, a parcel of land in Costa Rica
that had previously been selectively stripped of forest and converted into a cattle farm and
Figure 2. The Firestone
Reserve, otherwise known as
the Finca la Isla del Cielo, is
owned by Pitzer College and
located near Dominical in
southwestern Costa Rica
(Firestone Center for
Restoration Ecology, 2006).
8
monoculture plantation (Figure 2). A series of changes in ownership has allowed the
Firestone Reserve to have 13 years of continuous natural regeneration. Today, the preserve is
an ideal study site for investigating the process of regrowth. The variation in land patch
quality permits the juxtaposition of species in native riparian habitat versus secondary and
non-native plantations that have had over a decade to recover. The Joint Science Department
of the Claremont Colleges launched a student research program in the summer of 2005 and
has plans to establish long-term monitoring programs to track the regeneration progress.
Since it is unrealistic to study the impacts of human activities on all species in a given
area, indicator species are studied as representatives of a larger group. Indicators may belong
to any taxonomic group, but are commonly characterized by a degree of sensitivity to
disturbance that mirrors the responses of a wide variety of other species (Landres et al.,
1988). Anurans are ideal indicators because all stages of their life cycles are highly
dependent on environmental conditions. Most frogs and toads require a permanent water
source for reproduction, the development of young, and a source of food. Their skins are
permeable to permit survival in water and on land, leaving their bodies vulnerable to the
chemical balance around them. The majority of anurans also consume insects, a group known
to shift radically with a change in vegetation (Gibbs and Stanton, 2001). Based on these
characteristics and others, anurans are especially susceptible to habitat loss, chemical
contamination, climate change and the introduction of exotic species and disease, factors
known as the leading causes of decline in other amphibian species as well (Young et al.,
2001).
Although anurans are prized for the insight they provide regarding the health of an
ecosystem, the creatures are also the arch nemeses of many field scientists. Cryptic and
9
nocturnal, the typical frog or toad is a challenge to study in its natural environment. The
endemic poison dart frogs of Costa Rica, however, provide a colorful alternative to studying
cryptic indicators in the tropics. All members of the family Dendrobatidae are
aposematically colored, diurnally active and easily identified to species. Seventeen species of
Dendrobates have been identified in Central America and more than 100 species are known
to South America (Maxson and Myers, 1985). In addition to the two-continent family
distribution, many poison dart frog species have a range that spans several countries. Data
collected on frogs in one region can thus be readily applied to an entirely different area.
In an effort to measure the biodiversity status in the regenerating Firestone Reserve
habitats, the abundances of two poison dart frog species were measured. Both the granular
Error!
Figure 3. Many frogs utilize camouflage to hide from predators and field scientists alike
(left, Hyla versicolor), whereas Dendrobatids have conspicuous skin color and patterns to
contrast against their background (right, Dendrobates azureus). Their aposematic coloration
conveys to predators the consequences of a quick snack. Photos courtesy of
www.livingunderworld.org and www.webshots.com
10
Figure 4. The study species: Dendrobates granuliferus (top) and Dendrobates auratus
(bottom). Photographs by Keith Christenson.
11
(Dendrobates granuliferus) and the green and black (Dendrobate auratus) poison dart frog
occur naturally on the preserve (Figure 4). The population densities of these species were
assessed in riparian, secondary and bamboo forest habitats during the early wet season of
October 2006. Data were collected and calculated using the Hayne Estimator technique
(Hayne, 1949), which utilizes measurements taken from observations of sighting angle and
distance to each animal. The densities were then applied to approximate frog recolonization
in the secondary and bamboo habitats as compared to the more pristine riparian habitat.
Because the method assumes that an individual will flush and be readily noticeable as
the observer approaches, the Hayne Estimator is not well designed for the cryptic, nocturnal
habits of most amphibians but has been repeatedly employed to evaluate populations of birds
and mammals (Coulson and Raines, 1985; Pelletier and Krebs, 1997). The conspicuous
coloration and diurnal activity periods of Dendrobatid frogs makes them potentially
appropriate for the Hayne Estimator technique. This study explores the utility of poison dart
frogs as subjects for the Hayne Estimator while investigating the quality of vegetation
regrowth at the Firestone Reserve as a means of supporting native levels of biodiversity.
To account for the impact of abiotic factors on frog sighting frequencies, proximity to
water, as well as correlations with rainfall, air temperature and time of day, were considered.
