SMALL MAMMAL COMMUNITIES
IN AN EASTERN BRAZILIAN PARK
JODY R. STALLINGS
A THESIS PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN
PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
This project could not have been carried out without the help
of numerous people. Foremost, I thank John G. Robinson for
suggesting the research project in the Rio Doce State Forestry
Park. His perception and patience aided in all aspects of the
project and his visit to the project site in May, 1986, was a great
boost to my morale. I would also like to thank the other members
of my committee for their suggestions during the planning stage and
for helping me to interpret the data. I thank Drs. John F.
Eisenberg, Melvin E. Sunquist, Wayne R. Marion, and Nigel J. H.
Smith. John Eisenberg was a constant source of support through the
analysis and writing stages. His continued interest in the results
and concern for my welfare helped to carry me through difficult
periods. Kent H. Redford provided obscure and relevant
publications. He also visited the field site at the beginning of
the project and encouraged me to sample the eucalypt forests.
Funding for this project was provided from the Program for
Studies in Tropical Conservation at the University of Florida,
World Wildlife Fund-US, The Organization of American States, and
the Instituto Estadual de Florestas (IEF). I thank John Robinson,
Kent Redford, Russell Mittermeier, John Eisenberg, Jose Carlos
Carvalho, and Cello Valle for helping me to obtain the funds
necessary to carry out the project.
I would like to thank Jose Carlos Carvalho, president of IEF,
for allowing me to work in the Rio Doce State Forestry Park. Dr.
Cello Valle, Ilmar Bastos Santos, and Gustavo A. B. da Fonseca were
very instrumental in helping me to obtain my scientific expedition
visa to conduct research in Brazil. Celio Valle and Gustavo
Fonseca were more than helpful throughout the various stages of
fieldwork and allowed me to work in the mammalogy laboratory in
Belo Horizonte. Gustavo Fonseca was more than generous in sharing
his computers with me. Dr. Wilson Mayrink and his excellent staff
treated me in a concerned and professional manner. I thank them
for saving my nose.
The professional staff at the Rio Doce State Forestry Park
contributed to the project in several ways. Ademir Camara Lopes,
the park Administrator, helped us get settled into the Park. He
offered no cost housing for the duration of the project and
contributed a modest amount of the Park's gasoline supply for my
use in the project. He also was able to obtain funds from IEF for
the construction of arboreal platforms and 60 wire traps.
Hermogenes Ferreira S. Neto was extremely important in solving
daily problems and in handling our correspondence. Jose Lourenco
Ladeira, the park dendrologist, identified the shrub and tree
species that occurred within each trapping post. I am very
grateful for his participation and was amazed at his knowledge of
I thank Drs. Phil Myers, Mike Carleton, Guy Musser, Al
Gardner, and James Patton for identifying the voucher specimens.
Their prompt reply enabled me to begin the analysis shortly after
returning to the states.
I relied on several workers from the Rio Doce park to
implement the field project. Trails were opened in an expert
manner by Ivanil Moreira and Waldemar Queroga dos Santos. Ivanil
later became my field assistant and proved to be an extremely
valuable asset. He never once complained because of the incredible
numbers of ticks or because of all the small mammals that bit
chunks out of his fingers. I also thank Lidair Rufino for climbing
42 trees and placing the arboreal platforms.
Several students from the Departamento de Ecologia and
Zoologia participated in the fieldwork. I was aided greatly in the
data collection by 3 students. I thank Ludmilla Aguiar, Eduardo
Lima Sabato, and Luiz Paulo de Souza Pinto. I could not have
collected the appropriate data without their help. I also thank
Sonia Riquiera for her effort in helping us adjust to Minas Gerais,
and for participating in the initial stages of the fieldwork. She
was a pleasure to work with and always brought a positive attitude
to unpleasant working conditions.
My stay in Brazil was greatly facilitated by Gustavo and Ana
Fonseca. Their hospitality was more than generous and they always
opened their door to me and my family. Their hospitality was more
than "mineiro", it was more of true friendship. My wife and I will
always be grateful to all that the Fonsecas did to help us get
adjusted in Brazil.
To my wife and best friend, Cathie, I offer my deepest
gratitude for her patience, understanding, and moral support during
the 12 month field project. It was always a pleasure to come home
from the field and be met by a smiling and positive person. I am
very appreciative for Cathie's participation in the project and
wish that she could have had more freedom to pursue her own
TABLE OF CONTENTS
ACKNOWLEDGEMENTS ........................................... ii
I INTRODUCTION ......................................... 1
General Background ................................... 1
Study Organization .................................. 3
II SMALL MAMMAL INVENTORIES IN AN EASTERN BRAZILIAN PARK 5
Introduction ......................................... 5
Study Site ........................ ............ .... .... 7
Methods ............................................... 11
Results ........................................... 22
Discussion .............................. .............. 48
III TEMPORAL VARIATION IN TRAPPING SUCCESS OF
DIDELPHID MARSUPIALS IN AN EASTERN BRAZILIAN PARK ..... 57
Introduction ......................................... 57
Materials and Methods ............................... 58
Results ........................................... 66
Discussion .............................. .............. 81
IV SMALL MAMMAL ASSOCIATIONS AND MICROHABITAT
SELECTION IN AN EASTERN BRAZILIAN PARK ................ 88
Introduction .......................................... 88
Materials and Methods ............................... 89
Results ........................................ ....197
Discussion ............................................ 114
V FOREST FIRE AS A DETERMINANT OF SMALL MAMMAL
DIVERSITY IN A BRAZILIAN FOREST .......................122
Materials and Methods .................................126
Results ........................................ 138
Discussion .................................... 155
VI CONCLUSIONS AND SYNTHESIS ..................... 160
APPENDIX .............................................. 165
LITERATURE CITED .................................... 187
BIOGRAPHICAL SKETCH ................................... 200
Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
SMALL MAMMAL COMMUNITIES IN AN EASTERN BRAZILIAN PARK
JODY R. STALLINGS
Chairman: John G. Robinson
Major Department: Wildlife and Range Sciences
Forest Resources and Conservation
This dissertation examines small mammal communities in native
and non-native habitats in one of the largest remaining forested
tracts of land in the Brazilian Atlantic Forest. Small mammals
were live-trapped over a period of 12 months in grass, eucalypt,
and native forested habitats. Eucalypt and native forested
habitats were rich in rodent species, but marsupials were
numerically higher in terms of relative densities. Grass habitats
were rich in rodent species and were dominated by one rodent
species. Trapping success for marsupials was observed to increase
during the sharp dry season. Paucity of food resources may be the
factor responsible for the observed increase in trapping success.
The Rio Doce State Forestry Park has been subjected to frequent
crown fire in the recent past, and relatively little primary forest
remains. Marsupials may dominate in terms of abundance in this
Park because of the large amount of secondary forest. Small mammal
species diversity was calculated from forest stands that were
burned completely to the ground, in stands that were burned in a
mosaic fashion, and in primary stands. Species diversity was found
to be higher in intermediately fire disturbed native forested
habitats in comparison to heavily disturbed and primary forested
habitats. Mosaic habitats, composed of both secondary and primary
forest series, offer suitable habitat to more species than either
secondary or primary forest stands alone.
The Atlantic Forest of eastern Brazil is one of the most
threatened ecosystems in the World due to intense alteration of the
forested habitat (Mittermeler et al., 1982; Fonseca, 1983).
Fonseca (1983) estimates that approximately 3 5X of the native
forest remains today, with most of the remaining forest patches
occurring in protected areas and on privately owned lands. These
remaining forests are composed of mixed stands of primary, and
mostly, secondary forest series. The impact of these isolated and
heterogeneous forests on the wildlife communities has not been
studied. Only preliminary surveys have been conducted for the
primates that occur in this region and the remaining vertebrate
fauna is poorly known. This study focuses on small mammal
communities in a variety of habitats in one of the largest
remaining protected areas in the Atlantic Forest of eastern Brazil.
This study examines temporal and spatial variation in small mammal
communities in a region severely threatened by human activities.
The notion of prevailing stable environmental conditions in
the tropics helped to support the belief that tropical species
exhibit less temporal variation in reproduction, activity, and
feeding than temperate species. A current body of data reveals
that the tropics are much more dynamic than previously believed
(e.g., Leigh et al., 1982). This is especially true of tropical
areas that experience pronounced wet and dry seasons. The
Brazilian Atlantic region has such pronounced seasons. With this
in mind, this study investigated the effect of time on the trapping
success of small mammals.
Habitat structure has been shown to affect the species
composition and relative abundance of small mammals in forested
(Dueser and Shugart, 1978; August, 1984) as well as in non-
forested systems (August, 1984; Lacher et al., in press; Lacher and
Alho, 1988). Tropical forest small mammals form a diverse group
that has a varied diet, use of vertical space, and activity time.
In addition, South American small mammals can be composed of two
large groups, rodents and marsupials, in comparison to North
American communities. These factors will be taken into
consideration in determining how these small mammals allocate the
available space and resources in forest stands of differing
In some forest communities, disturbances can help to increase
floristic diversity (Connell, 1978; Horn, 1974). Connell (1978)
suggested that intermediate disturbances promote species diversity
in tropical forests and in coral reefs. This pattern has not been
observed for tropical vertebrate fauna, however, the relationship
is intuitive. Early secondary and primary forests are
characterized as spatially homogeneous, while forests disturbed in
an intermediate fashion are spatially heterogeneous. Forest series
characterized by a high degree of spatial heterogeneity should
provide more resources for more species in comparison to relatively
homogeneous environments. This study investigated small mammal
communities in a variety of forest habitats that were affected by
This dissertation is divided into four major chapters. Each
chapter reads as an independent paper, but all of the chapters have
a common theme. The theme is the small mammal communities in the
Rio Doce State Forestry Park in eastern Minas Gerais, Brazil. This
Park lies in the Rio Doce Valley and is part of the highly
endangered Atlantic Forest system of eastern Brazil. The primate
fauna of this region is highly endemic and very endangered
(Mittermeier et al., 1982). Inventories of other mammal species
are lacking for this region, although, preliminary species lists
indicated a highly endemic small mammal fauna.
Chapter II is concerned with intensive small mammal
Inventories in this park. This chapter describes the major habitat
types that were sampled, the sampling methodology, the effect of
trap position and type, and the small mammals that occurred in each
major habitat type. Chapter II sets the stage for the subsequent
Chapter III examines the temporal variation in trapping
success of marsupials. Many species of small mammals have a
"window of vulnerability" or a period of time during the year when
they are trapped with greater frequency. This chapter addresses
this question by examining the pattern of trapping success of
didelphid marsupials through time.
The fourth chapter addresses the effect of habitat structure
on the use of space by small mammals. I sampled small mammal
communities in 5 forested habitats. Individual species were
compared by their use of vertical space, diet, and time of
Chapter V examines the effect of forest fire in structuring
the small mammal communities in the Rio Doce Park. Tropical forest
fire has not been recognized as a determinant of species diversity.
In this chapter I suggest that even a low frequency fire can set
back succession in a forest system that may require 300 years to
reach "climax" conditions. Fires in humid forests can occur
because of unseasonably dry periods that produce optimal conditions
for intense crown fires. Gaps or patches produced by fires can
affect species diversity of small mammals. Disturbances of an
intermediate fashion, such as fire mosaics, should increase species
diversity because it produces a mixture of successional series that
are suitable for a number of small mammal species.
SMALL MAMMAL INVENTORIES IN AN EASTERN BRAZILIAN PARK
In comparison to its geographical size, few studies have been
conducted on small mammal communities in Brazil. Most published
reports on Brazilian mammals are preliminary species lists (e.g.
Avila-Pires and Gouvea, 1977) or inventories (e.g., Vieira, 1955;
Moojen, 1952). Some Brazilian studies have focused on densities
(Emmons, 1984), others have focused on abiotic effects on small
mammals (Borchert and Hansen, 1983; Peterson, in press), while
others have addressed such diverse subjects as plantation effects
(Dietz et al., 1975) and public health needs or mammals that carry
human diseases (e.g., Laemment et al., 1946; Botelho and Linardi,
1980; Dias, 1982).
