Citation
Patterns of small mammal species diversity in the Brazilian Atlantic forest

Material Information

Title:
Patterns of small mammal species diversity in the Brazilian Atlantic forest Gustavo A. B. da Fonseca
Creator:
Fonseca, Gustavo A. B. da
Publication Date:
Language:
English
Physical Description:
viii, 233 leaves : ill. ; 28 cm.

Subjects

Subjects / Keywords:
Mammals -- Brazil ( lcsh )
Habitat (Ecology) -- Brazil ( lcsh )
Mammals -- Habitat ( lcsh )
Forest Resources and Conservation thesis Ph. D
Dissertations, Academic -- Forest Resources and Conservation -- UF
Genre:
bibliography ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1988.
Bibliography:
Includes bibliographical references (leaves 216-232).
General Note:
Typescript.
General Note:
Vita.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright Gustavo A. B. da. Fonseca. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Resource Identifier:
001147581 ( ALEPH )
20442513 ( OCLC )
AFP7244 ( NOTIS )

Downloads

This item has the following downloads:


Full Text
PATTERNS OF SMALL MAMMAL SPECIES DIVERSITY
IN THE BRAZILIAN ATLANTIC FOREST
By
GUSTAVO A. B. DA FONSECA

A DISSERTATION 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

1988




ACKNOWLEDGEMENTS

First and foremost, I would like to thank Dr. John G. Robinson, my major advisor, for his dedication to my work way beyond duty. His advice during planning, field work, analysis and writing of this dissertation was invaluable, and I will be forever thankful. Drs. John Eisenberg, Larry Harris, Nigel Smith, Melvin Sunquist, and Charles Woods, members of my committee, provided support in the planning of this study and improvements in earlier drafts. Dr. Kent Redford also read and commented on the manuscript. I am grateful to all of them. My study also benefited from discussions with Drs. Thomas Lacher and Michael Mares.
Once more, Dr. Russell A. Mittermeler supported my research
efforts by advocating my cause with the World Wildlife Fund-US, which financed the most substantial part of the costs involved in this study. Small mammals are also people, just smaller. Additional financial support was provided by the Program for Studies in Tropical Conservation, of the University of Florida, and by the Research Council of the Federal University of Minas Gerais, Brazil. The National Research Council of Brazil (CNPq) awarded me with a doctoral fellowship.




Dr. Celio Valle encouraged the field work with his enthusiasm,
vision, and eternal optimism. Several people helped during field work, too many, and in too many ways, that I apologize beforehand for failing to mention. Cecilia Klerulff was my most dedicated and frequent field assistant, and for that I will be eternally in debt. My small mammal data from the Rio Doce Park was partially collected by Jody Stallings and his crew. This study greatly benefited from his friendship and discussions while I was in Brazil. I would also like to thank Ludmilla Aguiar, Ilmar Bastos, Sonia Rigueira, Carlos Alberto Pinto, Ederson Machado (sometimes), Silverio Machado, Ney Carnevalli, Luiz Fernando Mello, Gisela Herrmann, Jairo Vieira, Eduardo Veado, Eduardo Sabato, Luiz Paulo Pinto, and Maria Cristina Alves.
Mr. Feliciano Abdalla and Dr. Antonio Cupertino kindly allowed me to work on their farms and provided housing. "Santinho" let us stay in his house at the expense of family problems, and I am thankful to him. The State Forest Institute of Minas Gerais (IEF) provided me with accommodations and gasoline at the Rio Doce State Park. I am very grateful to the staff of the Park. Eng. Ftal. Jose Lourenco Ladeira identified the trees in and outside of the Rio Doce Park. I am also indebted to him. Drs. M. Carleton, K. Creighton, L. Emmons, G. Musser, P. Myers, and J. Patton generously identified the small mammal voucher specimens.
Even risking making her angry, I cannot help but to recognize
that my wife's love was as helpful as her intellectual stimulation and




support. My son Bruno tolerated my bad moods and I hope to be able to repay him in the future. My parents continued to provide me with love, help and support. The knowledge that I can count on them at all times is one of my most valuable strengths. To all of them, I emotionally dedicate this dissertation.
I would also like to dedicate this work to Drs. Celia Valle and Russell Mlttermeier. Together and unselfishly they managed, in less than 5 years, to train and put to work more Brazilian biologists than others had in the last three decades. The vast majority of young Brazilian ecologists, previously so prone to early infanticide, and who will be working in Brazil in the next decade, will certainly have crossed the paths of these two visionary people. The history of Brazilian biological conservation sciences will soon have to pay its dues to Cello and Russ.




TABLE OF CONTENTS
Page
ACKNOWLEDGEMENTS ......................................... ii
ABSTRACT ................................................ vii
CHAPTER
1 INTRODUCTION
General Background ................................ 1
Study Organization ................................ 4
2 BIOLOGY AND NATURAL HISTORY OF BRAZILIAN
ATLANTIC FOREST SMALL MAMMALS
Introduction ...................................... 7
Materials and Methods ............................. 9
Study Sites .................................. 9
Climate ...................................... 12
Trapping ..................................... 13
Results and Species Accounts ...................... 19
Discussion ........................................ 83
Seasonality and Resource Use ................. 83
Life History Patterns ........................ 89
Population Turnover .......................... 91
3 SMALL MAMMAL SPECIES DIVERSITY IN
BRAZILIAN TROPICAL PRIMARY AND SECONDARY
FORESTS OF DIFFERENT SIZES

Introduction ......................................
Materials and Methods .............................
The Region ...................................
Study Sites ..................................
Trapping .....................................
Habitat Variables ............................
Statistical Methods and
Data Analysis ...................................
Dependent Variables ..........................
Predictors of Community
Structure ..................................
Primary Versus Secondary Forests .............
Area Size Effects ............................
Results ...........................................
Discussion ........................................

93 97 97 98 100
102
104 104
115 116 117 118
145




Conservation and Management
Implications .................................... 154
4 THE RELATIVE ROLE OF HABITAT SELECTION, COMPETITION AND PREDATOR PRESSURE ON THE
STRUCTURE OF TROPICAL FOREST SMALL MAMMAL
COMMUNITIES
Introduction ...................................... 157
Materials and Methods ............................. 161
Study Sites .................................. 161
Trapping ..................................... 163
Habitat Variables ............................ 165
Statistical Methods and Data Analysis ........ 166
Results ........................................... 170
Species Composition and
Abundance Patterns ......................... 170
Habitat Selection ............................ 176
Species Interactions ......................... 183
Predation .................................... 184
Discussion ........................................ 192
5 CONCLUSIONS AND SYNTHESIS ......................... 210
LITERATURE CITED ........................................ 216
BIOGRAPHICAL SKETCH ..................................... 233




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
PATTERNS OF S14ALL MAMMAL SPECIES DIVERSITY IN THE BRAZILIAN ATLANTIC FOREST By
GUSTAVO A. B. DA FONSECA
August 1988
Chairman: John G. Robinson
Major Department: Wildlife and Range Sciences, School of Forest Resources and Conservation
The influences that habitat structure, area size, and species
interactions have on the species richness and diversity of small mammal communities in the Brazilian Atlantic forest were investigated in a 17-months study. Three primary and three secondary forest plots of different sizes were subjected to a capture/recapture program. Trapping effort totaled 5T,120 trap nights. A total of 19 small mammal species (rodents and marsupials) were registered during the course of the study, with species richness and diversity achieving their highest at the second-growth forest plot of large size. Small mammal species diversity only responded to the increase in area size among secondary forests, with primary forests being fairly poor communities. Species




richness, species diversity and total number of individuals in communities were found to be correlated with several habitat attributes characteristic of second-growth forests. The common opossum, Didelphis marsupialis, is a dominant species in these small mammal communities. Densities of other small mammal species were found to be depressed by opossums. The higher predator pressure on opossums by mammalian carnivores in the larger forest patches may be the factor responsible for allowing plots of increasing size to be more species rich and diverse.

viii




CHAPTER 1
INTRODUCTION
General Background
The Brazilian Atlantic forest is one of the most threatened
ecosystems in the world, with habitat reduced to small islands and very little remaining in undisturbed state (Mittermeier et al., 1982; Fonseca, 1985a). These isolated areas are characterized mostly by second growth under different regeneration stages. The consequences for the vertebrate fauna inhabiting these habitat fragments is not known, except for a few primate species. This is aggravated by the fact that not much information is available on the natural history of Atlantic forest mammalian species. By investigating species composition, diversity and relative abundances of small mammal communities of these tropical forests, this study will address these questions.
Small mammal community ecology research has revealed several parameters that, to variable degrees, explain underlying community characteristics and organization. Local species composition and relative abundances of small mammals and birds have been partially explained by several different measures of habitat structure and




2
heterogeneity (Rosenzweig and Winakur, 1969; Johnson, 1975; M'Closkey, 1976; Dueser and Shugart, 1978; James and Warner, 1982; August, 1983, 1984; Lynch and Whigham, 1984). As has been observed with plants (Horn, 1974), animal communities can sometimes be expected to decrease in diversity in late successional stages (Connell, 1978), possibly as a consequence of the lower degrees of spatial heterogeneity in climax forests. Therefore, the elimination of primary forests in the Brazilian Atlantic region may not have resulted in the decrease in small mammal species diversity. Within this framework, this study investigated the nature of small mammal species richness and diversity of second growth, which is increasingly a characteristic of the present day habitat patches of the Atlantic forest.
The size of habitat fragments has also been used as a determinant of mammalian species richness, especially in insular communities (Dueser and Brown, 1980; Lomolino, 1984). MacArthur and Wilson (1967) proposed an immigration-extinction equilibrium model to explain number of species in islands. As the size of an island increases, the probability of species extinction decreases. If the island is close enough to a potential source of colonists, extinction is somewhat compensated for by the frequent addition of new colonists, but if the degree of isolation is the same for two islands of different sizes, the larger one will tend to support richer communities. Several studies on different taxa have shown that continental habitat patches can sometimes behave like islands, if the surrounding habitat provides