Because past studies indicate that poison dart frogs do not depend on large bodies of water
(reviewed by Savage, 1968; Vences et al., 2000; Jowers and Downie, 2005), random
distribution was expected. Rainfall and time of day have both been identified as influential
factors, with some Dendrobatids occurring in larger quantities in the presence of rain and in
the early morning and late afternoon (Graves, 1999). Finally, the air temperature was not
12
anticipated to affect sighting frequencies because of its small range due to the tropical
climate.
13
MATERIALS AND METHODS
Study Area
Field research was conducted at the Firestone Center for Restoration Ecology with the
permission of Pitzer College and the Joint Science Department. Claremont Colleges. The
Firestone Reserve is a 60 ha protected preserve of lowland (15m – 303 m) Pacific Moist
Forest in southwestern Costa Rica near Dominical (16.684 N, 51.643 W). The reserve has a
unique history that makes the area a suitable research site for an examination of poison dart
frog populations in regenerating habitats. Beginning around 1950, the property was
completed deforested, with the exception of two precipitous stream canyons within which
circa 100 m wide strips of riparian forest were only selectively logged (Firestone Center for
Restoration Ecology, 2006). The land was utilized as a cattle farm until 1993, when the
property was purchased by Ms. Firestone and converted into a combined sustainable farm
and private biological preserve. At this time, livestock were removed and parts of the land
were replanted with monoculture crops, including 5.9 ha of bamboo (Guadua aculeata, G.
angustifolia, Dendrocalamus asper, and D. latiflorus), 1 ha of bananas (Musa acuminata), 1
ha of black palm (Bactris gasipaes), and 24.7 ha of mixed hardwood tree species.1 The
remaining 27.4 ha were allowed to regenerate naturally. In 2005, the property was donated to
Pitzer College and farming maintenance was abandoned. The land has since been left alone
to regrow and is currently used as a biological reserve for education and research by the
Pitzer Study Abroad Program and the Claremont Colleges Joint Science Department.
The division of the reserve into multiple sub-habitats makes it an ideal location for
the study of biodiversity in recovering natural and non-native vegetation. The Firestone
1 Refer to http://costarica.jsd.claremont.edu/biodiversity/trees.shtml for an up-to-date listing of identified species.
14
15
Reserve borders the Hacienda Baru National Wildlife Refuge to form a contiguous 390 ha
sanctuary dedicated to scientific study and eco-tourism with minimal biological impact
(Hacienda Baru National Wildlife Refuge, accessed 2006).
This study focuses on Dendrobatid presence in three types of habitat on the Firestone
Reserve: selectively-logged riparian forest (from hereon referred to as “riparian”), abandoned
pasture (“secondary”) and bamboo plantation (“bamboo”; see Figure 5). The riparian regions
consist of primary forest with tall vegetation and dense canopy cover. The ground in this
habitat is shaded during most of the day and was usually covered by thick leaf litter. No
records are available that describe the method of selective logging in this habitat, leaving no
way to judge whether the land is an accurate standard of natural vegetation. However, the
riparian habitat of the Firestone Reserve is visually indistinguishable from the primary forest
of the Hacienda Baru National Wildlife Refuge. Therefore, the riparian habitat was used in
this study as a representative of natural forest conditions, although it should be recognized
that there are potential influences of the past selective logging that are unmeasured in this
study.
The secondary forest is comprised of lower, thinner trees than the riparian habitat.
Large patches of sunlight are often observed on the ground, causing grasses to replace damp
leaf litter throughout much of the secondary forest (Figure 5). The bamboo habitat features
thick groves of tall culms that provide moderate amounts of shade. Sunlight filters through
the vegetation at a lower intensity than in the secondary forest, and dense, knee-high
vegetation covers the ground around the bamboo. Multiple sources of water exist in all three
sampling habitats. Small and moderate, 1-10 m wide streams run through both the riparian
and secondary forests, while three large ponds border the bamboo habitat.
16
Figure 5. A map of the Firestone Reserve habitats with images of the three habitats of study: bamboo (top left), riparian forest (bottom left) and secondary forest (top right). The
six transects are numbered and differentiated by color. Prominent water sources are represented by light blue symbols (filled polygon = pond; solid line = stream, known location;
dotted line = stream, estimated location [surveyed by McFarlane, 2001 {unpubl. data}]). Photographs by author.