Recently, work in Brazil using mark-release techniques has
addressed the use of space, longevity, diversity and social habits
of small mammals (see Alho, 1982). As Alho (1982) pointed out,
most of these studies were concentrated in the xerophitic Cerrado
and Caatinga habitats. Fewer studies have been carried out in the
humid forests. Carvalho (1965) live trapped small mammals in a
tropical humid forest in Sao Paulo. Malcolm (pers. comm.) trapped
small mammals in tropical humid forest near Manaus, Amazonia.
Fonseca (pers. comm.) worked on small mammals in a range of
habitat types in eastern Minas Gerals.
The purpose of this paper is to report on intensive small
mammal inventories in a tropical humid Brazilian forest, the Rio
Doce State Forestry Park. This Park occurs within the geographical
limits of the highly endangered Brazilian Atlantic Forest (Fonseca,
1983). Small mammal inventories are lacking from this park.
Gastal (1982) and Avila-Pires (1978) report preliminary results
from intermittent small mammal inventories in the Rio Doce park.
The Brazilian Atlantic forest has a highly diverse flora and fauna,
with many endemic species of trees (Mori et al., 1981), reptiles
(Muller, 1973), and birds (Haffer, 1974). In-depth mammal
inventories in the region are lacking. The mammalian fauna is
poorly known. Mitterme1er et al. (1982) and Klnzey (1982) report
on the high level of diversity and endemism found in the primates
of the region. Preliminary species lists for non-volant mammals in
this region suggest a very high diversity and endemism (Cabrera,
1957, 1961; Honacki et al., 1982; Moojen, 1952; and Vieira, 1955).
For this project, a preliminary checklist was prepared of the non-
volant mammal species that probably occur in the Atlantic Forest
region (Table A-1). These data indicate that for the region there
are at least 131 species, 50 of which (39X) are endemic.
Small mammal trapping was carried out in the Rio Doce State
Forestry Park (19 48'18" and 19 29'24" south latitude and 42
38'30" and 42 28'18" west longitude). The Park was created in 1944
at the request of Dom Helvecio, the bishop of the region (Gilhaus,
1986). The State Forestry Institute of the state of Minas Gerais
is the present administrative body.
The climate of the Park is classified as tropical humid
(Gllhaus, 1986) with a seasonal pulse of precipitation from
November through February and a pronounced dry season from June
through August. Average annual rainfall for a 20 year period
(CETEC, 1981) was 1480 mm, although the rainfall recorded during
this year of the study was considerably less (Figure 2-1). Mean
annual temperature averages 22 C (CETEC, 1981) and mean minimum
monthly temperatures vary greatly throughout the year (Figure 2-2).
The Park boundaries on the north and the east are two rivers,
the Rio Piracicaba and the Rio Doce, respectively (Figure 2-3).
The southern and western boundaries abut plantations of Eucalyptus
The predominant relief forms in the Park are hills originating
from fluvial plane dissection and valleys derived from fluvial
deposits (Gilhaus, 1986). The altitude in the Park varies from 230
to 515 m. CETEC (1981) reports that 21% of the Park is composed of
plains, 40% undulating to strongly undulating hills and 34%
strongly undulating hills to mountainous terrain.
WALTER AND LEBTH
MONTHLY RAINFALL (MM) MEAN TEMPERATURE ( )
JAN FEB MAR APR MAY JUN JUL AUG SEP OOT NOV oE8
AVERAGE ANNUAL RAINFALL 1480 mM
OATA PRU *964-'474
:ALTER AND LEITH
MONTHLY RAINFALL fMM.) MEAN TEMPERATURE O )
1 60 4"
AN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC
ANNUAL AVERAGE RAINFALL 804 mne
ATA PROM I086-1Oe
Figure 2-1. Walter and Leith climatic diagram characterizing
precipitation surplus and drought per month in Rio Doce State
Forestry Park, Minas Gerais, Brazil. A= data collected for a
twenty year period (1954-1974) and B= data collected during
MEAN MINIMUM & MAXIMUM
NOV DEC JAN FEB MAR APR MAY JUN JUL AUG SEP O(
MEAN MAXIMUM TEMP
MEAN MINIMUM TEMP
DATA FROM 1954 1974
Figure 2-2. Temperature graph showing pronounced decrease in
minimum temperature during June, July, and August in Rio Doce State
Forestry Park, Minas Gerais, Brazil.
RIO DOCE STATE
..::i... boundary ...
:::::::::::..... burned areas
"": : -: eucalyptus
0 1 2 3 4 km
Figure 2-3. Map of Rio Doce State Forestry Park, Minas Gerais,
Brazil. Numbers circled indicate trapping sites.
A unique feature of the Rio Doce State Forestry Park (PFERD)
is the system of Rio Doce Valley lakes that occur there. There are
approximately 40 lakes and numerous marshes in the Park (Saijo and
Tundisi, 1985). According to Saijo and Tundisi (1985) the lakes
were formed by damming the drainage river of the Rio Doce
watershed. The marshes in the Park are the result of the
sedimentation of previous lakes.
The vegetation of the park is classified as tropical semi-
deciduous (Gilhaus, 1986). Most of the emergent tree species lose
their leaves during the cool dry months (J. Stallings, per. obs.).
Forest fire has been a major threat to the vegetation and wildlife
in the park because of the litter that accumulates during the dry
season. In 1964 and 1967, major fires burned approximately 30Z of
the park (Lopes, 1982; Silva-Neto, 1984).
Gilhaus (1986) described 5 forested and 5 open/field habitats
for the Park (Table 2-1). All of the forested and open/field habitats
have been altered to some extent by fire, with the exception of the
Tall Primary Forest with Epiphytes.
I live-trapped small mammals in 4 general habitas types: (1)
homogeneous eucalypt forest, (2) eucalypt forest with native species
subcanopy, (3) wet meadow, and (4) native forested habitats. I
trapped in 5 sites within the native forested habitat type. Two of
the sites (RD/C and RD/T) were primary forests and corresponded to
the Gilhaus (1986) classification of Tall Primary Forest with
Epiphytes. Two other sites (RD/H and RD/M) were altered by forest
fire in 1967 and burned in an intermediate fashion which produced a
forest mosaic of short, secondary and tall forest. This habitat type
corresponded to the Medium to Tall Forest with Bamboos and
Graminoids (Gilhaus, 1986). The remaining site (RD/F) was burned
completely to the ground in 1967 and the resulting vegetation type
corresponded to the Medium Secondary Forest with Bamboos and
Graminoids (Gilhaus, 1986).
The wet meadow habitat (RD/B) corresponded to both the Low
Woodland and Low Tree and Scrub Tallgrass Savanna classified by
Gilhaus (1986). This habitat type occurs between the edge of
permanent marshes and secondary forest. Grasses from the family of
Graminae are the dominant vegetative cover and were introduced in
the region as food for cattle.
The eucalypt forest with native subcanopy habitat was planted
with Eucalyptus salina in 1954 after the original vegetative cover
was removed. The eucalypt forest was harvested selectively in 1964
and again in 1971. However, the eucalypt forest was never clear-cut
and the native species were allowed to regenerate, largely through
coppicing, and developed into a complex native subcanopy. The result
is a homogeneous eucalypt upper canopy and a native species
subcanopy or "mata su.a." Tall exotic grasses cover the ground.
Emergent eucalypt trees reach 20 m in height.
The homogeneous eucalypt forest was first planted with
Eucalyptus salina in 1954 and was harvested 3 times on a 7 year
rotation. There is no native woody herbaceous subcanopy, only
Table 2-1. Forested and open/field habitats that occur in the Rio
Doce State Forestry Park, Minas Gerais, Brazil. Definitions and
characteristics of habitats taken from Gilhaus (1986).
HABITAT TYPE % TOTAL THIS STUDY
TALL PRIMARY FOREST
MEDIUM TO TALL FOREST WITH
BAMBOOS AND GRAMINOIDS
MEDIUM SECONDARY FOREST WITH
BAMBOOS AND GRAMINOIDS
LOW SECONDARY FOREST
LOW TREE AND SCRUB TALLGRASS
EVERGREEN TALLGRASS FIELD
WITH TYDha SD.
eucalypt trees. Tall exotic grasses are the dominant ground cover.
Eucalypt trees reach approximately 15 m In height.
Individuals were snap trapped in order to obtain voucher
specimens, dietary, and reproductive information. Snap trapping was
carried out exclusively in wet meadow and secondary succession
habitats. All specimens were either preserved in 10X formalin or
made into museum study skins with corresponding complete skulls. I
also made study skins of individuals of species of uncertain taxonomic
status that were live trapped in habitats other than those where snap
trapping was carried out.
I placed one line of terrestrial Sherman live traps in the
homogeneous eucalypt habitat. Twenty-five Shermans, baited with dry
oatmeal, pineapple chunks and cotton balls soaked with cod-liver oil,
were opened for 4 consecutive nights on 5 occasions through the
study. Trap stations were separated by 15 m.
In the wet meadow habitat, live trapping started in February,
1986, and continued at monthly intervals through January, 1987. I
used two parallel trapping lines in this habitat. Each line was 280 m
in length and subdivided into 20 trap stations separated by 15 m. All
traps were placed on the ground, with even numbered stations having
only one Sherman live trap while odd numbered stations had one
Sherman and one locally made small wire live trap. Bait was
identical to that used in the homogeneous eucalypt forest.
In the eucalypt forest with native subcanopy and the native
forested habitats, the trapping design was identical (Table 2-2). In
each area, I cut 3 parallel lines 300 m in length through the forest.
Table 2-2. Trapping design used in native forested and eucalypt
forest with native species subcanopy habitats. Trapping lines are
A, B, and C. Numbers 1 16 represent trapping posts. Terrestrial
medium sized live traps = arboreal medium sized live traps = *,
small arboreal Sherman live traps = S, small terrestrial Sherman
live trap = s, large terrestrial live trap = L. Trapping lines
were separated by 100 m.
STATION TRAPPING LINES
A B C
Sixteen trapping stations were placed along the line separated by 20
m. I used Sherman live traps and locally made small (15 X 15 X 30
cm) and large (25 X 30 X 60 cm) wire live traps. All traps were
placed within 3.5 m of the trapping post. All trapping posts had a
terrestrial small live trap. Odd numbered trapping posts had a small
wire live trap placed in a tree or bush. Mean arboreal live trap
height was 1.2 m. Odd numbered trapping posts had a Sherman live
trap alternating between arboreal and terrestrial positions. The
exterior trapping lines were identical with respect to number, kind
and placement of traps. I did not use the large live traps on the
interior line. Aside from the large live traps, the exterior and
interior lines were equal in trap number. However, the positions of
the Sherman and small live wire traps were reversed for the interior
line. The Sherman live traps were introduced into each of the
forested habitats after the study was well underway in an attempt to
sample smaller bodied species. I placed Shermans in two of the
native forested sites in January, 1986, and introduced Shermans in the
remaining forested sites in May of the same year.
I also experimented with traps that were placed high in the
canopy by a pulley and platform device. This method was similar to
that developed by Malcolm (pers. comm.) for use in the Brazilian
Amazon. Mean trap height was 11.2 m. I used these arboreal traps
to sample the canopy dwelling small mammals. Traps were located at
trapping posts along the established lines. The post and exact trap
location in each tree were determined in a subjective manner. I
placed traps in trees which I thought had a high degree of canopy
connectivity and upper stratum vine density. Forty-two arboreal
platforms were spread across 4 native forested sites. The primary
forest sites, RD/C and RD/T, and one of the mosaic sites, RD/H,
each had 12 arboreal traps, 4 traps per line. The other mosaic site,
RD/M, had only 6 traps because I did not believe that there was
sufficient upper strata development to support canopy dwelling
species. Arboreal canopy trapping started in June, 1986, and
continued through October, 1986. Trapping coincided with the
schedule developed for the terrestrial and arboreal trapping.