3
effective isolation from the sources of colonists (e.g. Vuilleumeier, 1970; Seifert, 1975; Brown, 1978; Faeth and Kane, 1978; Harris, 1984; Blake and Karr, 1987). Notwithstanding the fact that a few attempts have been made (Malcolm, 1987), area-size relationships have yet to be demonstrated for tropical small mammal communities. The structure of the habitat may also influence community composition and relative abundances in islands of increasing size. As Williams (1964) pointed out, as the size of the area increases, the probability that additional habitats with their distinct species will be incorporated into samples also increases. Therefore, both factors have to be taken into consideration in investigating the mechanisms that promote species diversity.
Recently, studies have began to incorporate both area-size relationships and habitat structure characteristics into models explaining community parameters. These have occasionally successfully separated the "pure" or dynamic effects of the size of the area from those related to environment structure (e.g., Simberloff, 1976; Weaver and Kellman, 1981; Buckley, 1982; 1985; Boecklen, 1986). In addition, both area and habitat structure can have an underlying effect in the determination of patterns of species interactions within communities, which in turn can potentially affect species composition and relative abundances of these assemblages. Competitive interplays are known to exert influence upon small mammal community characteristics (Price, 1978; Dueser and Porter, 1986; Lacher et al., in press), but mixed




results have been obtained in the attempts to demonstrate competition (Murua et al., 1987).
I believe that previous models, although successful to various degrees in their predictive power, were commonly too simplistic. The result is that significant amounts of variation in species richness and diversity between communities are left unexplained. Furthermore, single parameter models may conceal the underlying mechanisms by which species diversity Is promoted and maintained (Connor and McCoy, 1979). Therefore, their application to conservation schemes can be potentially dangerous (see for example, Zimmerman and Blerregaard, 1986). The multivariate approach, which takes into account the many more potential influences arising from area-size relationships, habitat structure, and species interactions, was chosen for this study. This approach favors the information conveying value of the method rather than the quantitative models it can possibly generate. Although the method results in models that may have limited applicability to different habitats, they frequently have higher heuristic value and are more biologically realistic.
Study Organization
The core of this study is presented in three main chapters. They are meant to either stand alone as individual studies or constitute additive sections of a major work. Chapter 2 deals with the biology




5
and natural history of the small mammal species that were recorded in this study. It presents the general aspects of reproduction and population dynamics, substrate use, diet, movements and persistence in traplines, for the species with large enough sample sizes to allow analysis. Morphometric data on all species are also presented. This section serves as a data base for the understanding of the following chapters.
In Chapter 3, 1 deal mostly with the relation between habitat
structure and area size, and the way in which they affect diversity and relative abundance of small mammals. Communities are studied as collective entities, with very little emphasis placed on the individual species. Six forest plots of different sizes and levels of disturbance were sampled for small mammals. In addition to species richness, the total number of individuals in the communities, and the Shannon-Wiener species diversity index, I also developed a second index of community structure that takes into account species abundances. I used a large number of vegetation parameters as independent variables indicative of habitat structure. Sizes of the forest plots sampled constituted the second most important variable under study.
The third main chapter, Chapter 4, investigates the structure of small mammal communities in terms of requirements of individual species and the competitive interactions among these species. It also examines the possible influences that habitat requirements and size of the area can have on the outcome of community composition and relative




6
abundances. It explores the mechanisms by which patterns observed in the previous chapter are determined.
Finally, Chapter 5 offers a synthesis of the results, the
conclusions, and provides a summary of overall major findings. It also explores possible conservation applications of the results of this research.




CHAPTER 2
BIOLOGY AND NATURAL HISTORY OF BRAZILIAN ATLANTIC FOREST SMALL MAMMALS
Introduction
The neotropical region encompasses some of the most threatened ecosystems in the world, and yet the mammalian fauna of this area still remains very poorly known (Mares and Genoways, 1982). In South America, the Brazilian Atlantic forest has been the vegetation formation subjected to the highest rate of destruction, with less than
5 % of the region possessing some form of forest cover (Fonseca, 1985a), and probably less than 1 % in undisturbed state (Mittermeier et al., 1982). Nonetheless, the few natural history studies conducted on its fauna have indicated that species diversity is quite high, with many faunal elements being unique to the region (Mello-Leitao, 1946; Muller, 1973; Fonseca, unpublished data). The pioneer works of Moojen (1952), Vieira (1955), and Cabrera (1957; 1961) provide a general data base indicating that there are at least 129 species of non-volant mammals in the Atlantic forest region, about 40 % of which are endemic. Were the taxonomy of most of the groups better understood, other species would probably be described. There are at least 23 marsupial




8
and 57 rodent species described for the region, of which 39 % and 53 Z, respectively, are endemic to the region.
Because of past and present habitat destruction, the fauna is increasingly isolated into small patches, and many members of this unique ecosystem are now endangered. This situation has been documented for the most conspicuous elements of the mammalian fauna, the primates (Coimbra-Filho and Mittermeier, 1977; Mittermeier et al., 1982; Fonseca, 1985b). Nonetheless, the rich and highly endemic small mammal fauna of the Brazilian Atlantic forest region has been the subject of very few long-term studies, especially when compared to the Amazon region (e.g., Pine, 1973; Lovejoy et al., 1984; 1986; Terborgh et al., 1984; Malcolm, 1987) or the Cerrado (Alho, 1981; Alho et al., 1986; Fonseca and Redford, 1984; Lacher et al., in press; Nitikman and Mares, 1987). The only detailed study of Atlantic forest small mammals is now over 40-years-old (Davis, 1946).
Historically, the Brazilian Atlantic forest extended from the
coast to the eastern and portions of the western slopes of the coastal mountains (Hueck, 1972; Alonso, 1977), an area of approximately 700,000 square kilometers. Where rainfall permits the presence of tall, evergreens, this type of forest extends into the western slopes of the coastal formation. Due to a rain shadow, the vegetation of the western slopes, where this study was conducted, possesses a number of deciduous tree species, which lose their leaves during the approximately 6 months of the dry season. During the wet season, however, the forests of both




the eastern and western regions are physiognomically undistinguishable. In addition, the faunal elements are mostly the same, generally belonging to the same biogeographical region (Muller, 1973).
The objective of this study was to investigate several aspects of the biology and natural history of eastern Brazilian non-volant small mammals. During a period of 17 months, data on the general aspects of population dynamics, breeding, substrate use and movement patterns of small mammals (marsupials and rodents) were collected in six forest plots, at three main sites in the state of Minas Gerals, Brazil. Morphometric data were also collected. Species composition, relative abundances and other community structure patterns are presented elsewhere (Chapters 3 and 4).
Materials and Methods
Study Sites
Small mammal communities of six forest plots, two at each of three sites in the state of Minas Gerais, were the subjects of this study (Figure 2-1). At each site, a primary and a secondary forest plots were selected. A set of three parallel transect trapping lines was established in each. The primary forest plots showed vertical stratification, with an average canopy height of 19 meters. The herbaceous stratum was somewhat sparse, while the midstory was




Figure 2-1. Location of study sites: (1) Fazenda Esmeralda, Rio Casca county; (2) Parque Estadual Florestal do Rio Doce, Marileia county; (3) Fazenda Montes Claros, Caratinga and Ipanema counties.




generally well developed. The secondary forests, on the other hand, were mostly in their mid-stages of succession. Average canopy height was approximately 12 meters; herbaceous cover was frequently extensive, with masses of tangled vines being a common occurrence. Epiphytes and emergents were conspicuously absent from secondary forests.
The first site was Fazenda Esmeralda, located on the county of Rio Casca. The farm is under extensive agricultural use, with very little remaining under forest cover. Located along the plains of the Rio Doce River, the farm was covered almost completely by pristine forest as recently as 1964. Wood extracting rights were then sold to the largest steel industry of the state of Minas Gerais and by 1970 most of the farm was deforested. Forest patches located on the top of two hills were selected for the study, one with a 60 ha. second growth, and the other with 80 ha. of primary forest, known locally as "Lagoa Fria".
Fazenda Montes Claros was selected as the second site for this study. It is a coffee and cattle farm located within the counties of Ipanema and Caratinga. The total area of the farm is about 1,200 ha., 860 ha. of which remain under forest cover. A research station under the administration of the Brazilian Foundation for Conservation of Nature (FBCN) and the Federal University of Minas Gerais (UFMG) was established in the farm in 1983. A second growth forest patch at Fazenda Montes Claros, known as "Jao" was selected, and another under primary forest, "Matao," was also used.