Observations were systematically made within each habitat by following the pre-
established trails of the reserve as transects for observation. The use of four maintained trails
permitted an accessible and repeatable loop through the forest and covered all three types of
habitat. The trails were divided into six transects of varying lengths, with most transects
covering multiple habitats (Table 1). The order that transects were walked was randomized
when possible so as not to cause unintentional correlations with time. Transects 1 and 4 could
not be shuffled because they provided starting and ending access or connected path loops,
respectively.
Table 1. Length distributions for study transects.
Habitat length (m) Transect
Trail ID*
Total length (m) Riparian Secondary Bamboo
1 WT 275.9 275.9 0 0 2 B 949.9 84.1 865.8 0 3 C 838.6 721.3 117.3 0 4 B 646.2 194.5 451.7 0 5 BB 985.6 0 236.6 749.0 6 C 100.5 100.5 0 0
*Trail ID corresponding to the survey by McFarlane, 2001 (unpubl. data).
Study Species
The combined Firestone-Hacienda Baru area hosts at least 28 known species of
anurans (M. Ryan, pers. comm.), including Dendrobates granuliferus, the granular poison
dart frog, and D. auratus, the green and black poison dart frog. While both species are
abundant on the Firestone Reserve, D. granuliferus is internationally recognized as a
threatened species due to habitat loss and degradation as well as human harvesting of the
species (IUCN, 2006). The range of D. granuliferus is also limited, covering 5,579 km2 from
17
the mid-western coastal lowlands of Costa Rica to the northern border of Panama (Global
Amphibian Assessment, 2006; IUCN, 2006). Dendrobates auratus is considered to be of
lesser concern, largely because of its greater range of 11,944 km2 from northern Costa Rica
through northern Columbia and higher tolerance of habitat degradation.
Dendrobates granuliferus and D. auratus were selected as study subjects because of
their relevance to amphibian declines and their conspicuous appearances in the field.
Dendrobatids have many natural history characteristics typical of tropical amphibians. All
species are diurnal and commonly live among the low vegetation and leaf litter of moist
forests below elevations of 3,000 m (Savage, 1968). They are considered terrestrial anurans
because their life cycles are independent of large water sources. Dendrobatid eggs are laid on
land and tadpoles are carried on the backs of their parents to temporary puddles of water
among vegetation. Dendrobatids specialize in eating ants but also consume a large quantity
of mites, insects that are also characteristic of the diets of other tropical amphibians such as
Atelopus, Bufo and Bolitoglossus (Toft, 1981; Anderson and Mathis, 1999). They mate
during the wet season like many other tropical amphibians, and they are most active between
May and November (reviewed by Savage, 2002). Because many of the human impacts that
threaten D. granuliferus and D. auratus also affect other tropical amphibians and potentially
other groups of organisms, these two species serve well as indicators of the status of tropical
wildlife populations.
In addition, these species were chosen because their unique aposematic coloration
makes them convenient to study in the field. While many anuran species are nocturnal and
camouflaged to their environments, Dendrobatids are diurnal and have brilliantly colored
skin markings. The coloration serves as a signal to predators, warning them of the toxic
18
alkaloids that can be released from the frogs’ skin glands as a mechanism of defense
(Saporito et al., 2004). The distinctive patterns of D. granuliferus and D. auratus permit easy
sighting of individuals and allowed for a high confidence in the accuracy of the field
techniques used in this study.
Frogs were observed on the Firestone Reserve during the wet season between 6
October and 14 October 2006. The research period corresponded to the Dendrobatid mating
season and the peak of their activity throughout the year (reviewed by Savage, 2002).
Observing at this time guaranteed the highest number of frog sightings possible, leading to
elevated estimates of population densities and an overall optimistic perspective of the
Dendrobatid presence on the Firestone Reserve.
Population Density
The population density of frogs was measured with the Hayne Estimator (Hayne,
1949). To keep measurement technique consistent, all observations were made by the author.
Two sessions of observations typically occurred each day. The first session began at
approximately 7:00 and ended around 11:00 and the second began at approximately 13:30
and ended around 16:30. Transects were walked at a constant speed from start to stop without
pause, except to record frog measurements. Consequentially, transects with many frog
sightings took longer to walk than transects with few sightings.
Each observation followed the same protocol. When a frog was sighted, the observer
immediately took three measurements (Figure 6):
(1) The distance from the observer to the frog’s location at first sighting. Measurements
were made using a Leica Geosytems laser rangefinder accurate to ± 3mm and later
19
trigonomically corrected from incline distances (i.e. from the height of the hand-held
rangefinder) to true plan distances.
(2) The magnetic bearings of the transect and frog, using a Suunto sighting compass
readable to ± 0.5 degrees.