I used dry oatmeal, pineapple chunks and cotton balls soaked
with cod-liver oil for bait. Traps were set during the day and
remained open for 5 consecutive nights each month for one calendar
The first occasion that an individual was trapped was considered
a first capture. The first capture plus subsequent captures of each
individual were considered total captures. Minimum known alive
(MKA) was the number of individuals actually captured each month
whether the capture was the first capture or a recapture from a
Trapping success of small mammals was calculated in the
following manner. I multiplied the number of traps by the number of
nights the traps were baited and armed per site per month to
determine the number of trap nights. Trapping success was the
number of individuals, MKA or total captures of all species divided by
the number of trap nights and expressed in percentages. For
example, if 100 individuals were trapped during 1000 trap nights, the
trapping success would be (100/1000) X 100 = 10X. Recapture indices
were calculated by dividing total captures by first captures and
indicated the average number of times an individual of each species
I recorded the following information from each captured small
mammal: date, location on trapping line, position of trap, species, sex,
whether the trapped animal was a juvenile or adult, it's reproductive
condition, general condition, external parasitic load on a relative
scale, and behavior upon release. I also recorded standard body
measurements for each individual: body length, tail, ear, hind foot
and mass. I placed a numbered metal eartag in the left ear of each
individual upon the initial trapping of the individual.
Taxonomic determination of questionable species was made by
taxonomists specializing in various small mammal groups. Voucher
specimens were brought to the United States and the taxa were
distributed to the following people: cricetid rodents and marsupials of
the genus Marmosa were sent to Dr. Phil Myers, University of
Michigan; rodents of the subgenus Oecomvs were sent to Dr. Guy
Musser, American Museum of Natural History, and to Dr. Mike
Carleton, U.S. National Museum of Natural History.
I used body measurements, weight, reproductive condition and
pelage characteristics to determine the age class (juvenile or adult)
of each captured individual. An individual was considered an adult if
it was reproductively active. Female rodents were considered
reproductively active if they 1) had a perforated vulva, 2) were
pregnant, or 3) were lactating. Marsupial females were considered
reproductively active if they 1) were lactating or 2) had young
attached to the teat field. Male rodents were considered
reproductively active if the testes were descended. I could not
determine the reproductive status of male marsupials as the testes are
permanently descended. However, the activity state of the sternal
gland in marsupials can indicate the reproductive time of year.
Initially, I used the overall gestalt of each individual of each sex to
assign age classes. Later, I compared my initial classification with
the body measurements and mass. Body measurements and weights
were sorted for each sex of each species and plotted to determine if
recorded values could be grouped according to size. I then assigned
a body measurement value as the threshold for separating juvenile
and adult age classes. These age classes were then compared to the
initial age classes that were assigned in the field.
I used the General Linear Program (PC-SAS) ANOVA to test for
the equivalence of adult body measurements and mass means between
sexes for each species. This analysis enabled me to determine the
extent of sexual dimorphism for the external characters. Statistical
significance was set at 1 0.05.
Feeding categories were determined by stomach content analysis
(Charles-Dominique et al., 1981) and from the literature. I relied
heavily on information gleaned from the literature on food
preferences of small Neotropical mammals.
The use of vertical space (i.e., terrestrial, scansorial, or
arboreal) of each species was determined from 3 data sets. I
compared the proportion of captures in trees to that of captures on
the ground for each species. There were more terrestrial than
arboreal traps and this bias was corrected by adjusting the number of
total trapping opportunities. For this adjustment, I divided the
number of arboreal total captures by the total arboreal opportunities
(or total arboreal trap nights). The same was done for terrestrial
total captures. Results of these two divisions were summed and each
respective result (e.g., arboreal) was divided into the sum. The result
generated adjusted percentage success of arboreal and terrestrial
captures. The sum of total captures was multiplied by the adjusted
percentage in order to generate adjusted number of captures per
trapping stratum (on the ground or in arborescent vegetation). These
adjustments also were made for each species at each site as well as
pooled adjustments across all sites. I tested the null hypothesis that
there was no difference in the proportion of arboreal and terrestrial
captures for each species across all habitats as well as within each
trapping site. Prior to using Student's t-test, I arc-sin transformed
the proportions. I then compared the proportion of arboreal and
terrestrial responses upon release for each species across all habitat
types. These adjustments were made for all species in each habitat
I used Student's t-tests to test the null hypothesis that there
was no difference in the locomotory response upon release of each
species. These 2 data sets were then compared to determine if there
were any differences between where an individual was captured and
its locomotory behavior upon release across all habitat types. These
tests also were performed for each habitat sampled. A species was
considered to be arboreal if that species was found to have a high
proportion of arboreal captures and a high proportion of arboreal
behavior upon release. The opposite would be true for a terrestrial
species. A species would be scansorial if there were no significant
differences in the proportion of spatial captures and no significant
differences in the proportion of behaviors exhibited upon release. I
then compared my results obtained from the trapping data to the
available literature for each species.
I also recorded the presence and relative abundance of other
vertebrates while walking the trapping lines in each sampling area.
Upon encountering an animal, I recorded the species, distance in
meters from the trail, height in trees for arboreal species, hour of
day, group size and other natural history data. I moved at a pace of
approximately 1 km. per hour and covered a distance of 1050 m per
sampling area. These walks yielded 60 km of repeat censuses per
sampling area. This method does not allow for the computation of H'
indices because it is difficult to ascertain the identity of individuals
being censused. Because of this problem, I only calculated species
richness per taxonomic group, number of encounters of each species
through time, and number of encounters of each group per linear km.
These data are presented in Table A-3.
I also superficially sampled the bat fauna of the Park. I
mounted mist nets in various habitat types (e.g., lake edge, secondary
habitat, primary habitats, and manmade structures) in order to
determine the species that occurred in the park. Nets were mounted
one hour before dusk and opened for approximately 4 hours each
night. Mist netting occurred on a sporadic basis. A species list of
bats obtained by these methods is presented in Table A-4.
I logged 1,308 captures of 17 species of small mammals in 40,490
trap nights. The small mammal fauna of the park consisted of 6
species of marsupials and 11 species of rodents. The diet, use of
vertical space and habitat requirements are presented in Table 2-3 for
each species captured during the study.
Marsupials Family Didelphidae.
Didelphis marsupialis Linne. (1758). The black-eared opossum
ranges widely in South America from the Isthmus of Panama to
southern Brazil. This species occurs sympatrically with D. albiventris
throughout much of its range (Streilein, 1982). In the Rio Doce
Valley, however, D. marsupials inhabits moister habitats, while D.
albiventris occurs in the cerrado vegetation (Valle and Varejao, 1981).
A. Gardner (pers. comm.) suggested that the form of D. marsupialis
found in eastern Brazil is distinct and should be referred to as Q.
azarae. This species inhabits brushy and forested habitats (Alho,
1982; Nowak and Paradiso, 1983). Miles et al. (1981) found this
species to be nocturnal, with a preference for nesting in tree
cavities. I captured this species in all forested habitats in the park
(Table 2-4). Adult body measurements do not indicate sexual
dimorphism (Table A-2). Females have a well developed pouch. This
Table 2-3. Ecological role played by each species captured during
this study in the Rio Doce State Forestry Park. GM= grasslands and
wet meadows, B= brushy areas, S= secondary forests, P= primary
forests, F= fossorial or semifossorial, SA= semiaquatic, T=
terrestrial, S= scansorial, A= arboreal, HG= herbivore-grazer; FG=
frugivore-granivore; FO= frugivore-omnivore; IO= insectivore-
omnivore. Taxa endemic to the Brazilian Atlantic rainforest or
the eastern coastal area of South America.
SPECIES HABITAT ADAPTATION CLASSIFICATION
Didelphis marsupialis B, S, P T, S FO
Metachirus nudicaudatus S, P T IO/FO
Marmosa incana, B, S, P S IO
M. cinerea B, S, P A 10
M. microtarsus* S, P A IO
Caluromys philander S, P A FO
Oecomvs trinitatis S, P S FG
Orvzomys capito S, P T FG
Q. subflavus* GM, B, S T FG
0. nigripes* GM, B S FG
Akodon cursor* GM, B, S T IO
Calomvs laucha GM, B T FG
Nectomys sauamipes GM, B SA HG
Abrawavaomvs ruschii* S T FG?
Oxvmycterus roberti* GM, B, S F 10
Rhipidomvs mastacalis S, P A FG
Cavia fulgida* GM, B T HG
Euryzygomatomys soinosus* GM, B F HG
Table 2-4. Number of total captures and percent of total for each
species (SPP.) per habitat type. Numbers in parentheses represent
the percentage of captures per species per habitat percentages are
rounded to the nearest whole number.
SPP* HABITAT TYPES+
RD/F RD/H RD/M RD/T RD/C RD/E RD/B
DM 4 (3) 2 (1) 1 (1) 7 (6) 21(24) 7 (4)
MN 20(14) 44(16) 21(15) 30(23) 25(28) 18(11) -
MI 25(18) 75(27) 20(14) 28(22) 6 (7) 14 (9) 3 (1)
MC 74(52) 71(26) 69(49) 38(30) 31(35) 77(49) 1 (0)
MM 1 (0) -
CP 16 (6) 2 (1) 13(10) 3 (3) 15(10) 2 (1)
NS 6 (2) 9 (6) 5 (1)
RM 7 (6) -
AC 3(2) 46(17) 2(1) 1 (1) 25(16) 315(85)
OT 4 (3) 8 (3) 4 (8) 3 (2) 2 (2) -
OC 4 (1) 12 (9) 2 (2) 2 (1) 1 (0)
OS 12 (9) 1 (0) 31 (8)
ON 4 (1)
OR 1 (0) 2 (1) 4 (1)
AR 1 (0) -
CL 4 (1)
ES -2 (1)
142 276 142 128 89 158 373
+ RD/F= SECONDARY HABITAT BURNED COMPLETELY TO THE GROUND; RD/H=
SECONDARY HABITAT BURNED IN MOSAIC FASHION; RD/M= SECONDARY HABITAT
BURNED IN MOSAIC FASHION; RD/T= PRIMARY FOREST WITH LITTLE EFFECT
FROM FOREST FIRE; RD/C= PRIMARY FOREST; RD/E= EUCALYPT FOREST WITH
NATIVE SPECIES SUBCANOPY; RD/B= WET MEADOW.
* DM= DidelDhis marsupialis; MN= Metachirus nudicaudatus; MI=
Marmosa incana; MC= Marmosa cinerea: MM= Marmosa microtarsus; CP=
Caluromvs philander; NS= Nectomvs sauamipes; RM= Rhipidomys
mastacalis; AC= Akodon cursor; OT= Oecomvs trinitatis; OC=
Orvzomvs capitol; OS= Orvzomvs subflavus; ON= Orvzomvs nigripes;
OR= Oxvmvcterus roberti; AR= Abrawayaomvs ruschii; CL= Calomys
laucha; ES= Eurvzvyomatomvs spinosus.
species is terrestrial. There was a significant percentage of
terrestrial captures (Table 2-5) and terrestrial behavior upon release
(Table 2-6). I did observe several juvenile individuals and one adult
climb readily. Charles-Dominique (1983) reported that this species
exploits the lower stratum in forests, but can climb. This species is
basically an opportunistic feeder and feeds upon fruit and animal
matter (Charles-Dominique, 1983).
Metachirus nudicaudatus Geoffroy (1803). The brown four-eyed
opposum has a geographical distribution similar to that of 1.
marsupialis except that it is not found over much of Venezuela nor
in northeastern Brazil (Streilein, 1982). M. nudicaudatus can be
confused with Philander ODDosum as both have pale spots above the
eyes. In addition, there is considerable confusion over the taxonomy
of the two species. Nowak and Paradiso (1983) classified this species
as Philander nudicaudatus and Philander opossum as Metachirops
ODOSsum. I agree with Honacki et al. (1982) and follow their
classification. This species was captured in all forested habitats
(Table 2-4). Metachirus nudicaudatus is sexually dimorphic in its
mass and hind foot measurements (Table A-2). Females do not have
a pouch. This species is strongly terrestrial, rarely caught in
arboreal traps (Table 2-5) and rarely climbs upon release (Table 2-6).