The third site was the Rio Doce State Park, which has portions of its area within the county of Marilela. The park, with its 35,973 ha., constitutes the largest continuous area under tropical forest in the state of Minas Gerais. It was created in 1944 and has been ever since under the jurisdiction of the Government of the state of Minas Gerais. Since the creation of the State of Minas Gerais Forest Institute (IEF) in the late 1960s, the park has been under its administration. It has recently become one of the best maintained and protected areas under the Brazilian Parks system. Because of several extensive fires in the 1960s, a considerable area of the park is second growth. One of these, "Mata do Hotel" was selected as a study area. A pristine primary forest, known locally as "Campolina," constituted the second patch selected for study within the Rio Doce Park.
Climate
Climatological data were collected at all three sites, but
because of the extreme similarity in temperature and rainfall regimes among study sites, only information collected at Fazenda Montes Claros is presented here.
Hueck (1972) states that rainfall for the western slopes of the Atlantic forest region is always below 1,600 mm annually. In some areas, it can achieve a little over 1,000 mm. During the study period, the region experienced an unusually dry period, as we can observe in the climatogram of Walter (1971) constructed with data collected at




13
Fazenda Montes Claros (Figure 2-2): total rainfall for the first 12 months of trapping was 850 mm, while for the last 12 months rainfall totaled 931 mun. Rainfall is highly seasonal, being concentrated between the months of September and February (Figure 2-2). Average monthly precipitation for the rainy season was 128 mm, while dry season rainfall averaged only 30 nun monthly.
Mean minimum annual temperature during the study period was about 18 C, close to the average for the region (Hueck, 1972). Average daily differences between minimum and maximum temperatures are quite constant throughout the year. Rainfall maxima coincide with the warmest months of the year, while winters are usually very dry. For the purpose of this analysis, the dry season is considered to occur between the months of March and August, and the wet season between September and February.
Trapping
Three transect lines 300 meters long were established in each
plot at each of the three sites. These transects were as parallel to each other as possible and separated by 100 meters. Each transect line possessed 16 trapping stations 20 meters apart. Traps were placed in suitable locations within a 3.5 meter radius measured around center of station. A squirrel-size Tomahawk live trap (Tomahawk Live Trap Co., one-door folding trap, size 203, Tomahawk, WI) was placed on the ground at each station. At every other station, a second Tomahawk trap of the




MONTHLY RAINFALL (mm) 300

MEAN TEMPERATURE ( C)

250 200
150 100 50 0un ul AugSe
Jun Jul AugSe

S
0

120

RAINFALL SURPLUS

DROUGHT

6O
45
0

0
pOctNovDecJanFebMarAprMayJun Jul AugSepOct
MONTHS

Figure 2-2. Climate diagram of Walter (1971) for the forests of the western slopes of the Atlantic forest, built using data collected at Fazenda Montes Claros between June, 1985 and October, 1986. The diagram indicates periods of drought and of water surplus. The dry season can be defined between the months of March and August. Temperatures are given in oC, and rainfall in mm.




15
same size was wired either to a branch or vine at heights from 1 to 4 meters high. In addition to these traps, every other station possessed a mouse-sized collapsible Sherman trap (H. B. Sherman Traps, Inc., Tallahassee, FL), with alternation of ground and tree traps. Moreover, the two outermost transect lines had, at every other station, a large 80 x 30 x 30 centimeters wire home-crafted live trap. Therefore, each outer line possessed 16 ground Tomahawk traps, 8 tree-bound Tomahawk traps, 4 ground and 4 tree-bound Sherman traps, and 8 ground large wire traps. The total for each outer line was 40 traps. The mid-line did not have large traps, but a total of 24 Tomahawk and 8 Sherman traps. In summary, each forest plot had 48 trap stations disposed into 3 transects, of 16 stations each, and a total of 112 permanent based traps (Table 2-1). With the exception of Sherman traps, all traps were closed at the end of each five night trapping session and left in place. Sherman traps were removed, washed, and replaced each month.
Trapping took place between June of 1985 and October of 1986,
i.e., for 17 consecutive months. Each forest plot was trapped for five consecutive nights every month. During the course of the study, a total of 9,520 trap nights were accumulated for each plot. Therefore, trapping effort for all six forest plots together totaled 57,120 trap nights.
Fresh pineapples, oatmeal and a cotton ball soaked with a
commercial codfish oil solution were used as baits. Traps were checked




Table 2-1. Schematic representation of the first four stations in transect trap lines. Additional stations are replicates of the first four. Each trapline is 300 meters long, separated from each other by 100 meters. Stations are disposed within each line at every 20 meters.

LINE ONE
#1
-Ground Tomahawk
-Above-Ground
Tomahawk

--100m--

LINE TWO
#1
-Ground Tomahawk
-Ground Sherman

--100m--

LINE THREE
#1
-Ground Tomahawk
-Above-Ground Tomahawk

--20m--

#2
-Ground Tomahawk
-Large Trap
-Above-Ground
Sherman

#2
-Ground Tomahawk
-Above-Ground Tomahawk

#2
-Ground Tomahawk
-Large Trap
-Above-Ground Sherman

--20m--

#3
-Ground
Tomahawk
-Above-Ground Tomahawk

#3
-Ground Tomahawk
-Above-Ground Sherman

#3
-Ground Tomahawk
-Above-Ground
Tomahawk

--20m--

#4
-Ground Tomahawk
-Large Trap
-Ground
Sherman

#4
-Ground
Tomahawk
-Above-Ground Tomahawk

#4
-Ground Tomahawk
-Large Trap
-Ground
Sherman

#16

#16

#16




every morning for captures and for adequacy of bait, which was replaced as needed.
For each individual, the following information was recorded: 1. Species. If not readily identifiable, the individual was preserved for later taxonomic identification. The only individuals which were consistently prepared as skins and skulls were representative of Orvzomys trinitatis and Oryzomys nigrirpes. The former shows a high degree of variability in skin color and pattern and was considered, during field work, as belonging to more than one species. Whenever possible, animals that died in the traps were also preserved. Moreover, live trapping and snap-trapping were conducted on other locations within the same forest type for preparation of voucher specimens and determination of stomach contents. Taxonomic identification of voucher specimens was provided through the kindness of the following specialists: Drs. M. Carleton, C. Creighton, L. Emmons, G. Musser, P. Myers, and J. Patton.
2. Location in grid.
3. Individual identification. If already tagged with metal fish tags (Fish and small animal tag, size 1, National Band and Tag Co., Newport, KY), the animal was promptly released; otherwise the animal was tagged. Individuals with positive taxonomic identification were released at the same station where captured.
4. Sex.
5. Weight.




6. Body length.
7. Tail length.
8. Ear length.
8. Hind foot length.
9. Reproductive condition. Female rodents were checked for
perforated vaginas. I also noticed whether individuals were pregnant (in late stages) or lactating. Male rodents were considered in breeding condition if testes were descended. Female marsupials were considered in breeding condition if neonates were found in either pouch or attached to nipples, or if the nipple area showed signs of recent nursing by previous litter. Male marsupials with testes of reduced size were considered pre-pubertal, while for some species (Metachirus nudicaudatus, Marmosa incana, Marmosa cinerea, Marmosa agilis, and Philander opossum) abdominal or sternal scent gland activity indicated breeding condition. We also recorded any pouch young old enough to be sexed.
10. Behavior upon release.
Analysis of variance (Procedure GLM: PC-SAS, SAS Institute, Cary, NC) was used to test the null hypothesis of absence of sexual dimorphism, using the following morphological characters: weight, body length, tail length, hind foot length, and ear length. Kruskal-Wallis tests were performed to test for differences in persistence times in traplines among sexes.




Since the number of arboreal traps in each set of three traplines (36; 32 %) was different from that of terrestrial ones (76; 68 ), analysis of differential trapping success of arboreal versus terrestrial traps was conducted with corrections to account for the differential probability of trapping in the different strata. Where sample size permitted, chi-square tests were performed to test the null hypothesis of lack of difference in substrate use among sexes. For this analysis, all proportions were arc-sin transformed (Sokal and Rohlf, 1981).
Results and Species Accounts
During the course of this study, a total of 692 individuals of 8 marsupial and 11 rodent species were caught 1,366 times, in all six forest plots (Table 2-2). One individual of the species Monodelphis americana was trapped on one of the auxiliary lines at the primary forest of Fazenda Montes Claros. Mean trapping success, averaged over 17 months of trapping at all sites, was 2.4 %. For most species, trapping success was much higher during the cold, dry months, dropping sharply around October with the onset of rains.
Caluromys philander. Woolly opossums were only registered for the Rio Doce State Park. Although occasionally trapped on the ground, this species is mostly arboreal (Davis, 1947; Nowak and Paradiso,




20
1983). In this study, only one individual was captured on the ground. Upon release, all individuals climbed into canopies, disappearing from sight at heights of about 10 meters. Both the small number of arboreal traps (32 %), relative to terrestrial ones, and the absence of traps in higher strata of the forest, may have accounted for the low trapping success for this species. Only three males and three females were captured during the course of the study. One female was recaptured once. Morphometric data on these individuals are presented in Table 2-3.
Didelphis marsupialis. Opossums occurred at all sites and represented the most common small mammal species captured in the majority of forest plots. This is a generalist species, found in all habitat types (Hunsaker, 1977; Miles et al., 1981; Alho et al., 1986). Adult males were significantly heavier than females (Table 2-4). Hind feet were also sexually dimorphic. Other biometric parameters did not vary between sexes.
Opossums have been previously described as being mainly terrestrial, although they are able to climb opportunely while foraging (Miles et al., 1981). These generalizations were confirmed here. Overall, 73 % of D. marsupialis captures were from terrestrial traps, while 93 % of individuals released also remained on the ground. However, young opossums tend to use aerial




Table 2-2. Trapping results for all six forest plots together during 17 months of trapping.
SPECIES Number of Total Number of
First Captures Captures
Caluromys philander 6 7
Didelphis marsupialis 144 404
Marmosa agilis 7 11
Marmosa cinerea 43 120
Marmosa incana 127 220
Marmosa microtarsus 1 1
Metachirus nudicaudatus 105 156
Philander opossum 14 29
Abrawayomys ruschi 1 1
Akodon cursor 29 53
Nectomys squamipes 10 14
Oryzomys capito 1 2
Oryzomys nigripes 5 5
Oryzomys subflavus 1 1
Oryzomys trinitatis 84 99
Oxymycterus roberti 2 2
Rhipidomys mastacalis 1 1
Echymys sp. 1 1
Proechimys setosus 110 239
TOTAL 692 1366




Table 2-3. Biometric Caluromys philander.
measurements in mm.

data for three males and three females of Weight is given in gms, and other body

MALE

FEMALE

Weight

Body length Tail length Hind Foot Ear length

Mean St.Dev.
171.3 57.6 210.7 30.0 301.0 25.9

36.3

Mean St.Dev. 180.5 40.5 206.8 11.4 298.8 20.7

36.3 2.6

33.3 1.5 30.5 1.9




23
Table 2-4. Biometric data for adult male and female Didelphis marsupialis. Weight is given in gms, and other body measurements in mm. N=number of individuals, St. Dev.=Standard Deviation. P refers to p-value associated with analysis of variance for sexual dimorphism (n.s.=p>0.05;*=p<0.05; **=p<0.01; ***=p<0.001).
MALE FEMALE

Weight Body length Tail length Hind Foot Ear length

Mean 1,024.0 354.0 348.5 58.1 51.2

St. Dev. 361.4 40.4 30.9
3.9
4.2

N
63 63 62 62 62

Mean 946.0 354.0 349.6 56.1 50.0

St. Dev.
157.0 29.7
27.4
4.3 3.6

N
34 34 31
34 34

P
**
n.s.
n.s.
n**s.
n.s.