(3) The time of the sighting.
Occasionally, when a frog was observed well beyond the first possible point of contact, the
observer back-tracked her steps until she reached the location where the frog first came into
view. For example, if a frog was first noticed when the observer was directly beside it, the
observer retraced her steps until she could first view the frog amidst the vegetation.
Obscurities due to vegetation occasionally caused sighting difficulties, but errors were most
likely not frequent enough to largely impact data. This technique corrected for the limitation
Figure 6. A visual representation of the
Hayne Estimator data collection
technique, showing the distance from the
observer to the frog (ri) and the
corresponding measured sighting area
(shaded red).
20
of being able to view only one side of a transect at a time.
The location of each frog was recorded relative to a surveyed map of the reserve
paths. A survey of current habitat borders was mapped during the study period and overlaid
on the original path survey. Survey information was used to relate frog sighting to habitat
type for use in the Hayne Estimator. Total transect length and average segment (i.e. the
distance between transect turns) length were also collected from the survey.
The population densities of D. granuliferus and D. aruatus for each observation
session were calculated for each habitat using the unmodified Hayne Estimator:
Dh =n
2L1n
1rii= t
n
∑⎛
⎝ ⎜
⎞
⎠ ⎟ ,
where Dh is the Hayne density estimate, n is the number of animals observed, L is the
transect length, and ri is the sighting distance to the ith animal. The standard deviation was
calculated by taking the square root of the variance, calculated as:
Variance(DH ) ≈ DH2 var(n)
n2 +
1r,
− R⎛
⎝ ⎜
⎞
⎠ ⎟
2
i= t
n
∑R2n n −1( )
⎡
⎣
⎢ ⎢ ⎢ ⎢ ⎢
⎤
⎦
⎥ ⎥ ⎥ , ⎥ ⎥
where R is the mean of the reciprocal of the sighting distances and calculated as:
R =1n
1rii= t
n
∑ ,
Circular statistics on sighting angles were computed using the StatistiXL Excel Add-
In (http://www.statistixl.com/). Population densities were analyzed with VassarStats (Lowry,
2007) for statistical differences between the species and habitats using One-Way Independent
ANOVA and Tukey HSD tests.
21
Distribution
To relate frog sightings to actual geographical features, ArcGIS Version 9.1 was used
to project the Firestone trail survey into a satellite image of the Firestone Reserve (obtained
from Digital Globe, Inc.) with reference to GPS coordinates collected at the site. Habitat
zones were constructed using several older habitat maps of the reserve as well as records of
current habitat boundaries taken during the study. Frog sighting points were imported and
displayed with graduated symbols to represent point densities. Water sources (streams and
ponds) were approximated and drawn by hand according to the trail survey (McFarlane,
2001, unpubl. data) and satellite image.
The proximity of each sighting to water was determined using COMPASS software
(Version 5.05; Fish, 2005) The straight-line distance between each sighting location and the
closest water source was measured and then correlated to the number of frogs sighted at the
location using a linear regression calculated with VassarStats (Lowry, 2007).
Abiotic Factors
To determine whether the transect distribution proportionally represented the amount
of water in each habitat, an analysis of transect-resource proportionality was conducted using
measurements from the Firestone maps created with ArcGIS to compare the ratio of the
habitat area within 50 m of a water source to the total habitat area versus the transect length
(by habitat) to the total transect length (by habitat). In other words,
area of habitat within 50m of watertotal area of habitat
: tran sec t length in habitat within 50m of watertotal length of tran sec t in habitat
22
Rainfall and air temperature (from now on referred to as “reserve temperature”) were
measured every 2 hours by a Davis Weatherlink meteorological station on the Firestone
Reserve. Average reserve temperature was calculated for each increment as the arithmetic
mean of the high and low temperatures. To test for a correlation between reserve temperature
and the frog sighting frequency, data were analyzed using a linear regression calculated with
Vassar Stats (Lowry, 2007). VassarStats was also used to determine whether a correlation
existed between rainfall and frog sighting frequency with a Pearson’s chi-square 2x2
contingency table test. To evaluate overall trends in frog sighting frequency, data from both
species were combined and compared to the time of sighting.