Miles et al. (1981) found this species to be nocturnal and construct
nests on the forest floor or in ground hollows. There are very little
data on the feeding habits of this species due to the small numbers
that have been reported to be trapped. Preliminary data indicate
Table 2-5. Student's t-tests between adjusted and arcsin
transformed percentages of terrestrial and arboreal captures of
small mammals in all forest types. NS = non significant.
Table 2-6. Results of Student's t-tests between terrestrial and
arboreal behavior upon release of small mammals captured in all
NS= non significant.
X TERRESTRIAL % ARBOREAL
SPECIES N BEHAVIOR N BEHAVIOR V T P <
DM 33 73.2 3 16.7 34 3.265 0.01
MN 150 80.7 4 9.1 152 4.934 0.001
MI 95 53.5 52 36.5 145 3.448 0.001
MC 11 10.8 294 78.9 303 7.743 0.001
CP 2 18.4 18 71.6 18 2.488 0.05
NS 12 90.0 0 0.0 10 10.882 0.001
RM 1 24.0 5 67.8 4 1.393 NS
AC 70 90.0 0 0.0 68 26.282 0.001
OT 9 64.8 2 25.3 9 1.763 NS
OC 13 90.0 0 0.0 11 11.326 0.001
OS 10 90.0 0 0.0 8 9.933 0.001
that this species is an insectivore-omnivore (Robinson and Redford,
1986) or frugivore-omnivore (Hunsaker, 1977).
Marmosa incana Lund 1840. This mouse opposum is endemic to the
Brazilian Atlantic Rainforest (Streilein, 1982). M. incana occurs in
both secondary and primary forest habitat. This species is small in
size (adults, average weight = 62 g) and strongly sexually dimorphic
in body size and color. Males tend to have larger ears and hind feet
(Table A-2) while females tend to have a more rose colored venter
and less pronounced face mask (P. Myers, pers. litt.). Females do not
have a true pouch. This species tends to use both the ground and
arborescent vegetation. I classify the spatial adaptation of this
species as scansorial (Table 2-3). There was no significant difference
in the proportion of terrestrial and arboreal captures (Table 2-5, p >
.90, df=169), however; individuals tended to remain on the ground
upon release (Table 2-6, p < 0.001, df=145). No data exist on the
feeding category of this species. Other species of Marmosa which
have similar body mass are classified as insectivore-omnivores.
Stomach content analysis (n=3) showed 100% insects from two orders,
Coleoptera and Orthoptera (Table 2-7). I classify this species as an
insectivore-omnivore based on the relationship found between body
mass and dietary classification (Robinson and Redford, 1986).
Table 2-7. Stomach content analysis of small mammals captured in
Rio Doce State Forestry Park, Minas Gerais, Brazil (N= number of
SPECIES FRUIT SEEDS GRASS INSECTS
Marmosa incana (n=3) 100%
Akodon cursor (n=23) 19% 19% 8% 52%
Orvzomys subflavus (n=1) 5% 95% -
Orvzomys nigripes (n=5) 34% 11% 21% 34%
Oxvmvcterus roberti (n=2) 100%
Calomys laucha (n=1) 100% -
Nectomvs squamipes (n=2) 50% 50 -
Marmosa cinerea Temminck (1824). M. cinerea has a disjunct
geographical distribution in South America; it occurs in northern
Venezuela through the Guianas, and It occurs In the Brazilian
Atlantic Rainforest extending into Paraguay (Streillen, 1982). M.
cinerea occurs in brushy and forested habitat, ranging from secondary
to primary. This species is a large bodied Marmosa (Table A-2,
average weight= 105g) and is highly sexually dimorphic based on
external body measurements. Males tended to have larger body, tail,
ear and foot (Table A-2). Females do not have a pouch. This
species is strongly arboreal and exploits the high forest stratum
(Charles-Dominique, 1983; Miles et al., 1981). Miles et al. (1981)
found this species to be nocturnal and to construct open arboreal
nests rather than use cavities. M. cnerea tended to be caught a
greater proportion of the time in arboreal traps (Table 2-5, p <0.001,
df=356) and tended to exhibit arboreal rather than terrestrial behavior
upon release (Table 2-6, p < 0.001, df=303). This species is primarily
an insectivore-omnivore (Robinson and Redford, 1986).
Marmosa microtarsus Wagner (1842). This species is restricted to the
Brazilian Atlantic Rainforest (Streilein, 1982). M. microtarsus differs
from its congener M. agilis by possessing a pure colored white patch
of hairs on the throat and chin (Tate, 1933). There are insufficient
data in the literature to determine if this species is terrestrial,
scansorial, or arboreal. I only recorded one capture during the study,
in an intermediately disturbed habitat (Table 2-4). However, the
species has a long prehensile tail and short wide feet which suggest
an arboreal lifestyle. This species is probably an insectivore-
Caluromys philander Linne (1758). This species has a disjunct
geographical distribution with populations in Venezuela, the Guianas
and northern Brazil and in southeastern Brazil (Streilein, 1982). The
woolly opossum is classified as a forest dwelling species (Nowak and
Paradise, 1983). This species was present in the eucalypt, mosaic,
and primary forested habitats (Table 2-4). Based upon the external
body measurements, there was no sexual dimorphism in adults (Table
A-2). Females lack a true pouch. According to Charles-Dominique
(1983) and Miles et al. (1981) this species exploits the high forest
stratum and is nocturnal. My data showed that there was a higher
proportion of arboreal captures (Table 2-5, p< 0.001, df=49) and that
this species tended to climb more than remain on the ground upon
release (Table 2-6, p < 0.05, df=18). Fruit makes up a large portion
of this species' diet (Charles-Dominique, 1983; Robinson and Redford,
Rodents Family Cricetidae
Oecomys (Oryzomys) trinitatus = (Q. concolor) Wagner (1845). This
genus is in need of revision and the subgenus Oecomys is currently
being revised (P. Myers, pers. comm.). This species was previously
called Orvzomvs concolor and was known, within Brazil, as an
Amazonian species (Alho, 1982). However, Nitikman and Mares (1987)
reported trapping this species in gallery forest in the Brazilian
cerrado. I captured this species in all native forested habitats (Table
2-4). The species is not sexually dimorphic (Table A-2).
Gyldenstolpe (1932) and Moojen (1952) stated that this species is
"more or less adapted for arboreal life." My data suggest that this
rat is scansorial; there were no significant differences in the
proportion of terrestrial and arboreal captures (Table 2-5, p < .10,
df=19) and no significant differences in terrestrial and arboreal
behavior upon release (Table 2-6, p < .20, df= 9). Most species of
the genus Orvzomvs are frugivore-granivores (Robinson and Redford,
Orvzomvs subflavus Wagner (1842). This species is distributed
throughout the Gulanas, southeastern Brazil and eastern Paraguay
(Honacki et al., 1982; Alho, 1982). In Brazil, it occurs in the
cerrado, caatinga and Atlantic Rainforest (Alho, 1982). This species
was captured in wet meadow and heavily disturbed secondary habitat
(Table 2-4). There were no differences in body measurements
between sexes in adult individuals (Table A-2). I classify this species
as terrestrial. This species tended to be captured more on the
ground than in the trees (Table 2-5, p < 0.01, df=46) and was never
observed to climb upon release (Table 2-6). Stomach content analysis
(n=1) showed 95% grass and 5% fruit (Table 2-7).
Oryzomys capitol Olfers (1818) = Q. goeldi, laticeos and intermedius.
This species has a wide distribution throughout the Neotropics and
occurs in a variety of habitats ranging from agricultural fields
(Moojen, 1952) to humid forests (Alho, 1982). Orvzomvs capitol was
primarily captured in humid forests ranging from intermediate levels
of disturbance to primary forests in the park (Table 2-4). There
were no significant differences in body measurements between sexes
for adults (Table A-2). My capture and release data are in
accordance with Alho's (1982) classification for this species. Q.
capito tended to be caught more on the ground than in trees (Table
2-5, p < 0.05, df=19) and tended to remain on the ground upon
release (Table 2-6, p < 0.001, df=11).
Orvzomvs (Oliiorvzomvs) niarioes (eliurus) Wagner (1845). This
small-bodied rodent (Table A-2) occurs in grassland, wet meadow and
secondary forest habitat in northern Argentina, eastern Paraguay,
southern Brazil and the Bolivian Beni (Honacki et al., 1982). In the
park, all captures were made in the wet meadow habitat (Table 2-4).
All captures were made on the ground (n=4), however; the individuals
climbed readily in captivity (J. Stallings, pers. obs.). The results
from the stomach analysis (n=5) revealed a wide range of foodstuffs
Abrawavaomys ruschii Cunha and Cruz (1979). This species is only
known from the type locality in Espirito Santo, eastern Brazil. It is
endemic to the Brazilian Atlantic Rainforest. The single capture of
this species was recorded in an intermediately disturbed forest (Table
2-4). There is very little information available regarding the ecology
of this species and there are only three study skins found in
museums (A. Gardner, pers. comm.).
RhiDidomvs mastacalls Lund (1840). Climbing mice range south from
Margarita and Tobago Islands to Venezuela and Gulanas to
northeastern and east central Brazil (Honacki et al., 1982). This
species was only captured in a relatively undisturbed primary forest
(Table 2-4). Sample size was too small to detect any differences
between terrestrial and arboreal captures and behavior upon release.
However, as the common name implies, this species climbs readily. I
captured two individuals in my house in the park, a commonly cited
"exotic" habitat for this species (Nowak and Paradiso, 1983).
Nectomvs sauamipes Brants (1827). The neotropical water rat occurs
in aquatic habitats either in grasslands and wet meadows or in
forests. This species distribution ranges from the Gulanas to
Colombia to Peru and in Brazil, Paraguay and northeastern Argentina
(Honacki et al., 1982). This species tended to be caught more on the
ground (Table 2-5, p < 0.05, df=17) and exhibited a significant
tendency to remain on the ground upon release (Table 2-6, p < 0.001,
df=10). Stomach content analysis (n=2) showed 50% grass and stems
and 50X fruit (Table 2-7).
Akodon cursor Winge (1887). This species occurs in several habitat
types from southeastern and central Brazil to Uruguay, Paraguay and
northern Argentina. The wet meadow habitat in the park was the
primary habitat to capture this species (Table 2-4). A. cursor was
formerly included in A. arviculoides (Honackiet al., 1982). This
species is sexually dimorphic in tail (p < 0.01) and body (p < 0.008)
length (Table A-2). A. cursor is strongly terrestrial (Table 2-5 and
Table 2-6). Analysis of stomach contents (n=23) revealed a high
proportion of insects, seeds and fruit (Table 2-7).
Calomys laucha Olfers (1818).. This species occurs in grassland and
wet meadows in southern Bolivia, southeastern Brazil, Paraguay,
central Argentina and Uruguay. I only captured this species in the
wet meadow habitat in the Park (Table 2-4). The results from one
stomach sample revealed 100% seeds (Table 2-7).
Oxvmvcterus roberti Thomas (1901). The burrowing mouse occurs in
a variety of habitats but is usually associated with moist substrate in
open or brushy habitats. I captured this species in wet meadow and
secondary habitats in the Park (Table 2-4). This species is endemic
to eastern Brazil. This semifossorial mouse is described as an
insectivore (Nowak and Paradisio, 1983). Stomach analysis (n=2)
revealed 100% insects (Table 2-7).
Cavia fulaida Wagler (1831). This species of cavy is endemic to the
open grasslands and wet meadows of the Atlantic Rainforest of
eastern Brazil (Honacki et al., 1982; Nowak and Paradiso, 1983). I
captured this species in grassland and wet meadow habitats in the
Park, however; it was not trapped in site RD/B. For this reason this
species does not appear in Table 2-4. Cavies are terrestrial and are
herbivore-grazers (Nowak and Paradiso, 1983).