Frequency

0 1 2 3 4 5 6 7 8 9 10 11 12 Number of months in lines Males -- Females
Figure 2-3. Frequency of individuals of D. marsupialis persisting within traplines.




25
substrate more often than adults. Approximately 23 % of the individuals released (11 out of 37) climbed into trees and/or onto vines. The same result was obtained by Davis (1947). There were no significant differences in substrate use between adult males and females.
A total of 79 males and 64 females were captured 196 and 204 times, respectively, with sex ratios of both number of individuals and total captures deviating very little from 1:1. Based upon recaptures, males did not persist for long periods on grids, averaging 1.74 months. Females stayed within the traplines an average of 2.6 months, but this difference is not significant (p=0.14; Figure 2-3). Record persistence times were achieved by one female and one male, which were recorded for a total of 12 months. Fleming (1972) found similar low persistence times for Didelphis marsupialis in Panama. If the data from Sunquist et al. (1987) with D. marsupialis from Venezuela is extrapolated to the present study, the low persistence times can probably be attributed to high mortality rates rather than to dispersal.
The distance travelled between successive captures were not significantly different between males and females. The overall mean was 143.4 meters (maximum=400 meters), slightly higher than the figure given by Fleming (1972). However, similar to grid trapping methods, traplines underestimate actual distances




26
travelled in a night, which was found by Sunquist et al. (1987), using radio transmitters, to average one kilometer.
Even though trapping success during the first three months of study was very high, population levels, as reflected by the number of captures, did not appear to vary between dry and rainy seasons (Figure 2-4). Reproductive activity, on the other hand, was highly seasonal, with most females breeding just prior and well into the rainy season. The number of juveniles recorded also followed this same pattern (Figure 2-5). Two females were recorded as having two consecutive litters, the first in August, and the second in October. The same has been observed by Collins in Brazil (1973, in Nowak and Paradiso, 1983). The occurrence of two consecutive litters has also been reported elsewhere for 0. marsupialis in Colombia (Tyndale-Biscoe and Mackenzie, 1976) and southeastern Brazil (Davis, 1946).
An interesting result was the decrease in opossum abundances during the second years's late dry season, especially in the small forest plots. Smaller plots are where D. marsupialis was most found to be most abundant during the first year of trapping, and capture success was much lower than for the previous year (Figure 2-4). The reason for this apparent population crash is not yet clear, but may be due to a reduction in food resources that occurred as a result of a severe dry season. Although the reduction In abundance was observed across all plots, smaller




100 1
80
60
40
20
Jun Jul AugSepOctNovDecJanFeoMarAprMayJun Jul AugSepOct
SNumber of Captures M Rainy Season Limits
Figure 2-4. Number of captures of D. marsupialis according to monthly trapping period. Straight lines enclose boundaries of rainy season (September to February).




Jun Jul AugSepOctNovDecJanFebMarAprMayJun Jul AugSepOct =Breeding Females MINumber of Juveniles -X--Mean Rainfall (cm)

Figure 2-5. Number of breeding females and juveniles of D. marsupialis, according to monthly trapping period. Line represents mean daily rainfall (mm).




29
forests experienced the highest decrease, lending support to this hypothesis, since populations inhabiting smaller plots should be more susceptible to a decrease in food resources.
Average litter size of pouch young was 8.6, with a M:F sex ratio of 1.7:1.0, highly skewed towards males (Table 2-5). Overall, two thirds of females produced male biased litters. This average litter size is much higher than those obtained in Colombia by Tyndale-Biscoe and Mackenzie (1976) or by Fleming (1973) in Panama. Both studies reported averages of approximately 6.5 young. In addition, these studies found that sex ratios of pouch young did not differ from a 1:1 ratio. Austad and Sunquist (1986) were able to experimentally induce male-biased sex ratios in D. marsupialis by providing dietary supplementation. Therefore, relative to the sites studied by Fleming (1973) and Tydale-Biscoe and Mackenzie (1976), the forests sampled in the present research may be more productive for common opossums, which may have resulted in the observed skewed sex ratios.
Marmosa agilis. This small arboreal marsupial can be fairly common in gallery forests of the Cerrado region (Nitikman and Mares, 1987), but was infrequently caught in this study. The first individual of the species was recorded only after 12 months of trapping. In total, three males and four females were caught 11 times in two forest plots. Ten of the eleven captures (91 %) were




Table 2-5. Litter sizes and pouch young.

sex ratios of Didelphis marsupialis

ID Number Number of Number of Total Sex Ratio
Female Young Male Young Young M:F
0076 9
0088 9
0055 7
0072 9
0087 9
0126* 3
0156 7
0162* 4
0067 7
0192 7
0193 10
0209 3 6 9 2.0:1.0
0072** 9
0081 11
0220 8
0251 8
0257 8
0258 8
0274 3 5 8 1.7:1.0
0209** 2 7 9 3.5:1.0
0278 5
0281 3 4 7 1.3:1.0
0227 5 5 10 1.0:1.0
0480 5 5 10 1.0:1.0
14/10 2 7 9 3.5:1.0
42P 2 5 7 2.5:1.0
11/12 4 4 8 1.0:1.0
Means 8.6 1.7:1.0

injuries. calculating means.

*Females were in poor condition, with multiple Therefore, their litter sizes were not used in
**Second consecutive litter.




31
on traps placed on trees or vines, which is a capture rate on arboreal traps similar to that obtained by Nitikman and Mares (1987). Alho et al. (1986) also found M. agilis to be preferentially arboreal in gallery forest habitat.
Except for two males and one female, which were recorded in
traplines for two months, no other animal was recaptured in subsequent trapping sessions. The only individual for which data on movement between trapping stations were obtained travelled 40 meters in consecutive nights, a figure very close to that determined by Nitikman and Mares (1987). Biometric data on Marmosa agilis are presented in Table 2-6.
Marmosa cinerea. This species is one of the largest within the genus (Nowak and Paradiso, 1983) and was fairly common at the Rio Doce Park, but was conspicuously absent from both Fazenda Montes Claros and Fazenda Esmeralda. Most individuals were captured during the cold, dry months (Figure 2-6), with trapping success during the rainy season being fairly low. Lactating females, on the other hand, were only trapped during the rainy season, indicating that the high trapping success of the following months reflected the addition of recently born young into the population.
A total of 8 males and 35 females were caught 120 times
(Table 2-2), making it the species with the highest recapture rate in




32
Table 2-6. Biometric data for four males and one female Marmosa agilis. Weight is given in gms, and other body measurements in mm.
MALE FEMALE
Mean St.Dev. N Mean St.Dev. N
Weight 29.5 11.3 4 25.0 1
Body length 109.3 13.2 4 99.0 1
Tail length 155.3 7.1 4 151.0 1
Hind Foot 18.3 1.5 4 17.0 1
Ear length 20.5 0.6 4 21.0 1




Jun Jul AugSepOctNovDecJanFebMar AprMayJun Jul AugSepOct

SNumber of Captures

M Rainy Season Limits

Figure 2-6. Number of captures of M. cinerea according to monthly trapping period. Straight lines enclose boundaries of rainy season (September to February).




Differential Trapping Success (in %)

Males Females

EMTerrestrial Traps

=Arboreal Traps

Figure 2-7. Differential trapping success (in percentages) of M. cinerea in arboreal and terrestrial traps.




35
this study (31 X for males and 37 % for females). While sex ratio for first captures was 1.6:1.0, total captures approximated a 1.0:1.0 ratio. This might indicate that females are captured more frequently than males, which may be due to the fact that males are more transient than females (chi-square=5.17; df=l, p 1.8 months in traplines. Females, on the other hand, persisted in the areas for an average of 3.8 months. The longest tenancy in traplines was also by a female that was recorded present for 14 months. In addition, males also travelled farther between successive captures than females (males=142 meters, maximum:380 meters; females=80 meters, maximum:180 meters), which also supports the contention that transient males are a frequent occurrence.
On average, adult males were found to be slightly heavier than females. However, the species did not show sexual dimorphism for any biometric parameter measured (Table 2-7). Although trapped frequently on the ground, M. cinerea appeared to be primarily arboreal, especially the females (Figure 2-7). The behavior upon release also revealed that, while 56 % of the males remained on the ground, only 27 % of the females displayed a similar behavior. Terborgh et al. (1984) list M. cinerea as being able to exploit several forest strata, from understory to canopy. Miles et al. (1981) determined M. cinerea to be mostly arboreal, with all nests also located above ground.