The air temperature in each habitat (from now on referred to as “habitat temperature”)
was measured using four temperature loggers (Stow Away XTI). One logger was attached to
a tree in each habitat and the sensor was oriented to hang freely (Figure 7). The loggers were
positioned so that they received light levels typical of the particular habitat (i.e. not in full
sunlight). A fourth control logger was set in a deforested meadow on the reserve to measure
the highest possible daily temperature (i.e. full sunlight). Loggers were set to record data for
each day and night of the study period and measured the habitat temperature every 5 or 20
minutes, depending on the format available on the logger. Temperature data from all the days
in the study period were averaged to find the 24-hour mean temperature fluctuation for each
habitat. The fluctuations of all habitats were then compared to determine whether a large
difference in temperature in any of the habitats may have influenced poison dart frog activity
levels.
23
Figure 7. Locations of the temperature loggers in each habitat: bamboo (left top), secondary
(bottom left), riparian (top right) and exposed meadow (bottom right). Photographs by
author.
24
RESULTS
Inconsistencies in data collected at the start of the study period have led to the
exclusion of several days of data from the final analysis. The number of frog sightings in the
first three days was significantly lower than sightings during the remainder of the observation
days (an average of 6 ± 6 observed frogs/km in contrast with 85 ± 39 observed frogs/km).
Additionally, no significant differences were found in abiotic factors such as rainfall or
temperature between the first 3 days and the subsequent days of observation. The lack of
disparity suggests that the initial low number of sightings was likely a result of the observer’s
learning period. A test run was not conducted ahead of time, leading the observer to learn the
sighting and measurement techniques during the official study period. Therefore, only data
from 9 October through 14 October 2006 were included in the analysis (Appendix A). Data
collected from 6 October to 8 October are listed in Appendix B but are not considered valid
data or incorporated into the thesis.
Population Densities and Distributions
A total of 166 D. granuliferus and 109 D. auratus were observed, resulting in a total
sample size of 275 frogs. The average population densities for both species were larger in the
riparian forest than in the secondary or bamboo habitats (Figure 8). For D. granuliferus, the
riparian density was estimated to be 68 times greater than the secondary density and 23 times
greater than the bamboo density. The D. auratus riparian density estimate was 155 times
greater than the secondary density but only three times greater than the bamboo density.
Bamboo densities were larger than secondary densities for both species, with density for D.
granuliferus in bamboo reaching an estimate that was three times larger than for secondary
25
forest and the density for D. auratus in bamboo estimated to be 47 times larger than in
secondary forest. ANOVA analysis indicated significant effects of habitats on densities for
both D. granuliferus and D. auratus (F = 20.31, df = 2, P < 0.0001; F = 28.62, df = 2, P <
0.0001, respectively).
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
2.2
Riparian Secondary Bamboo
Habitat Type
Den
sit
y (
fro
gs/
ha)
Figure 8. Average population densities of D. granuliferus (solid) and D. auratus (open) by
habitat. Vertical lines represent one standard deviation.
26
Population density trends differed through habitats between the two poison dart frog
species. The presence of D. granuliferus was nearly twice that of D. auratus in the riparian
forest, while the density of the D. auratus was more than four times larger than in the
bamboo. The population densities of both species in the secondary forest were very low,
although results indicated that D. granuliferus was sighted more often than D. auratus.
Tukey HSD tests found densities of riparian versus secondary habitats and riparian versus
bamboo habitats to be significantly different, but secondary versus bamboo habitats to be not
significant (P < 0.01, P < 0.01, P > 0.5, respectively).
A clear correlation between the number of frog sightings and proximity to water was
apparent for both species (Figure 9). Analysis with a linear regression indicated a significant
negative relationship between the number of sightings and the distances of the sighting
locations to a water source (y = -0.0476x + 13.478, df = 1, r2 = 0.207, P < 0.01). Three
particularly dense clusters of frog sightings are apparent in the riparian and secondary
habitats near streams, while sightings in the bamboo habitat did not appear to be correlated to
water (Figure 10).
The distributions of D. granuliferus and D. auratus were generally very similar.
Individuals of both species were found simultaneously at the same locations on multiple
occasions. Only two locations throughout the study site indicated the dominating presence of
one species without the other (Figure 11). For one, there is a distinct difference in the number
of D. auratus (n = 8) found in bamboo compared to D. granuliferus (n = 2). However, the
small sample size undermines the strength of this disparity. A second conspicuous
dissimilarity in distribution occurred on the southernmost stream where the trail dips towards
the southern stream.