Eurvzvqomatomvs soinosus Fischer (1814). The single species of
guiara is endemic to southeastern Brazil, Parguay and northeastern
Argentina (Honacki et al., 1982). I captured this species in the wet
meadow habitat (Table 2-4). This species inhabits open grasslands
and wet meadows, is considered terrestrial or semifossorial (Alho,
1982), and is most probably a herbivore-grazer.
Tables 2-8, A-5, and A-6 present the capture results by species
for the three main habitat types: native forested, eucalypt with
native species subcanopy and wet meadow habitats, respectively. As
a group, marsupials represented 79.2% and 83.3X of first and total
captures, respectively, in native forested sites, and 67.7% and 82.9% in
eucalypt forest with native species subcanopy. Rodents represented
97.3% and 98.4% of the first and total captures, respectively, in the
wet meadow habitat.
In the native forested habitat, Marmosa cinerea represented over
40% of the marsupial captures (Table 2-8), while in the eucalypt
forest, this species represented more than 58% of the marsupial
captures (Table A-5). Akodon cursor was the major contributor to
the rodent captures in all three habitats. This species only
Table 2-8. Capture results from native forested plots in Rio Doce
State Forestry Park, Minas Gerais, Brazil. RECAP INDEX= total
captures/first captures, and represents the average number of times
that an individual of species X is captured. Numbers in
parentheses represents percent of contribution of capture per
species per taxonomic group. Species abbreviations are explained
in Table 2-4.
TOTAL FIRST RECAP
SPECIES CAPTURES % TOTAL CAPTURES % TOTAL INDEX
DM 35 4.5 ( 5.4) 32 7.8 ( 9.9) 1.1
MN 140 18.0 (21.6) 91 22.3 (28.1) 1.5
MI 154 19.8 (23.8) 90 22.1 (27.8) 1.7
MC 283 36.4 (43.7) 92 22.5 (28.4) 3.1
MM 1 0.1 ( 0.2) 1 0.2 ( 0.3) 1.0
CP 34 0.4 ( 5.3) 18 4.4 ( 5.5) 1.9
647 83.2(100.0) 324 79.4(100.0)
NS 15 1.9 (11.5) 9 2.2 (10.7) 1.7
RM 7 0.9 ( 5.4) 3 0.7 ( 3.6) 2.3
AC 52 6.7 (40.0) 27 6.6 (32.1) 1.9
OT 21 2.7 (13.8) 19 4.7 (22.6) 1.1
OC 18 2.3 (13.8) 15 3.7 (17.9) 1.2
OS 13 1.7 (10.0) 7 1.7 ( 8.3) 1.9
OR 3 0.4 ( 2.3) 3 0.7 ( 3.6) 1.0
AR 1 0.1 ( 0.8) 1 0.2 ( 1.2) 1.0
130 16.8(100.0) 84 20.6(100.0)
Table 2-9. Trapping success of
type in Rio Doce State Forestry
small mammals calculated by habitat
Park, Minas Gerais, Brazil.
NUMBER OF NUMBER OF
HABITAT TYPE TRAP NIGHTS CAPTURES % SUCCESS
represented about 7X of the total captures in the native forested
habitat, but 40% of the rodent captures (Table 2-8). In the eucalypt
forest, A. cursor represented about 16X of the total captures and 93%
of the rodent captures (Table A-5). This rodent was the dominant
species captured in the wet meadow habitat, representing about 85%
of both the total and of the rodent captures.
In both the native and eucalypt forested habitats, marsupials in
general were recaptured at a high rate. Didelphis marsupialis and
Marmosa microtarsus both showed low recapture rates and reflect the
small sample size. Especially noteworthy was the relatively high
recapture rate of Marmosa cinerea (Tables 2-8 and A-5). No
individuals of other species, neither rodent nor marsupial, were
recaptured as frequently as individuals of this species. In the native
forested habitat, individuals of M. cinerea were recaptured on the
average 3.1 times, while in the eucalypt forest, individuals of this
species were recaptured on the average 7.7 times.
Akodon cursor was the only rodent that had a relatively high
number of captures and recapture rate (Tables A-5 and A-6). This
species had a recapture rate of 1.6 and 3.0 in the eucalypt and wet
meadow habitats, respectively.
Table 2-9 presents the trapping success by habitat type.
Trapping success was calculated for small mammals in the native
forested habitat. The platform trapping data was excluded. It must
be kept in mind that the sampling effort in each general habitat
category was different, however, comparisons of trapping success are
COMPARISON OF TRAPPING
SUCCESS IN THREE HABITATS
NOV DEC JAN FEB MAR APR MAY
JUN JUL AUG SEP OCT
-- NATIVE FOREST
forest in the
Comparison of trapping success of small mammals in
types: eucalypt forest, wet meadow, and native
Rio Doce State Forestry Park, Minas Gerais, Brazil.
the result of the number of captures relative to the number of
trapping opportunities or nights. Overall, the wet meadow habitat
yielded the highest trapping success (18.8X) while the homogeneous
eucalypt habitat generated the lowest trapping success (0.02%).
Figure 2-4 compares the progression of the trapping successes of
the native forested habitat (without the platform data), the wet
meadow habitat and the eucalypt forest with subcanopy habitat.
Although these three habitats have unequal sampling effort and
trapping design, these habitats were sampled for the period of one
year and show important temporal trends. From the overall gross
comparison portrayed in Figure 2-4 and the percent success presented
in Table 2-9, in contrast to the wet meadow habitat, it appears that
the forested habitats, both native and exotic, were quite similar in
overall percent success and monthly trapping success. The wet
meadow trapping success fluctuated greatly throughout time, from a
high approaching 45% in March and April, to a crash lower than 10X
in May, October, and November.
I plotted the total captures, minimum known alive and first
captures through time for each of the three habitats. Figure 2-5
shows the capture curves for the native forested sites. Trapping
success was relatively low and stable from November through April,
with a noticeable increase in June, July and the early part of August.
This peak in trapping success dropped back to the levels observed
prior to the increase. Figures 2-6 and 2-7 compare the trapping
success of the eucalypt forest and the wet meadow habitat,
respectively. The highest number of captures at any one time in the
CAPTURE CURVES FOR ALL
NUMBER OF CAPTURES
NOV DEC JAN FEB MAR APR MAY JU JUL A SEP
NOV DEC JAN FEB MAR APR MAY JUN JUL AUG SEP
-TOTAL CAPTURES MKA CAPTURES -*-FIRST CAPTURES
Figure 2-5. Small mammal capture curves for all native forested
sites in Rio Doce State Forestry Park, Minas Gerais, Brazil.
Capture curves include total captures, minimum known alive (MKA),
and first captures.
CAPTURE CURVES FOR
RD/E EUCALYPT SITE
NUMBER OF CAPTURES
0 --1-- 1--1---1- '1------:44-- -- --
NOV DEC JAN FEB MAR APR MAY JUN JUL AUG SEP OCT
TOTAL CAPTURES MKA CAPTURES +- FIRST CAPTURES
Figure 2-6. Small mammal capture curves for eucalypt forest in
the Rio Doce State Forestry Park, Minas Gerals, Brazil. Capture
curves include total captures, minimum known alive (MKA), and
RD/B WET MEADOW S1TE
NUMBER OF CAPTURES
DEC JAN FEB MAR APR MAY JUN JUL AUG SEP OCT
-- MKA CAPTURES
* FIRST CAPTURES
Figure 2-7. Small mammal capture curves for wet meadow habitat
in the Rio Doce State Forestry Park, Minas Gerais, Brazil.
Capture curves include total captures, minimum known alive (MKA),
and first captures.
Table 2-10. Trapping success by trap type for all species in all
habitat types. Trap types are arranged according to trapping
location: terrestrial or arboreal. Trap types are as follows: 1ST=
small terrestrial Sherman live trap; 3MT= medium sized terrestrial
live trap; 5LT= large terrestrial live trap; 2SA= small arboreal
Sherman live trap; 4MA= medium sized arboreal live trap; 6PA=
arboreal platform trap.
NO. NO. PERCENT
TRAP TYPE CAPTURES TRAP NIGHTS SUCCESS
1ST 370 4080 9.1
3MT 553 18000 3.1
5LT 42 5760 0.7
965 30480 3.2
2SA 48 2640 1.8
4MA 228 8640 2.6
6PA 66 1050 6.3
342 12330 2.8
wet meadow habitat was double the highest in the eucalypt forest.
However, the trends were similar. Both habitats showed two
pronounced peaks in their respective capture curves that corresponded
to the same months throughout the year. There was a peak in
February, March, and April followed by a crash, and another peak in
June, July, and August followed by another crash.
Overall, trapping success was higher with terrestrial traps than
with arboreal ones (Table 2-10). Caution should be used in these
comparisons as the number of trap nights are unequal; almost three
times the number of terrestrial trap nights as the number of arboreal
ones. However, I feel that some degree of comparison can be drawn
from this analysis based upon the number of captures relative to the
number of trap nights.
Trapping success by trap type varied considerably (Table 2-10).
Small terrestrial Sherman live traps were the most successful (9.1%),
medium terrestrial traps represented the trap with the greatest
number of captures and trapping opportunities. and large terrestrial
live traps were relatively unproductive.
Arboreal trap type success varied. Small arboreal Shermans
were the least productive arboreal trap type with only 1.8% success.
The arboreal platform traps had a trapping success of 6.3%.
Table 2-11 reports the number of total captures of each species
and the number of captures per trap type. Odd numbered traps
represent terrestrial traps and even numbered ones represent arboreal
Table 2-11. Trap response by species across all habitat types.
Trap types are explained in Table 2-10. Species abbreviations are
explained in Table 2-4.
TOTAL 1ST 2SA 3MT 4MA 5LT 6AP
42 0 0 24 2 15 1
158 0 0 129 3 26 0
171 13 14 106 38 0 0
361 2 29 145 150 0 35
51 0 0 9 13 0 29
1 0 0 0 1 0 0
19 4 0 13 1 1 0
7 0 0 3 3 0 1
392 312 2 78 0 0 0
21 1 3 9 8 0 0
21 8 0 11 2 0 0
10 7 0 3 0 0 0
1 0 0 0 1 0 0
44 18 0 20 6 0 0
2 2 0 0 0 0 0
4 3 0 1 0 0 0
2 0 0 2 0 0 0
1307 370 48 553 228 42 66
28.3 3.7 42.3 17.4 3.2 5.0
traps. In general, the number of captures per species in terrestrial
and arboreal traps represented the spatial adaptation for each species.
For marsupials, Didelphis marsupialis and Metachirus nudicaudatus
were captured principally in terrestrial medium live traps (3MT),
while Marmosa incana was captured in all trap types except for the
large terrestrial and arboreal platform traps. Marmosa cinerea was
captured in all trap types except the large terrestrial traps.
Caluromys philander was trapped principally in arboreal medium live
traps and arboreal platform traps. The sample size for rodents was
too small to allow for a clear trapping trend. Akodon cursor is the
clear exception. The small terrestrial Sherman live trap was most
effective for this species. This was mostly a consequence of the
approximately 80X of all Akodon captures made in the wet meadow
habitat (Table 2-10). Orvzomvs subflavus was trapped principally in
terrestrial small shermans and medium live traps.
The trapping success realized in this study for neotropical humid
forests falls within the range of observed success rates (Table 2-12).
However, one striking difference between this and other neotropical
small mammal studies was the high number of marsupial captures
relative to rodent captures (Table 2-12). All reported studies show
rodent biases, and usually high captures of rodents relative to
marsupials. Only Emmons (1984) reported a marsupial to rodent
capture ratio approaching equality (Table 2-12). Fonseca (pers.
comm.) reported marsupial biased trapping results from a variety of
Table 2-12. Percent capture by taxonomic group, trapping success
and number of trap nights by habitat type for this study compared
to other Neotropical field studies. % M= percent of total
marsupial captures; X R= percent of total rodent captures; % T.S.=
percent trapping success; # T.N.= number of trap nights.
STUDY SITE % M % R % T.S. # T.N.