Table 2-7. Biometric data for adult male and female Marmosa cinerea. Weight is given in gms, and other body measurements in mm. P refers to p-value associated with analysis of variance for sexual dimorphism (n.s.=p>0.05; *=p<0.05; **=p<0.01; ***=p<0.001).
MALE FEMALE

Weight Body length Tail length Hind Foot Ear length

Mean
117.3
176.1 258.9 28.9
30.4

St. Dev.
29.4 20.0 26.0 3.9 2.8

N
24 24 24 24 24

Mean
99.7 177.1 261.7 28.0 29.7

St.Dev.
16.7
20.5 17.9 3.1
2.1

N 11 11 11 11 11

P
n.s.
n.s.
n.s.
n.s.
n.s.




Frequency 60
50 40 30 20 100
0 1 2 3 4 5 6 7 8
Number of months in ines Males Females
Figure 2-8. Frequency of individuals of M. incana persisting within traplines.




38
Marmosa incana. This was the second most frequently caught species in this study. It was present in all six forest plots sampled. Eighty males and 44 females were captured a total of 220 times. Three individuals escaped before their sex could be determined. Sex ratio for first captures was 1.8:1.0, and 2.9:1.0 for all captures. Males also persisted on the traplines more than females (chi-square=8.1, df=l, p<0.005), on average 1.9 and 1.2 months, respectively (Figure 2-8). Therefore, in contrast with M. cinerea, male M. incana appear to be more sedentary than females. M. incana males travelled an average of 64.7 meters between successive captures (maximum=200 meters), while the only female recaptured in the same trapping session travelled 40 meters.
Marmosa incana adult males were sexually dimorphic for all morphometric characters measured, with males being, on average, 20 X heavier than adult females (Table 2-8). Body length of males was also 14 % larger than for females. Adult females also lack the sternal gland, which is functional in males during the breeding season.
As with M. cinerea, trapping success increased markedly with the end of the rainy season (Figure 2-9). This population growth is accounted for by the increase in the number of juveniles trapped (Figure 2-10). The hypothesis that increased trapping success is due to juvenile recruitment is further supported by the observation




Table 2-8. Biometric data for adult male and female Marmosa incana. Weight is given in gms, and other body measurements in mm. P refers to p-value associated with analysis of variance for sexual dimorphism (n.s.=p>O.O5;*=p<0.05; **=p
MALE

FEMALE

Mean St.Dev. N

Mean St.Dev. N

Weight Body length Tail length Hind Foot

82.3 23.6 62 68.4 23.1

23 **

150.5 18.0 62 141.0 15.5 23 ***
203.9 18.2 62 191.2 18.2 23 ***

23.4

Ear length 28.1

2.0 61 21.5 3.9 23 2.4 61 26.5 3.0 23 *




Jun Jul AugSepOctNovDecJanFebMar AprMayJun Jul AugSepOct

Number of Captures

Rainy Season Limits

Figure 2-9. Number of captures of M. incana according to monthly trapping period. Straight lines enclose boundaries of rainy season (September to February).




108
6
4
2
0
Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct
= Number of Juveniles E Mean Rainfall (cm)
Figure 2-10. Number of juveniles of M. incana according to monthly trapping period. Line represents mean daily rainfall (m).




Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct

EM reeding Males

=Breeding Females 9 Mean Rainfall (cm)

Figure 2-11. Number of breeding males according to monthly trapping period. rainfall (mm).

and females of M. incana Line represents mean daily




Differential Trapping Success (in %)

Males Females

EMTerrestrial Traps

=Arboreal Traps

Figure 2-12. Differential trapping success (in percentages) of M. incana in arboreal and terrestrial traps.




44
that the occurrence of breeding males and especially breeding females is tightly tied to mid to late rainy season (Figure 2-11).
Marmosa incana can be characterized as a scansorial species, as its use of arboreal and terrestrial traps was approximately evenly split (Figure 2-12). The same result was obtained using behavior upon release as measure of differential substrate use. Approximately 53 % of individuals released climbed trees or vines, while 47 % remained on the ground. There were no statistical differences in substrate use between males and females (chi- square= 3.19; p>O.05).
Marmosa incana appears highly insectivorous. Three stomachs analyzed in this study contained only insects, mostly belonging to the orders Coleoptera and Orthoptera. Nowak and Paradiso (1983) reported that most members of the genus Marmosa are insect and fruit eaters, although vertebrates are also occasionally consumed.
Marmosa microtarsus. During the last month of the survey, October of 1986, an adult male of this species was caught in an arboreal trap in the secondary forest of the Rio Doce Park. The species was previously described as being quite abundant in both secondary and primary forests of the Atlantic forest region (Davis, 1947). Marmosa microtarsus is similar in morphology to Marmosa agilis, and it is usually difficult to distinguish them. Morphometric data from this individual are as follows: weight=31




45
grams; body length:106 millimeters; tail length=148 millimeters; ear length=14 millimeters; hind foot=17 millimeters.
Metachirus nudicaudatus. This relatively large-bodied
terrestrial didelphid was the third most common marsupial species trapped in this study (Table 2-2). The brown four-eyed opossum was present in all six forest plots surveyed. A total of 60 males and 45 females were caught, respectively, 88 and 68 times. The sex ratio for first captures was equal to that of recaptures (1.3:1.0). Adult males were, on average, larger than females, with most biometric parameters measured proving sexually dimorphic (Table 2-9). Males were, on average, 1/4 heavier than adult females.
Even though the species was relatively common throughout the year, a trapping success peak was observed following the rainy season (Figure 2-13). This peak probably coincides with the onset of breeding in mid- rainy season and the beginning of the dry months. Both lactating females and males with functional abdominal glands were frequently caught at this time (Figure 2-14). Metachirus nudicaudatus may also be able to produce a second litter in the same year. One female had a litter in March and the second in October. The number of pouch young ranged from 5 to 9, with an average of 7.2 young (Table 2-10). For the two females in which




46
Table 2-9. Biometric data for adult male and female Metachirus nudicaudatus. Weight is given in gms, and other body measurements in mm. P refers to p-value associated with analysis of variance for sexual dimorphism (n.s.=p>O.05; *=p MALE FEMALE
Mean St.Dev. N Mean St.Dev. N P
Weight 352.7 94.0 51 280.8 57.1 33 **
Body length 252.0 28.7 51 232.1 25.4 32 **
Tail length 324.1 30.5 51 315.9 34.0 31 n.s.
Hind Foot 44.9 3.1 51 42.3 3.3 32 *
Ear length 37.4 3.9 50 36.5 2.9 32 n.s.




Jun Jul AugSepOctNovDecJanFebMar AprMayJun Jul AugSepOct

i Number of Captures

M Rainy Season Limits

Figure 2-13. Number of captures of M. nudicaudatus according to monthly trapping period. Straight lines enclose boundaries of rainy season (September to February).




108
6
4
2
Jun Jul AugSepOctNovDecJanFebMarAprMayJun Jul AugSepOct
Breeding Males Breeding Females e Mean Raintall (cm)
Figure 2-14. Number of breeding males and females of M. nudicaudatus according to trapping period. Line represents mean daily rainfall (mm).




Table 2-10. Litter sizes and pouch young.

sex ratios of Metachirus nudicaudatus

ID Number Number of Number of Total Sex Ratio
Female Young Male Young Young M:F
271 9
322 5
338 8
346 3 2 5 0.7:1.0
338* 6 3 9 0.5:1.0
Means 7.2 0.6:1.0
* Second consecutive litter.




50
young were old enough to be sexed, there was a biased sex ratio towards females.
Nowak and Paradiso (1983) regard brown four-eyed opossums as
being arboreal, but the species was only once in 156 captures caught in an arboreal trap. Furthermore, only once was an individual observed to use aerial substrate upon being released. In fact, the large and non-graspable hind feet and clumsy behavior on above ground support do suggest a complete terrestrial life for the species. Using a spool-line device, Miles et al. (1981) also found Metachirus nudicaudatus to be completely terrestrial. Terborgh et al. (1984) also regard gray four-eyed opossums as a species confined to the ground.
Male and female brown four-eyed opossums did not differ in
persistence times in traplines, with both sex classes averaging 1.7 months. Two males and two females were also recorded in traplines over a nine month period (Figure 2-15). Distances travelled between successive trapping were among the lowest for marsupials, averaging 36.7 meters (maximum=120 meters; N=6) for males and 40 meters (maximum=80 meters; N=T) for females.
Monodelphis americana. Short-tailed opossums have a fairly wide distribution in Brazil (Streilein, 1982). Only one female was caught in an auxiliary line established in the primary forest at Fazenda Montes Claros, for the purpose of collecting voucher




Frequency

0 2 4 6 8
Number of months in lines

Males Females
Figure 2-15. Frequency of individuals of M. nudicaudatus persisting within traplines.




52
specimens. The capture site did not differ physiognomically from the regular traplines, and therefore the species might also occur consistently at this site. However, it is felt that trapping methods were not proper to adequately represent the species. Its small body size and foraging habits on the forest litter may have accounted for its low representation in the sample (Davis, 1947). The biometric data on this individual are as follows: weight=19 grams; body length=92 millimeters; tail length=46 millimeters; hindfoot=16 millimeters; ear length=13 millimeters. Comparisons with data provided by Nowak and Paradiso (1983) indicate that this individual was probably a juvenile.
Philander opossum. Gray four-eyed opossums were trapped in all forest plots, except at the Rio Doce Park. Fourteen individuals were caught during the course of this study. The occurrence of this large dldelphld marsupial is apparently tied to the presence of standing or running water (Davis, 1947; Handley, 1976; Nowak and Paradiso, 1983; Alho et al., 1986). As only a few transects occurred close to streams, this may explain the low trapping success for this species. As with other species, trapping success was much higher during the dry season (Figure 2-16).
While only three individual females were recorded, these had a much higher recapture rate; 11 males were caught 17 times, while
3 females were captured In 12 different occasions. The longest




Jun Jul AugSepOctNovDecJanFebMar AprMayJun Jul AugSepOct
= Number of Captures M Rainy Season Limits
Figure 2-16. Number of captures of P. opossum according to monthly trapping period. Straight line encloses boundaries of rainy season (September to February).