27
Figure 9. A scatterplot showing the negative relationship between the number of sightings
(D. granuliferus and D. auratus combined) and the distance of the sighting location to a
water source in all three habitats. A linear trend line has been fit to the dat
y = -0.0476x + 13.478r2 = 0.2072
20
25
30
35
200 250 300
)
of
sigh
tin
g
0
5
10
15
0 50 100 150
Distance to water source (m
Nu
mber
s
28
Figure 10. Distribution of all frog sightings through the habitats of the reserve (D. granuliferus and D. auratus combined). Where multiple frogs were seen in the same location,
sighting frequencies are symbolized by graduated circles (see legend). Prominent water sources are represented by blue symbols (filled polygon = pond; solid line = stream, known
location; dotted line = stream, estimated location [surveyed by McFarlane, 2001 {unpubl. data}]).
29
Figure 11a. Distribution of Dendrobates granuliferus sightings through the habitats of the reserve. Where multiple frogs were seen in the same location, sighting frequencies are
symbolized by graduated circles (see legend). Prominent water sources are represented by light blue symbols (filled polygon = pond; solid line = stream, known location; dotted
line = stream, estimated location [surveyed by McFarlane, 2001 {unpubl. data}]).
30
31
Figure 11b. Distribution of Dendrobates auratus sightings through the habitats of the reserve. Where multiple frogs were seen in the same location, sighting frequencies are
symbolized by graduated circles (see legend). Prominent water sources are represented by light blue symbols (filled polygon = pond; solid line = stream, known location; dotted
line = stream, estimated location [surveyed by McFarlane, 2001 {unpubl. data}]).
Rainfall
No significant correlations were found between rainfall (i.e. rain falling at the time of
observation) and D. auratus sightings (Pearson=1.41, P=0.24). A significant correlation was
found for D. granuliferus sightings (Pearson=5.53, P=0.02), indicating that the frequency of
frog sighting increased in the absence of rainfall and decreased during rainfall (Figure 12).
0
10
20
30
40
50
60
Present Absent
Rainfall status at time of observation
Num
ber o
f tra
nsec
ts w
ith fr
ogs
obse
rved
Figure 12. A bar graph showing the negative correlation between rainfall (rain falling when
frogs observed) and D. granuliferus sightings (open bars = transects on which frogs were
observed; shaded bars = transects on which frogs were not observed).
32
Temperature
Regression analyses found no significant relationship between temperature and frog
sightings for either D. granuliferus or D. auratus (y = -0.0031x + 0.0962, df = 17, P > 0.05,
r2 = 0.1378; y = -0.0012x + 0.0434, df = 17, P > 0.05, r2 = 0.0587, respectively). Temperature
varied only 7ºC according to the Firestone meteorological station during the time of
observation and ranged from 23ºC and 30ºC.
No large temperature differences were found between the riparian, secondary and
bamboo habitats. The temperatures of the study habitats consistently remained within 1˚ of
each other (Figure 13). In contrast, temperatures recorded in the deforested meadow
remained higher than in the study habitats, peaking at 6.6˚C higher than in the other habitats.
The temperature loggers indicated that the temperature in the three study habitats ranged
from 22.5˚C to 28˚C while the temperature in the deforested meadow ranged from 23.5˚C to
34.0˚C.
33
20
22
24
26
28
30
32
34
36
0:00
2:00
4:00
6:00
8:00
10:0
0
12:0
0
14:0
0
16:0
0
18:0
0
20:0
0
22:0
0
Time
Avera
ge a
ir t
em
pera
ture
(˚
C)
BambooSecondaryRiparianDeforested meadow
Figure 13. Mean daily temperature fluctuations in each study habitat and control
environment (deforested meadow).
Time of day
To evaluate overall trends in frog sighting frequency, data from both species were
combined and compared to the time of sighting (Figure 14). A large increase in the sighting
frequency was observed in the early morning (7:00 to 8:00), followed by relatively constant
rates in the later morning and early afternoon (8:00 to 12:00 and 13:00 to 14:00; no data was
collected from 12:00 to 13:00). From 14:00 to 15:00, a sharp decrease in sighting frequency
34
occurred, followed by a substantial increase in the late afternoon (16:00 to 17:00) and a sharp
decrease in the mid afternoon (14:00 to 15:00).
Data separated by species show less extreme patterns of change over time. Sighting
frequency of D. granuliferus peaked in the early morning (7:00 to 8:00) and rose again in the
later morning (10:00 to 11:00) and mid afternoon (15:00 to 17:00), but remained
approximately constant at all other times of the day (Figure 15). Dendrobates auratus
sighting frequency also peaked in the early morning (7:00 to 8:00) but remained relatively
constant throughout the rest the day, with low dips in activity in the mid morning (10:00 to