DIETZ ET AL., 1975
NITIKMAN & MARES,
LAEMMENT ET AL.,
LACHER & ALHO,
BORCHERT & HANSON,
HOMOGENEOUS EUCALYPT FOREST
DIETZ ET AL., 1975
BRAZIL 83.0 17.0
EUCALYPT FOREST WITH NATIVE SUBCANOPY
native forest sites in eastern Minas Gerais. Avila-Pires (1978)
captured 245 rodents and 40 marsupials from the Rio Doce Park.
Gastal (1982) reported that 5 species of marsupials were captured in
the Park but gave no comparative data for rodents. Dias (1982)
trapped more rodents than marsupials in the Rio Doce Valley in
Hunsaker (1977) stated that marsupials require considerable
effort to trap. Perhaps one explanation for the observed high
marsupial/rodent capture ratio could be due to the habitat type found
in the Park. There is very little primary habitat in the Park relative
to secondary habitat (Table 2-1). Most of the forest habitat in the
Park has been altered by fire in the recent past (Lopes, 1982). The
primary forest plots that I sampled yielded the lowest species
richness and absolute captures of didelphid marsupials relative to the
other secondary forested habitats (J. Stallings, in prep.). Charles-
Dominique (1983) suggested that didelphid marsupials can reach high
local densities in areas of abundant food resources. He postulated
that these species are r-strategists and are adapted to the "unstable
environment of secondary forests." In Panama, Didelphis marsupialis
tended to occur at higher densities in primary forest, while
Caluromvs and Philander were found at higher densities in secondary
forest than primary forest (Fleming, 1972).
Marsupials were recaptured with great frequency in this study,
especially Metachirus nudicaudatus, Marmosa incana and Marmosa
cinerea. My recapture data on marsupials agree with data reported
by Fleming (1972, 1973), August (1984) and to some degree with that
found by O'Connell (1979). My data do not agree with Hunsaker
(1977) who stated that didelphid marsupials are difficult to recapture.
For example, individuals of Marmosa cinerea were captured on the
average 7.7 times in the eucalypt forest with native subcanopy
habitat. This high recapture rate could be explained by the fact that
this habitat in effect was surrounded by habitat unsuitable for
arboreal species. In essence, this habitat was a forested island. On
one side there was a monoculture of Eucalyptus salina with no
subcanopy, on another a marsh turned into a rice field, and the other
two sides of the forest were bounded by pasture. Arboreal species
should spend more time in arborescent vegetation than on the ground.
These species should be more hesitant to disperse than terrestrial
Another obvious difference between this study and other
inventories conducted in neotropical native forests was the absence of
echymid rodents. Species of the genus Proechimvs are the most
widespread taxa of the family Echymidae in the neotropics
(Hershkovitz, 1969). These forest species are terrestrial and usually
appear on species lists from forest inventories. In Panama,
Proechimvs was a common forest capture (Glanz, 1982; Eisenberg and
Thorington, 1973). Handley (1976) reported Proechimvs as a common
species in Venezuela. Emmons (1984) and Terborgh et al. (1986)
reported captures of Proechimvs from forested sites in Peru and
Ecuador. There are several reports of Proechimvs captures in
Brazilian tropical moist forests. Laemmert et al. (1946), Emmons
(1984), Carvalho (1965), Miles et al. (1981) and Malcolm (pers. comm.)
reported Proechimvs in their inventories from the Brazilian Amazon.
In the Atlantic Forest, Davis (1945) and Fonseca (pers. comm.)
reported two species of Proechimvs from their studies in the states
of Rio de Janeiro and Minas Gerais, respectively. Dias (1982) trapped
one species of Proechimvs in 3 study areas in the Rio Doce Valley.
I did not capture one individual of Proechimvs from the Park in
approximately 35,000 trap nights from 1985-86, nor from an additional
30,000 trap nights from 1986-87 (J. Stallings, in prep.). One
explanation could be the presence of predators in the Park.
Eisenberg (1980) speculated that the abundance of rodents in some
neotropical sites and the paucity of rodents in other sites could be
the result of the presence or absence of top predators. Hershkovitz
(1969) stated that species of the genus Proechimys "are the basic
source of protein for lowland predators in the Brazilian subregion."
The felid community in the Park is intact. All of the felids have
been observed by field workers in the recent past. I saw spoor from
jaguar, puma, ocelot, Geoffroy's cat, and jaguarundi.
In the wet meadow habitat my findings were consistent with the
literature (Table 2-12). In every other study, rodent captures were
higher than marsupial captures. In this study, Akodon cursor was the
dominant species in terms of absolute numbers and captures.
O'Connell (1981) reported that Zvgodontomvs was the dominant rodent
in grass habitat in Venezuela and represented 85X of the total rodent
captures. One difference between this study and others was the
observed high trapping success. As can be observed from Table 2-4,
Akodon cursor made up 85% of the captures and was responsible for
the high trapping success.
The eucalypt forest with native subcanopy habitat yielded
surprising results. I did not expect to find many small mammals in
this habitat because of "plantation effects." However, I captured 7
species of small mammals, most of these being marsupials. In fact,
the marsupial/rodent capture ratio was similar to that observed in the
native forested habitat (Table 2-12). I could not find any studies in
the literature to compare to the results from this habitat. Perhaps
the study that sampled homogeneous eucalypt forest would be the
most appropriate (Dietz et al., 1975). Dietz et al. (1975) captured
two species of terrestrial rodents, Orizomvs nigripes and Akodon
cursor, in the homogeneous plantations with grass/bamboo
undergrowth. The authors caught a total of 5 species, only one of
which was not strictly terrestrial, in the two native forested habitats.
The plantation habitats in this and my study are similar in that they
both were eucalypt plantations of similar age and that the terrestrial
substrate of both was covered by grass. The major difference was
the native species subcanopy. I captured a relatively high number of
terrestrial rodents and marsupials and a high number of arboreal
marsupials. I did not capture any arboreal rodents. The native
species subcanopy could be considered a secondary forest sere, if the
emergent eucalypt stratum is ignored. Following Charles-Dominique's
(1983) hypothesis of increased didelphid marsupial abundances in
secondary habitat, it is not surprising that I captured marsupials at a
rate consistent with the native forested habitat.
The temporal capture results suggested that there were two
systems operating in the Park. The trapping data showed a
pronounced peak in the total number of captures, mka and first
captures for the native forested habitat during the cool, dry winter.
Davis (1945) reported a similar trend in the trapping results and
suggested that this trend was the result of more younger individuals
present in the trapping pool or because of a paucity of natural food
items during this time of year. My data do not support the
hypothesis that more younger individuals explain the pronounced
increase; rather it appears that food resource paucity results in the
increase (Chapter III).
The eucalypt and wet meadow habitats showed similar trapping
trends with a peak in the cool, dry winter, and a peak in late
summer. These two habitats might have yielded similar capture curves
because of the grass substrate. Akodon cursor was an important
component of both habitats and is an insectivore/omnivore that is
reported to use a high proportion of grass and grass seed in its diet
(Nowak and Paradiso, 1983). Marmosa incana and Metachirus
nudicaudatus are insectivore/omnivores and frugivore/omnivores,
respectively, and perhaps track insect availability in grass substrate.
The grass species did not produce seeds until late May. Thus,
perhaps the peak observed in February, March and April can
be explained by the lack of food for both rodents and marsupials.
The second peak, which occurred in June, July, and August, could
also be explained in terms of a decrease in food availability. Insect
and fruit availability are usually low during the hibernal period in
seasonal neotropical forests (e.g., Janzen and Schonener, 1968). The
marsupial species rely heavily on these food resources. The results
of a preliminary stomach content analysis on Akodon, suggest that
insects are important items in this species diet (Stallings, in prep.).
Graminoids in this habitat were dry and seeds were not as readily
available as they were during April and May.
The trap type data analysis revealed that terrestrial small
Sherman live traps were very productive in the wet meadow habitat
but yielded relatively few captures in the forested habitats. The
same results were obtained for arboreal small Shermans in the
forested habitats. Large terrestrial live traps were unproductive in
the forested habitats. The most productive trap types were the
medium sized terrestrial and arboreal traps and the arboreal platform
traps. Some individuals of species that are considered terrestrial
were captured in arboreal traps. This can be explained by some low
arboreal traps that were connected to the ground by either a vine or
The use of the arboreal platform traps did not allow me to trap
additional species that were not already trapped using the terrestrial
and low arboreal traps. However, I was able to increase the
frequency of capture of the highly arboreal marsupial Caluromvs.
Malcolm (per. comm.) obtained similar results with platform traps in
Manaus. Perhaps the first use of arboreal platforms for trapping
small mammals were in the studies described by Davis (1945) and
Laemmert et al. (1946). Unfortunately, I could not determine the
success of these traps nor the species captured from these studies.
In total, I logged 49 captures of this Caluromvs in both the eucalypt
with native species subcanopy and the native forested habitats. In
the latter habitat, I only trapped this species 5 times in the
terrestrial and low arboreal traps. I trapped 29 Caluromvs in the
native forested habitat by using the arboreal platforms. I would have
underestimated the presence of this species had I not used the
platform traps. Marmosa cinerea was trapped also with relative high
frequency in this trap type.
I was surprised to find such a high trapping frequency of
Caluromvs in the eucalypt forest with native species subcanopy.
Although this species can be quite common in native species forested
habitats, I found it unusual that this highly arboreal frugivore would
be inhabiting an exotic monoculture plantation. There obviously was
sufficient-food resources available from the native species subcanopy.
The fact that this species was captured in terrestrial and low
arboreal traps, without using platforms, suggests that this species was
using the subcanopy.
The results of this small mammal inventory suggest that
marsupials play an important role in the community structure of small
mammals in one of the largest remaining native tracts of Atlantic
forest in Brazil. Wet meadow habitat in this region is speciose in
rodents, and perhaps dominated by one or two species. Eucalypt
forests with native species subcanopy can play an important role in
the conservation of small mammal communities in a region greatly
altered by monocultural plantations.
TEMPORAL VARIATION IN TRAPPING SUCCESS OF
DIDELPHID MARSUPIALS IN AN EASTERN BRAZILIAN PARK
Tropical rainforests are classified as stable and evergreen, with
little seasonal change in the flora and fauna (Richards, 1952; Smith,
1974). However, tropical seasonal and semideciduous forests exhibit
seasonal changes in floral phenology (e.g., Augspurger, 1982; Foster,
1982; Garwood, 1982) and consequently variation in faunal populations
and activities (e.g., Howe, 1982; Worthington, 1982; Smythe et al.,
1982; Glanz et al., 1982). These changes have been related to
variation in the abiotic environment, such as changes in photoperiod,
precipitation, and temperature (Sinclair and Norton-Griffiths, 1979;
Leigh et al., 1982).
Tropical forests that exhibit a pronounced dry season seem to
have a peak of fruiting at the onset of the wet season immediately
following the dry season (Foster, 1980). While the reasons for this
peak in the fruiting period are not clear, the result is that fruit is
more abundant at certain times of the year. This increase in the
availability of fruit drastically affects the animals that use these
resources for food.
Population fluctuations interpreted by trapping results of small
mammals must be carefully analyzed to determine if the population is
actually fluctuating or if animals are responding to trap bait during
periods when food is available (Hunsaker, 1977). Previous research
on didelphid marsupials in the neotropics (Davis, 1945; O'Connell,
1979; Fleming, 1972, 1973; August, 1984) have found that trapping
success of marsupials increased during the hibernal and prevernal
months. This increase in trapping success has been attributed to the
paucity of food resources during this period. Another interpretation
of the results could be linked to reproduction, such as greater
movements of males or a large number of immatures. These might be
related to the initiation or the cessation of reproductive activity.
One objective of this paper is to test the null hypothesis that
trapping success of marsupials does not vary with season in the
Atlantic Forest of eastern Brazil. Another objective is to test the
null hypothesis that any peak in trapping success was a consequence
of an increase in the population or reproductive activities.