54
persistence times on the traplines were 4 and 5 months, achieved by two adult females.
All females caught were lactating, and two still had pouch young. Nowak and Paradlso (1983), based on several studies, concluded that P. opossum breeds aseasonally. However, the small sample size of the present study does not allow any conclusions as to seasonal ity of reproduction. One female had pouch young in February, while the remaining were caught lactating in August and September. Each had 5 young attached to the nipples. The figure given by Davis (1947) in the Atlantic forest was an average number of pouch young of 4.5, with a maximum of seven, while litter sizes in Nicaragua were found to be slightly larger (Phillips and Jones, 1965). For one litter which sex could be determined, there were four female and one male young.
Even though the small sample size precluded statistical
analysis of sexual dimorphism, adult males were on average over 30 % heavier than females (Table 2-11). As noticed by Nowak and Paradiso (1983), in this study the species also proved to be primarily terrestrial, with only 17 % of captures in arboreal traps, and 7 % of individuals climbing trees after being released. Based on field observations, however, it is felt that gray four-eyed opossums are able, if needed, to efficiently make use of arboreal substrate. The same was suggested by Miles et al. (1981) and Crespo (1982).




Table 2-11. Biometric data for seven males and three females of gray four-eyed opossum, Philander opossum. Weight is given in gms, and other body measurements in mm.
MALE FEMALE
Mean St.Dev. Mean St.Dev.
Weight 394.9 98.7 295.0 39.7
Body length 282.0 51.3 285.0 48.2
Tall length 325.5 21.6 295.0 18.0
Hind Foot 45.3 2.6 40.3 4.0
Ear length 32.7 3.0 29.3 3.1




Table 2-12. Biometric data for adult Akodon cursor. Weight is given in gms, and other body measurements in mm. P refers to pvalue associated with analysis of variance for sexual dimorphism (n.s.=p>0.05; *=p<0.05; **t=p<0.01; ***=p<0.O01).

MALE

FEMALE

Mean St.Dev. N

Mean St.Dev. N

Weight Body length Tail length Hind Foot

48.1

107.8 10.1

98.8 26.9

9.9 1.3

14 45.5 6.8 13 103.0 14.3 12 94.9 10.4 13 25.9 1.3

1.6 14 18.8 0.9

8 n.s
8 n.s.
8 n.s.
8 n.s.
8 n.s.

Ear length 18.1




57
Abrawayomys ruschi. This rare monotypic murid rodent, endemic to the Atlantic forest of eastern Brazil, is only known through its type specimen, collected in the state of Espirito Santo (Nowak and Paradiso, 1983). A single adult male, with descended testes, was collected at the secondary forest of the Rio Doce State Park in January of 1986, for which measurements are as follows: weight=63 grams; body=128 millimeters; tail length=146 millimeters; hind foot=31 millimeters; ear length=20 millimeters.
Akodon cursor. The genus Akodon comprises over 40 species, and A. cursor is among the largest. Sexual dimorphism is lacking in A. cursor (Table 2-12), although males weighted more than females in another study (Nitikman and Mares, 1987). This species occurred frequently in the Rio Doce Park secondary forest, but was also occasionally caught at other sites, especially in some auxiliary lines which were located in humid grasslands. Within the traplines, a total of 29 individuals were caught 53 times (Table 2-2), 27 of which were trapped at the Rio Doce Park. Sex ratios for first captures and all captures were, respectively, 2.3:1.0 and 3.4:1.0.
Akodon cursor was previously described as being completely terrestrial (Crespo, 1982; Alho et al., 1986; Nitikman and Mares, 1987). However, the species demonstrated scansorial ability in this study, with approximately one-third of trapping success being obtained at arboreal traps.




58
A distinct peak in population density was found between May and July. Very few individuals were recorded during the rainy season (Figure 2-17). This was attributed to recruitment of young into the population just at the end of the wet and in the dry season. This conclusion is supported by the observation that the number of males with descended testes closely follows that of trapping success, and coincides with mid dry season (Figure 2-18). Gestation and subsequent weaning are close to five weeks (Nowak and Paradiso, 1983), and the influence of reproduction activity on population density was readily noticeable in terms of increased trapping success.
Akodon cursor appears to be primarily insectivorous. Individuals trapped in both grasslands and forests turned out high amounts of insects in their stomachs, which possibly indicate that the species makes use of insects in their diets in all habitats. The species also makes use of seeds, fruits, and vegetative parts, especially those of the Graminae (Table 2-13).
Akodon cursor turnover rates appear very high, with 78 % of
individuals only being recorded during one trapping session. Maximum persistence was achieved by two individuals, but for only three and four months. The two individuals for which data on travel distances were available, moved 20 and 40 meters between successive captures.




16
14 12 10
8
6
4
2
0
Jun Jul AugSepOctNovDecJanFebMar AprMayJun Jul AugSepOct
SNumber of Captures M Rainy Season Limits
Figure 2-17. Number of captures of A. cursor according to monthly trapping period. Straight line encloses boundaries of rainy season (September to February).




10
8
6
4
2
0
Jun Jul AugSepOct NovDec Jan Feb Mar Apr May Jun Jul Aug Sep Oct
EM Breeding Males e Mean Rainfall (cm)
Figure 2-18. Number of males of A. cursor with descended testes, according to trapping period. Line represents mean daily rainfall (mm).




61
Table 2-13. Stomach contents of four Akodon cursor.
PERCENTAGE OF
ID Number Insects Leafy Material Seeds Fruits
CEM1/10* 33 33 0 33
47P** 100 0 0 0
101P* 100*** 0 0 0
106P* 96*** 0 4 0
$ Forest traplines.
** Grasslands.
*** Coleoptera and ants.




62
Echymys sp. Members of this genus are entirely arboreal
(Hershkovitz, 1969). This may explain the low success in recording this species in the forests sampled. It is possible that it preferentially occupies higher canopy strata. Davis (1947) never trapped Echymys below 5 meters, while Miles et al. (1981) found nests of Echymys chrysurus in tree cavities located at the canopy level.
Only one adult male was caught on the primary forest of Fazenda Montes Claros in July of 1986, in an arboreal trap. The voucher specimen could not yet be identified beyond genus, but may possibly either be Echymys brasiliensis (sensu Moojen, 1952) or a species which has not been described (L. Emmons, pers. comm., 1988). This specimen had the following measurements: weight=225 grams; body length=215 millimeters; tail length=205 millimeters; hind foot=38 millimeters; ear length=14 millimeters.
Nectomys squamipes. This semi-aquatic rat was occasionally trapped in the vicinities of small streams or flooded areas within forests. Alho et al. (1986) obtained 93 % of all the captures of Nectomys squamipes in the Cerrado region on flooded areas. A few individuals in the present study were found over 500 meters from any source of water, suggesting that the species may occasionally exploit non-aquatic habitats. The three stomachs analyzed contained only vegetative material, two exclusively fruit pulp.




63
Nectomys squamipes is a fairly large-bodied rodent, with males attaining over 250 grams (Table 2-14). A total of eight males and one female were recorded in tree plots, one at each of the study sites. Another individual escaped before sex could be determined. All but one individual was caught during the mid-dry season, i.e., between May and August.
Although adapted for semi-aquatic life, N. squamipes were trapped 38 % of the times in arboreal traps. Nests which were assumed to belong to N. squamipes were found one meter above ground. Individuals were also captured at stations away from water sources. Two males persisted in the areas for three and four months, respectively.
Oryzomys capito. A widespread habitat generalist, this rodent is extensively distributed in South America (Handley, 1976). It is possibly mostly terrestrial, and in the Cerrado region was found more commonly at dense forests (Alho et al., 1986; Nitikman and Mares, 1987). However, it was a rare species in the forests sampled in this study and was represented by only one male and one female at the Rio Doce Park, both caught in September of 1986. All captures were on ground traps. The measurements of these individuals are, respectively: weight=60 and 63 grams; bodylength=122 and 132 millimeters; tail length=115 and 129 millimeters; hind foot= 32 and 34 millimeters; ear length=22 and 21 millimeters.




Table 2-14. Biometric data for five squamipes. Weight is given in gms,

MALE

males and one female Nectomys and other body measurements in mm.

FEMALE

Mean St.Dev. N

Mean St.Dev. N

Weight

Body length Tail length Hind Foot

197.0 67.0 188.4 19.9 207.8 24.9

50.0

Ear length 22.0

2.8

5 165 5 190 5 210 5 50

2.2 5 22

- 1
- 1
- 1
- 1
- 1




65
Oryzomys nigripes. This small cricetid rodent commonly occurred at auxiliary traplines located in grasslands and it has wide distribution among Cerrado habitats (Alho et al., 1986; Nitikman and Mares, 1987). It was, however, relatively uncommon in forests. Only 5 individuals, two males and three females, were caught in forest traplines. All trappings were done on mid- to late rainy season, i.e., between January and April. Two of these captures were on arboreal traps, although it is felt that the species is certainly more terrestrial and/or scansorial, as it was also observed by Crespo (1982). Alho and Pereira (1985) in Cerrado gallery forests and Veiga-Borgeaud (1982) in the Atlantic forest region determined that Oryzomys nigripes (=eliurus) makes extensive use of low shrubs, with nests located about 1 meter from ground.
Stomach contents revealed high frequency of insects, complemented by fruit, seeds and leafy material (Table 2-15). Barlow (1969, in Dalby, 1975), and Crespo (1982) also found insects as part of O.nigripes' diet, even though in lower proportions. Biometric data are presented in Table 2-16.
Reproduction occurs throughout the year, albeit it may increase in frequency at some periods. Two females, one caught in February and the other in August, both had 4 fetuses, coinciding with two major reproductive peaks observed by Veiga-Borgeaud (1982). A third female, trapped in April, had 5 implanted fetuses.