Materials and Methods
Small mammals were trapped from October, 1985, through
September, 1986, in the Rio Doce State Forestry Park, Minas Gerais
state, Brazil. The Park lies between the coordinates 19 48'18" and 19
29'24" south latitude and 42 38'30" and 42 28'18" west longitude. The
climate of the Park is classified as tropical humid with a rainy season
between the months of November and February and drought during
the months of June, July, and August. Mean annual temperature is
approximately 22 C, with mean minimum monthly temperatures
fluctuating greatly throughout the year (Figure 3-1). Both the mean
MEAN MINIMUM & MAXIMUM
Figure 3-1. Temperature graph showing both minimum and maximum
mean temperatures per month in the Rio Doce State Forestry Park,
Minas Gerais, Brazil. Winter and summer seasons are indicated
along the X axis.
maximum and minimum monthly temperatures reach their respective
lows during the drought season.
For the purposes of this paper, I divided the year into two
seasonal periods. This division is based on the monthly mean
minimum temperature. If the mean monthly minimum temperature was
above 17 C. the month was classified as summer, otherwise it was
considered the winter period. A comparison of my seasonal
classifications and those of Davis (1945) shows thermal and intuitive
consistencies (Figure 3-1). My winter season corresponds to the
period from April September and the summer season from October -
March. My classification corresponds to those of Davis in the
following manner: my winter season includes the autumnal, hibernal
and prevernal seasons, while my summer season includes the vernal,
aestival and serotinal seasons.
The phytogeographical domain of the Park has been classified by
several authors. For example, according to Ab Saber (1977, in
Gilhaus, 1986), the Park occurs within the Tropical Atlantic Domain.
Rizzini's (1963, in Gilhaus, 1986) classification defines the Park's
domain as the Atlantic Province, Austro-Oriental sub-province,
Cordilheira sector. Gilhuis (1986) correctly pointed out that both
classification schemes locate the Park in a transitional zone bordering
the more humid Tropical Atlantic Domain and the drier Cerrados
Domain. I found tree species from both domains to be present in the
Rio Doce Park. Trees are more deciduous than would be expected in
Tropical Atlantic Domain forest. Most of the emergent tree species
and some others in the upper strata lose their leaves during the cool
and dry hibernal and prevernal seasons (Gilhuis, 1986; Stallings, pers.
obs.). The forest of the park is classified as Tropical Semideciduous
by Alonso (1977, in Gilhaus, 1986), Tropical Broadleaved by Azevedo
(1969, in Gilhaus, 1986) and Tropical Pluvial Seasonal by Lima (1966,
in Gilhaus, 1986).
Small mammals were live-trapped at 5 forested sites within the
Park boundary. Additional trapping was conducted in and around the
Park and the methodology is described in Chapter II. For the
purposes of testing the above mentioned hypotheses, I recorded the
following observations on each captured individual. The initial
capture of each individual was a first capture and a numbered ear tag
was placed on the left ear of all animals to identify individuals that
were captured previously. Total captures were first captures plus
subsequent captures of the same individual. The minimum known
alive (MKA) were the number of individuals captured on the first
occasion and those individuals recaptured only on one occasion during
each trapping period. Sex, age and reproductive information was also
The probability of capturing a small mammal is affected by 3
factors: 1) temporal effects, 2) behavorial effects, and 3)
heterogeneity effects (White et al., 1982). The effects of
heterogeneity (e.g., number and placement of traps in an animal's
home range or variation due to social dominance) vary among animals,
but capture probabilities for each animal remain constant per
occasion. In contrast, temporal (e.g., variation over time because of
environmental conditions or trapping effort) and behavior (e.g., trap
avoidance or fascination) effects can alter capture probabilities after
the initial capture of an individual.
First capture data are more affected by heterogeneity and
temporal effects than behavioral effects (White et al., 1982). An
increase in the number of juveniles in first capture data through time
suggests that the population is increasing. However, if young
individuals do not account for an increase in first capture data, then
an alternative explanation is that the increase reflects individuals
present in the population that were not previously captured.
Total capture and minimum known alive data are influenced by
temporal, behavioral, and heterogeneity effects. Total capture data
are the result of all captures per occasion or through time, inclusive
of first captures, successive recaptures, and recapture of individuals
separated by time. These data should not be interpreted as a good
estimator of population abundance because they are strongly affected
by behavioral and temporal factors. An increase in recapture data
through time can be interpreted as a change in the capture
probabilities of marked individuals due to trap fascination, rather
than trap shyness which would reduce capture probabilities.
The minimum known alive data are the best population estimator
of the three. These data incorporate first captures and only the first
recapture of an individual captured in a previous occasion for each
trapping period. This estimator, in effect, uses the number of new
captures plus survivorship of animals captured previously to obtain an
estimate of abundance.
I only used marsupial capture data to test the null hypotheses
that trapping success does not vary by season and that any expected
increase is due to reproduction. Marsupials comprised 83X of the
total small mammal captures and 80% of the first captures (Chapter
II). Marsupial species included in the analysis were Didelohis
marsupialis (azarae), Metachirus nudicaudatus, Marmosa incana,
Marmosa cinerea, and Caluromvs philander. Fruit and insects are
important components of this group's diet (Chapter II).
I used correlation analysis and simple regression to determine if
there was any association between trapping success and mean
minimum monthly temperature. I quantified trapping success of first
captures, minimum known alive and total captures for each season.
Chi-square tests for goodness of fit were then used to test for
significant deviation from expected capture frequencies between
I looked at (1) an increase in the number of juveniles during
the winter season, (2) an increase in the number of males between
seasons, (3) a male biased sex-ratio during the winter season, and (4)
an increase in the number of lactating/pregnant females during the
winter season to test the null hypothesis of an expected increase due
to reproduction. For this analysis I used Chi-square tests goodness
of fit to test for significant deviation from the above expected
Individuals were placed into either juvenile or adult subjective
age classes. Initially, the age class placement was based on the
overall morphological gestalt of each individual per species. I then
used the recorded body measurements (Chapter II) to determine the
threshold between the two age classes and I compared field
determined ages to ages derived from body measurements. Body mass
and hind foot measurements were relied on heavily to determine age
Age class categories of the 5 species of marsupials are presented
in Table 3-1 and are taken from Chapter III. For Didelphis
marsupialis. juvenile and adult age categories were based on the hind
foot size. This species can attain a body mass of 2 kg. The hind
foot would grow slower in comparison to smaller bodied species and
body mass would provide a better indicator of age. Animals with a
hind foot of 51 mm or smaller were classified as juveniles (average
weight 417 gm). Lengths greater than 51 mm (average weight 1049
gm) were classified as adults. I used body mass to determine the
threshold for Metachirus nudicaudatus, Marmosa incana, and Marmosa
cincera. These species are relatively small and foot size reaches its
maximum at an early age. Body mass provided the better estimator
because mass increased as the individuals grew older. Metachirus
nudicaudatus was placed into age categories based upon a threshold
weight of 90 gms. Age class placement was determined by a weight
of 35 gm and 50 gm for Marmosa incana and M. cinerea, respectively.
All Caluromvs philander that were captured were considered adults.
The number of individuals of each sex were sorted for first,
MKA, and total capture data between and within seasons.
Comparisons were made for sex biased trapping success between
Table 3-1. Criteria used to place marsupials into either juvenile
or adult age categories. Individuals per species were classified
as juveniles if designated measurements were less than cut-of
measurements used in table. Species were not sexually dimorphic
based on body measurements listed.
SPECIES MEASUREMENT USED MEASUREMENT
Didelphis marsupialis HIND FOOT 51 MM
Metachirus nudicaudatus MASS 90 G
Marmosa incana MASS 35 G
Marmosa cinerea MASS 50 G
seasons for all capture data. In addition, comparisons were made for
sex biased trapping success within seasons.
Female marsupials were examined to determine if they were
lactating and for the presence of young attached to the teat field.
Females with young attached were considered pregnant. Frequencies
of lactating/pregnant females were compared on a seasonal basis.
There was a significant negative association observed between
mean monthly temperature and trapping success through time. This
association is significant only if we look at the number of total
captures (f=28.444, p < .0003) (Figure 3-2) or the minimum known
alive (f=9.939, p < .007) (Figure 3-3). First capture data did not
show a significant association (Figure 3-4).
The first capture data (0.05
0.001) and total captures (p< 0.001) of marsupials were greater than
expected in the winter season than in the summer season (Table 3-2).
Juveniles were observed throughout the year. Juveniles were
represented from four of the five species of marsupials (Table 3-3).
There was no significant difference in the minimum known alive or
first captures of juveniles between seasons. A significant difference
(p < .001) was observed in the total captures of juveniles between
seasons. More total captures of juveniles were observed in the
winter season than expected.
The number of juveniles captured per species between seasons
was fairly constant (Table 3-3) from minimum known alive and first
TEMPERATURE AND TOTAL CAPTURES
NUMBER OF CAPTURES
0 I-I I
0 5 10 15 20
TEMPERATURE (C )
Figure 3-2. Relationship between temperature and total captures
of didelphid marsupials in the Rio Doce State Forestry Park,
Minas Gerais, Brazil.
TEMPERATURE AND MKA CAPTURES
NUMBER OF CAPTURES
5 10 15 20
TEMPERATURE (C )
Figure 3-3. Relationship between temperature and
alive captures (MKA) of didelphid marsupials in
State Forestry Park, Minas Gerais, Brazil.
the Rio Doce
TEMPERATURE AND FIRST CAPTURES
NUMBER OF CAPTURES
Figure 3-4. Relationship between temperature and first captures of
didelphid marsupials in the Rio Doce State Forestry Park, Minas
Table 3-2. Comparison of the first captures, minimum known alive
(MKA) and total captures of five species of didelphid marsupials
between seasons in the Rio Doce State Forestry Park, Minas Gerais,
Brazil. Significant at 0.05-0.01. ** Significant at 0.01-
WINTER SUMMER CHI-SQUARE
FIRST CAPTURES 206 151 4.23*
MKA CAPTURES 276 157 16.35**
TOTAL CAPTURES 392 190 35.05**
Table 3-3. The number of first captures (FIRST), minimum known
alive (MKA) and total captures (TOTAL) of juvenile didelphid
marsupials per season. N.S= Non significant.
FIRST MKA TOTAL FIRST MKA TOTAL
Q. marsupialis 3
M. nudicaudatus 7
M. incana 15
8. cinerea _
P < 0.001
capture data. The total capture data per species indicated that there
was a recapture bias during the winter season, although the
difference was not significant. Figure 3-5 shows that first captures
of adults, not juveniles, clearly represented the majority of captures
during May, June, and July.
The number of captures of each sex between seasons is
presented in Table 3-4. There were no significant differences
between the number of males captured between seasons or the number
of females captured between seasons using the first capture data.
Minimum known alive and total captures data showed significant
differences in the expected number of males or females between
seasons. More males and more females were captured during the
winter season than during the summer season, with a higher
proportion of males being captured during the winter season. For the
first capture data, minimum known alive and total capture data there
were no significant differences in the male/female sex ratio within
each season (Table 3-5).
More lactating females, whether carrying young or not, were
captured during the summer season than the winter season (Table
3-6). However, this difference, although approaching significance,
was not statistically significant (x2 =2.89; .10 < p <.05). One
striking observation was the obvious decrease in the occurrence of
lactating/pregnant females during the months of June, July and
August (Figure 3-6). This decrease is also observed when comparing
number of juveniles and lactating females per species through time
(Figures 3-7, 3-8, and 3-9).
OF ADULTS AND YOUNG BY MONTH
NUMBER OF CAPTURES
NOV DEC JAN FEB MAR APR MAY JUN JUL AUG SEP OCT
i NO. YOUNG
Figure 3-5. Comparison of first captures of adult and young
didelphid marsupials per month in the Rio Doce State Forestry Park,
Minas Gerais, Brazil.
Table 3-4. Between season comparisons of the number of first
captures (FIRST), minimum known alive (MKA) and total captures
(TOTAL) of didelphid marsupials per sex. Significant at 0.05-
0.01. ** Significant at 0.01-0.001.