66
Table 2-15. Stomach contents of Oryzomys nigripes.
PERCENTAGE OF
ID Number Insects Fruit Leafy Material Seeds 105P 100 0 0 0
56P 20 0 0 80
CEM3 6 90 4 0
CEM2 30 12 58 0




67
Table 2-16. Biometric data for two males and three females of Oryzomys nigripes. Weight is given in gms, and other body measurements in mm.
MALE FEMALE
Mean St.Dev. N Mean St.Dev. N
Weight 19.5 0.7 2 18.3 2.9 3
Body length 87.5 0.7 2 83.0 5.3 3
Tail length 126.0 4.2 2 116.8 6.1 3
Hind Foot 23.5 0.7 2 24.0 1.0 3
Ear length 16.0 0.0 2 17.3 0.6 3




Oryzomys subflavus. Only one female of this otherwise fairly common Cerrado species (Melo, 1977; Alho and Pereira, 1985; Alho et al., 1986) was caught in an arboreal trap in the forest traplines. The measurements of this individual are as follows: weight=92 grams; body length=165 millimeters; tail length=173 millimeters; hind foot=34 millimeters; ear length=25 millimeters.
Oryzomys trinitatis (=concolor). This cricetid was the most frequent rodent found in the forests sampled, being present at all sites. It is widely distributed, from Costa Rica to Paraguay (Nowak and Paradiso, 1983), and found mostly in forested habitats (Alho and Pereira, 1985; Nitikman and Mares, 1987). A total of 84 individuals were captured 99 times (Table 2-2). Due to coat color variation, most individuals were prepared as skins. Therefore, recapture figures underestimate true recapture rates. Sex ratios at first captures were 1.35:1.00. In contrast to other small mammal species observed in this study, females were slightly heavier and longer than males. Other biometric parameters were found not to be significantly dimorphic (Table 2-17).
The species appears to be mostly arboreal. Approximately 74 % of males and 62 % of females were captured in arboreal traps (Figure 2-19). This figure is very close to that observed by Nitikman and Mares (1987) in a gallery forest of Central Brazil.




Table 2-17. Biometric data for adult male and female Oryzomys trinitatis. Weight is given in gms, and other body measurements in mm. P refers to p-value associated with analysis of variance for sexual dimorphism (n.s.=p>0.05; *=p<0.05; **=p<0.01; ***=p<0.001).

MALE

FEMALE

Mean St.Dev. N

Weight

Body length Tail length Hind Foot

63.2
129.1

Mean St.Dev. N

8.2 35 70.0 10.9 28
8.8 35 136.4 8.1 28

152.3 10.2 35 159.4 18.2 27

30.4

Ear length 18.6

4.1 35 29.1 2.0 28
3.0 35 19.0 1.8 28

n.s.
n.s.
n.s.




Differential Trapping Success (in

Males Females

Terrestrial Traps

= Arboreal Traps

Figure 2-19. Differential trapping success of male and female 0. trinitatis in arboreal and terrestrial traps.




71
Males and females did not differ in degree of arboreality (chi-square=2.40; df=l; p>0.05). Since most individuals were prepared as voucher specimens, sample size was too small to evaluate the differential use of substrate by analyzing behavior upon release. However, the few individuals that were released demonstrated climbing ability. Even though relatively uncommon in the forests of Manu National Park in Peru, Oryzomys trinitatis (=concolor) was listed by Terborgh et al. (1984) as being able to use most all forest strata, from the understory to canopy.
Although there were variations in trapping success at a monthly basis, there does not seem to exist any season characterized by high population, in contrast to most other species (Figure 2-20). This might be a consequence of the absence of seasonal breeding. Reproduction seems to take place throughout the year, as breeding males and females were captured in 14 out of 17 months of study (Figure 2-21).
The two stomachs available for analysis indicated a high
degree of insectivory. The stomach contents of an individual were 100 % insects, while a second had 48 % unidentified insects and coleoptera larvae, and 52 % fruit pulp.
Oxymycterus sp.. This genus includes semi-fossorial
cricetids. The genus is represented in these forests by only one juvenile male and one adult female of an yet undetermined species




Jun Jul AugSepOctNovDecJanFebMarAprMayJun Jul AugSepOct

W Number of Captures =Rainy Season Limits
Figure 2-20. Number of captures of 0. trinitatis according to monthly trapping period. Straight line encloses boundaries of rainy season (September to February).




10
8
6
4
2-7
Jun Jul AugSepOctNovDecJanFec Mar Ar May Jun Jul Aug Se Oct
SBreeding Males = Breeorng Females -- Mean Raintall (cm)
Figure 2-21. Number of breeding male and female Q. trinitatis according to monthly trapping period. Line represents mean daily rainfall (mm).




74
(Table 2-2). Members of this genus are mostly insectivorous (Borchert and Hansen, 1983; Redford, 1984); two stomachs analyzed in the present study yielded 100 % insects, especially Coleoptera larvae and ants. Species of the genus Oxymycterus are more commonly found in grasslands and inundated savannas (Borchert and Hansen, 1983; Fonseca and Redford, 1984; Redford, 1984), a habit which can account for the rarity of the species in the forests of this study. Biometric data for the male and female are, respectively: weight=57 and 85 grams; body length=124 and 245 millimeters; tail length= 111 and 112 millimeters; hind foot=27 and 30 millimeters.
Proechimys setosus. Spiny rats were the third most abundant small mammal in this study (Table 2-2). Only opossums surpassed Proechimys setosus in number of recaptures. Even though species of the genus Proechimys were also found to be quite common in other studies in the Atlantic forest (Davis, 1947; Carvalho, 1965; Avila-Pires and Gouvea, 1977; Botelho and Linardi, 1980; Miles et al., 1981; Fonseca et al., 1987), in the Cerrado (Fonseca and Redford, 1984; Alho et al., 1986) and in the Amazon region (Bishop, 1974; Emmons, 1982; 1984; Terborgh et al., 1984; Malcolm, 1987), spiny rats were surprisingly absent from the Rio Doce Park traplines. This can possibly be attributed to higher predation rate by mammalian carnivores, owls and other predators at the Rio




75
Doce Park (see Chapter 4). The park has a larger and richer carnivore fauna which exert a larger impact upon P. setosus populations than the comparatively depauperated predator community of the smaller forest plots.
Sex ratios of first captures deviated little from a 1:1 sex ratio, with 56 male and 49 female individuals being trapped. Five individuals escaped before their sex could be determined. Females were recaptured at a slightly higher rate than males, yielding a sex ratio for all captures of 0.83:1.00.
Although spiny rats were caught in all 17 months of the
study, a distinct mid to late dry season peak in trapping success was notable (Figure 2-22). This may result from the increase in the recruitment rate of juveniles into the population at this time (Figure 2-23). However, reproduction does not seem to be strictly seasonal, as it was also observed by Bishop (1974) in Mato Grosso, Brazil. Pregnant and lactating females were present in all trapping sessions, although an increase in the frequency of breeding individuals could be observed both in mid-rainy season and in mid-dry season (Figure 2-24). A few females had two litters in the same year, coinciding with the above mentioned breeding peaks.
There is no sexual dimorphism in this species of spiny rat, as none of the morphological characters measured significantly differ between sexes (Table 2-18). Proechimys setosus, is entirely




Table 2-18. Biometric data for adult male and female Proechimys setosus. Weight is given in gms, and other body measurements in mm. N=number of individuals, St. Dev.=Standard Deviation. P refers to pvalue associated with analysis of variance for sexual dimorphism (n.s.=p>O.05;*=p MALE FEMALE
Mean St.Dev. N Mean St.Dev. N P
Weight 270.7 36.2 43 259.6 43.4 41 n.s.
Body length 200.2 14.2 42 196.5 18.9 40 n.s.
Tail length 212.3 13.9 39 208.3 15.4 39 n.s.
Hind Foot 51.2 2.1 43 50.7 4.6 40 n.s.
Ear length 29.2 2.1 43 30.6 2.8 40 n.s.




Table 2-19.

Analysis of seven stomachs of Proechimys setosus.

PERCENTAGE OF
ID Number Fruit Insects Seeds
39P 81 0 19
93P 64 36 0
95P 100 0 0
90P 0 100* 0
91P 100 0 0
92P 100 0 0
79P 47 0 53
* Termites and Coleoptera larvae.




Jun Jul AugSepOctNovDecJanFeoMarAprMayJun Jul AugSepOct

SNumber of Captures

M Rainy Season Limits

Figure 2-22. Number of captures of P. setosus according to monthly Strapping period. Straigh line enclose boundaries of rainy season (September to February).




U.-> I
7f7

'4

Jun Jul AugSepOct NovDec Jan Feb Mar Apr May Jun Jul Aug SepQOct
= Number of Juveniles 9 Mean Rainfall (cm)

Figure 2-23. Number of P. setosus juveniles according to monthly trapping period. Line represents mean daily rainfall (mm).

LA




1614 12 10
8
6
4-1
2
Jun Jul AugSepOctNovDecJan Feb MarAprMayJun Jul AugSepOct
M Breeding Males = Breeding Females -9- Mean Rainfall (cm)
Figure 2-24. Number of breeding male and female P. setosus according to monthly trapping period. Line represents mean daily rainfall (nunmm).




hreauen,.