BETWEEN SEASON COMPARISON
SUMMER SEASON WINTER SEASON X2
NUMBER OF MALES
FIRST CAPTURES 62 83 1.52
MKA CAPTURES 78 152 11.90**
TOTAL CAPTURES 97 215 22.30**
NUMBER OF FEMALES
FIRST CAPTURES 66 70 0.05
MKA CAPTURES 79 124 4.98*
TOTAL CAPTURES 92 174 12.63**
Table 3-5. Within season comparison of the number of first
captures, minimum known alive captures (MKA) and total captures of
didelphid marsupials per sex. N.S.= Non significant. Significance
level set at p < 0.05.
Table 3-6. The frequency of lactating/pregnant didelphid
marsupials trapped per season in the Rio Doce State Forestry Park,
Minas Gerais, Brazil. N.S.= Non Significant.
D. marsupialis 4 1
M. nudicaudatus 11 12
M. incana 9 0
M. cinerea 12 3
Caluromvs philander 1 3
TOTALS 37 19
NOV DEC JAN FEB MAR APR MAY JUN JUL AUG SEP OCT
P M. cinerea
I O Caluromys
Figure 3-6. Comparison of lactating females of five species of
didelphid marsupials per month in the Rio Doce State Forestry Park,
Minas Gerais, Brazil.
= M. incana
FREQUENCY OF YOUNG,
AND LACTATING FEMALES
NOV DEC JAN FEB MAR APR MAY JUN JUL AUG SEP OCT
Figure 3-7. Comparison of the number of captures of young and
lactating females of the didelphid marsupial Metachirus
nudicaudatus in the Rio Doce State Forestry, Minas Gerais, Brazil.
FREQUENCY OF YOUNG
AND LACTATING FEMALES
NOV DEC JAN FEB MAR APR MAY JUN
! I YOUNG
JUL AUG SEP OCT
JUL AUG SEP OCT
= LACTATING FEMALES
Figure 3-8. Comparison of the number of captures of young and
lactating females of the didelphid marsupial Marmosa incana in the
Rio Doce State Forestry Park, Minas Gerais, Brazil.
AND LACTATDNG FEMALES
NOV DEC JAN FEB MAR APR MAY JUN JUL AUG SEP OCT
= LACTATING FEMALES
Figure 3-9. Comparison of the number of captures of young and
lactating females of the didelphid marsupial Marmosa cinerea in the
Rio Doce State Forestry Park, Minas Gerais, Brazil.
There is obviously a connection between animal population
numbers and the availability of food resources (Leigh et al; 1982).
An intuitive model is that population increases should track and peak
with maximum production of food resources. By timing reproduction
prior to the peak of available food resources, individuals maximize
their opportunities for the survival of their offspring.
There was a clear peak in the trapping success of marsupials
during the winter months in the Rio Doce State Forestry Park.
There was a significant negative association between the number of
total captures or minimum known alive each month and mean
minimum monthly temperature. However, the first capture data did
not support these previous findings. These data suggest that the
recapture of individuals plays an important role in the trapping
success numbers and that the numbers of new individuals did not
increase during the winter season.
August (1984) found an increase in the capture success of
Marmosa robinsoni during the dry season in Venezuela, while
O'Connell (1979) reported an increase in the minimum known alive for
Marmosa robinsoni, M. fuscata, and DidelDhis marsupialis during the
same period in Venezuela. Fleming (1972) observed an increase in
the number of captures of three species of marsupials during the dry
season in Panama. Davis (1945) reported more captures and
recaptures of two species of didelphid marsupials, Didelphis
marsuDialls and Caluromvs philander, during the dry, cold season in
Research on neotropical marsupial mating behavior suggests that
typically males are more active prior to and during the reproductive
or mating season (Atramentowicz, 1982; Hunsaker, 1977). This may be
a consequence of the polygynous mating systems observed for these
small mammals (Eisenberg, 1981). There was no significant difference
between seasons in the number of first captures of male marsupials.
The same results were obtained for female first captures. There were
more male and female minimum known alive captures during the
winter season than the summer season. The same results were
obtained for the total captures of male and female marsupials during
the winter season, but at a higher level of significance. These data
again suggest that there is a recapture bias in the trapping data, and
that observed and significant differences in the number of captures
and individuals of each sex between seasons is a function of being
recaptured and not related directly to reproduction. However, the
minimum known alive and total capture data show male captures in
the winter season to be more significant than females captures during
the same period. These findings could be interpreted as increased
male activity over female activity at the onset of the reproductive
A direct measure of population growth is the number of young
observed at any point in time. My results indicate that during the
winter season the number of young individuals captured was not
greater than in the summer season, but more young individuals were
recaptured during the this season than in the summer season.
O'Connell (1979) reported an increase in the number of individuals of
Didelphis marsupia1is during the wet season in Venezuela. She
attributed this increase to the first captures of juveniles.
The last measure that was used to determine population growth
was that of the occurrence of lactating/pregnant females through
time. Statistically, there was no difference in the occurrence of
reproductively active females between seasons, although there was a
suggestion of more reproductively active females in the summer
season than in the winter season. O'Connell (1979) found that most
of the species she studied to be sexually inactive during the dry
It is clear from the previous analyses that there was a
pronounced increase in the number of animals known to be alive and
the overall trapping success during the hibernal and prevernal seasons
(Figure 3-10). However, the data do not support the hypothesis that
this significant increase is due to more juveniles present in the
population, nor does it result from increased male activity or more
lactating/pregnant females. I conclude that population growth of
marsupials was constant if we consider only first capture data.
Clearly, the trapping success curve was being driven by adult
individuals; individuals that were being recaptured more during a
certain period in time than at another.
I did not measure the availability of fruit or insects during the
study. Phenological studies (CETEC, 1981) in the park from 1977-
1981 found that trees were in flower and fruit throughout the year,
FREQUENCY OF CAPTURES
APR-MAY JUN-JUL AUG-SEP
MONTHS GROUPED ACCORDING TO SEASON
SEASONS GROUPED BY MONTHS (DAVIS, 1945)
Figure 3-10. Frequency of captures of didelphid marsupials by
season in the Rio Doce State Forestry Park, Minas Gerais, Brazil.
with a noticeable, but not statistically significant increase from
October through December (Figure 3-11). These studies also Indicated
that the months of June, July and August represented the lowest
incidence of tree species with flowers and fruits. There was also a
low incidence of tree species with flower and fruits in January.
There are no quantitative data available on the seasonality of insect
density in the park. However, studies in other Neotropical areas
report a positive association between increased insect abundance and
precipitation and temperature (Janzen and Schonener, 1968; Wolda,
1978, in August, 1984; Davis, 1945). It is possible that such a trend
occurs in the park due to the high seasonality of rainfall and
In other areas were neotropical didephids have been studied
(Fleming, 1972; O'Connell, 1979; August, 1984; Charles-Dominque,
1983), the dry season represented the period of the year when food
resources were scarce. In contrast, the three month period from
October December in the park showed more species in fruit and
flower than any other three month period. These trends are
interesting when compared to the low and peak three month trapping
success periods. The highest trapping success coincides with the
lowest number of tree species in fruit and flower and likewise the
lowest period of trapping success coincides with the maximum period
of fruit and flower production.
However, it must be stated that trapping success remained
relatively constant for the period of December and January when the
number of tree species with fruit and flower were observed to be
FREQUENCY OF FRUIT
NO. SPECIES IN FRUIT
OCT-NOV DEC-JAN FEB-MAR APR-MAY JUN-AUG AUG-SEP
SEASONAL PERIODS IN MONTHS
Figure 3-11. Frequency of species of trees and shrubs in fruit and
flower per seasonal period in the Rio Doce State Forestry Park,
Minas Gerais, Brazil.
low. This two month period has the highest mean temperature and
precipitation of the year. If insect density is high at this time, then
perhaps marsupials are able to switch to insect prey or other foods
in response to the unavailability of fruit.
I would interpret my results of a significant increase in total
trap success as a response to the lack of fruit during the subtropical
winter and not due to (1) greater movement of males (i.e., more
captures of males than females during this period) nor (2) a
significant increase in the number of juveniles captured nor (3)
because of more reproductively active females during this period.
Marsupials fell into traps easier during this period because they were
searching for food. This interpretation is consistent with the
findings of Atramentowicz (1982), Charles-Dominique et al. (1981),
Charles-Dominique (1983), O'Connell (1979), Fleming (1972, 1973),
Davis (1945) and August (1984) and suggests that reproduction and
weaning of young marsupials occur at the end or after the dry season
when food resources are more abundant.
SMALL MAMMAL ASSOCIATION AND MICROHABITAT
SELECTION IN AN EASTERN BRAZILIAN PARK
There are several studies on small mammal habitat selection in
temperate zones (e.g., Dueser and Porter, 1986; Dueser and Shugart,
1978, 1979; Hallett, 1982; Rosenzweig and Winakur, 1969), but there
are relatively few small mammal habitat studies conducted in the
neotropics. August (1983, 1984) studied the relationship between
small mammals and habitat structure in Venezuela. Lacher and Alho
(in press) and Lacher et al. (1988) reported habitat selection by small
mammals in grassland habitats in southern Brazil. Nitikman and
Mares (1987) reported microhabitat preferences of small mammals in a
gallery forest in central Brazil. The purpose of this paper is to
examine the effect of habitat structure on habitat use by small
mammals in an eastern Brazilian forest, the Rio Doce State Forestry
Preliminary analysis of the habitat in this Park suggested that
there are distinct differences in the forest structure across several
forested habitats (Gilhaus, 1986). Forest fire, in the form of intense
crown fire, has played an important role, at least in the recent past,
in structuring the forest. This study tested the hypothesis that small
mammal abundances vary with habitat and that species select
microhabitats within each habitat.
Materials and Methods
This study was conducted in the Rio Doce State Forestry Park
(Rio Doce), Minas Gerais, Brazil (19 48' 18" and 19 29' 24" south
latitude and 42 38' 30" and 42 28' 18" west longitude). The Park
contains over 35,000 ha and elevation ranges from 230 to 515 m. The
mean annual precipitation for the Park was 1480 mm. from 1954 to
1974, however annual precipitation during the study amounted to 947
mm. (Figure 4-1). Mean monthly maximum and minimum temperatures
vary appreciably throughout the year (Figure 4-2).
There are several distinct forested and open/field habitats in the
Park (Gilhaus, 1986). Slope, soil quality and moisture, and elevation
all affect habitat type. However, forest fire has played the major
role affecting the forested habitat in this Park. The vegetation of
the Park is classified as tropical semi-deciduous and most of the
emergent tree species lose their leaves during the cool dry months.
In 1964 and 1967, major forest fires burned approximately 30X of the
forest (Lopes, 1982; Silva-Neto, 1984). Fire is important because leaf
litter accumulates during the dry season.
During the course of this study, I surveyed small mammal
communities in 5 distinct forested sites which represent 3 habitat
types in the Park (Figure 4-3). The following is a brief description
of each study site.
WALTER AND LEITH
MONTHLY RAINFALL (MM) MEAN TEMPERATURE (0) 14
200 I O
JAN FEB MAR AP MAY JUN JUL AUG SEP OOT NO OE
AVERAGE ANNUAL RAINFALL U8O mm
a5 RnM ta-1SI" 4
WALTER AND LEITH
MONTHLY RAINALL (MMt)
MEAN TEMPERATURE (C
JAN FEB MA APO uA JUN JULL AUG SEP OCT it0 OE-.
ANNUAL AWEMAGE RAINFALL 80 mn
am FROM a101u-au
Figure 4-1. Walter and Leith climatic diagram characterizing
surplus precipitation and drought by month in the Rio Doce State
Forestry Park, Minas Gerais, Brazil.
TEMPERATURE (C )
DEC JAN FEB MAR APR MAY JUN JUL AUG SEP
MEAN MAXIMUM TEMP MEAN MINIMUM TEMP
DATA FROM 1954 1974
Figure 4-2. Mean minimum and maximum monthly temperatures in the
Rio Doce State Forestry Park, Minas Gerais, Brazil.
I I I-