20-1
0-I
0

1 2 3 5 ,
umber of mont
Males

7 8 9 hs in lines Females

Figure 2-25. Frequency of individuals of P. setosus persisting within traplines.

11 12




82
terrestrial, lacking any scansorial ability. All 239 spiny rat captures were conducted in terrestrial traps. Analysis of stomach contents indicates that spiny rats are primarily frugivorous, but also make opportune use of insects and seeds (Table 2-19). Emmons (1982) found Proechimys to be highly frugivorous, with insects being frequent in stomach samples.
Although, in average, females persisted in traplines more than males (respectively, 2.7 and 2.0 months), there were no significant differences between sexes (chi-square=O.58, df=1, p>O.05). The records for persistence times were obtained by two females who stayed, respectively, 10 and 12 months within traplines (Figure 2-25). Males and females also did not differ in mean travelled distances between successive captures with, respectively, 98.7 (maximum=220 meters) and 87.5 meters (maximum=240 meters).
Rhypidomys mastacalis. This arboreal rat (Hershkovitz, 1969) was only caught once in this study, in the month of September. It is found mostly in moist, forested habitats (Davis, 1947; Dietz, 1983; Fonseca and Redford, 1984; Alho et al., 1986), even though it can also invade households (Nowak and Paradiso, 1983; J. Stallings, pers. comm., 1988). The adult male measurements are as follows: weight=72 grams; body length=137 millimeters; tail length=161 millimeters; hind foot=27 millimeters; ear length=18 millimeters.




Discussion
Seasonality and Resource Use
One of the most striking phenomena common to several small mammals of the western slopes of the Atlantic forest is the occurrence of seasonal reproduction. With the exception of the two most abundant rodents (Oryzomys trinitatis and Proechimys setosus), all four species with sufficient sample size to allow analysis proved to concentrate breeding in the late dry season and into the early to mid-wet season. These were all marsupials. The evidence for these patterns derives from both the increase in trapping success starting in late wet season, represented mostly by the addition of juveniles, as well as the observation of adults in breeding condition just previous to that period. Furthermore, the number of first-time captures of all small mammals is significantly and negatively correlated with average monthly precipitation (r=0.55; p Davis (1946) and Laemmert et al. (1946) observed a pattern of cyclic reproduction among several Brazilian Atlantic forest small mammal species. Breeding was concentrated in the late dry winter




84
and in the wet summer months, albeit less marked than in the present study. The possible reason for this may be that both these studies were in the eastern side of the Brazilian coastal mountains, where the dry season is shorter and less severe (Hueck, 1972). These two studies have also revealed the lack of seasonality for some rodent species, especially ones from the genus Proechimys, which probably breed year-round throughout its range.
A large number of African rodents seem also to closely tie
the onset of reproductive activities to just before the end of the wet season (Delany, 1986). Cerrado species appear to be equally divided between seasonal and year-round breeders (Dietz, 1983; Alho, 1982; Alho and Pereira, 1985). In Panama, over 50 % of mammals are seasonal breeders (Fleming, 1973), although reproduction is concentrated in the three months of the dry season, which results in most young being weaned at the beginning of the rains.
Other things equal, reproductive output and juvenile survival of small mammals would be maximized if reproduction occurs when the environment is best in terms of resource availability (Fleming, 1975b). In fluctuating environments the energetic costs of pregnancy and lactation, coupled with the needs of adequate resources for newly weaned young, should pose constraints on the timing of reproduction. Lee and Cockburn (1985) state that all tropical didelphlds are seasonal breeders and the timing of




85
reproduction is linked with availability of food. Marsupial food resources, in turn, are tied to the fluctuation in rainfall regimes of seasonal environments (Charles-Dominique, 1983). Information on small mammal diets in this study is not sufficient to provide an accurate and longitudinal picture of diet composition. However, it is important to notice that all but the semi-aquatic Nectomys squamipes were observed to consume insects in variable quantities. All didelphid marsupials appear to be predominantly insectivorous (Nowak and Paradiso, 1983), and even large-bodied species such as Didelphis marsupialis and Philander opossum depend heavily on insect prey (Charles-Dominque, 1983). Metachirus nudicaudatus, Marmosa sp., and Caluromys philander have also been listed as being consumers of ants and termites (Redford, 1987), although the latter species may rely more heavily on fruits (Atramentowicz, 1982).
Results of studies on insect population fluctuations in the neotropics vary (see Elton, 1975; Bigger, 1976; Janzen and Schoener, 1968; Janzen, 1973; Wolda, 1978), but most authors agree that samples obtained in wet season months are larger than comparable ones collected during the dry months (Davis, 1946; Wolda, 1978; Smythe, 1982; Charles-Dominique, 1983). Therefore, the trends of this study certainly support the notion that small mammal reproduction, especially that of marsupials, is influenced by availability of insect prey. However, until the variation in




86
the cycles of prey populations across the year is described, explaining timing of reproduction of small mammals as a consequence of abundance of insect prey in the late wet to early rainy season will remain speculative.
Fruits and seeds were the other frequent items present in the stomachs of small mammals in this study, although it is felt that rodents may depend more heavily on these resources than the more insectivorous marsupials. If the number of trees with fruits can serve as a measure of food resource availability for small mammals, the timing of reproduction of most small mammal species seems to track that of resources. In the Brazilian Atlantic forest asynchroniously fruiting trees may be found at every month of the year (Davis, 1946). Nonetheless, fruit production of several plant species of seasonal neotropical environments has been shown to be influenced by precipitation, usually peaking just prior to the start of the rainy season (Foster, 1982; Smythe et al., 1982; Charles-Dominique, 1983). A second peak in fruit productivity can also be present at the end of the rainy season (Davis, 1946). A preliminary study conducted in the Rio Doce Park (CETEC, 1981) also indicated that there are tree species flowering and fruiting throughout the year, and also that there is a slight peak in the number of fruiting species in September and October, i.e., in the early rainy season.




87
I expected that marsupials, which use invertebrates to a larger extent than rodents, should display a higher degree of seasonality due the more fluctuating nature of their food resources. This was observed in the present study. Rodents, on the other hand, can probably rely on resources which are more seasonably stable in the Atlantic forest, such as fruits, seeds and leaf tissue, and thus can reproduce throughout the year. The two most common rodents, Proechimys setosus and Oryzomys trinitatis, do not show seasonality in their reproductive patterns. Since they have also shown to be able to use and at some periods heavily rely on insects, the consequent larger resource spectrum should place lower limits on the reproduction of rodent species.
While some small mammals may exploit resources on the forest floor, most of the tropical forest productivity is located above ground (Eisenberg and Thorington, 1973). Therefore, even for primarily terrestrial species, some level of scansorial ability should prove advantageous. There was a high degree of overlap in substrate use among small mammal species in this study. Twelve species are shown to use both aerial and ground substrate regularly, even though their relative degrees of arboreality varied. Among marsupials, only Metachirus nudicaudatus was confined to the ground level of the forest. Only four of eleven rodent species can be safely classified as terrestrial, while the remaining are at least marginally arboreal. This is consistent




88
with descriptions of other neotropical small mammal fauna, where the majority of species show some level of climbing ability (see August, 1984; Fonseca et al., 1987; Nitikman and Mares, 1987).
It should be stressed that these findings do not imply that arboreality is the dominant mode of life among Atlantic forest small mammals. Of the species caught in the present study, only Caluromys philander, Marmosa cinerea, Marmosa agilis, Oryzomys trinitatis, Rhipidomys mastacalis and Echymys sp. can be regarded as predominantly arboreal. It does, however, indicate the widespread ability of a greater fraction of the community in exploring a tri-dimensional environment. This may be especially important during certain periods of the year. Charles-Dominique (1983) provided data indicating that the insect faunas of the canopy and of the undergrowth can fluctuate asynchroniously with each other. Therefore, if a particular food resource is undergoing a period of seasonal shortage, those species with versatile habits would be at an advantage. It has been demonstrated before that in the highly seasonal savanna region of central Brazil gallery forests play an important role in maintaining overall mammalian species diversity (Fonseca and Redford, 1984), and become crucial during periods of stress. Part of the reason for this may be linked to the higher productivity of the tri-dimensional gallery forest environment, when compared, during periods of moisture deficit, to that of the cerrado savannas.




Life History Patterns
While the occurrence of pouch litters with skewed sex ratios in neotropical marsupial species has not often been reported, it was interesting to find Didelphis marsupialis with litters having a predominance of males. Skewed sex ratios have only been achieved experimentally under a regime of diet supplementation, a procedure which does not always produce the predicted male bias outcome (Austad and Sunquist, 1986). Therefore, it is important that future investigations address this question. The female-biased sex ratio obtained for Metachirus nudicaudatus pouch young is not based on a large enough sample size and therefore remains inconclusive.
Three out of four marsupial species, for which sample size was large enough to allow statistical inference, proved sexually dimorphic. Male Didelphis marsupialis has also been found to be larger than females elsewhere in South America (O'Connell, 1979). Other Marmosa species also have larger and heavier males (Nowak and Paradiso, 1983), the same being true for Philander opossum. No references have been found to indicate size differences for Metachirus nudicaudatus, except for the present study. While larger males have been usually associated with polygynous species in which males strongly compete for females (Ralls, 1976), there was no evidence of territorial defense by any of the marsupials studied here. However, this does not necessarily preclude males




90
from actively competing for females at overlapping ranges. Fierce fighting for estrous females has been observed elsewhere in Dideiphis marsupialis (Austad and Sunquist, 1986). Since marsupial breeding in this study was usually confined to a certain season, the potential for male-male competition for estrous females is li kel1y.
Rodents, on the other hand, are seldom dimorphic (see
Eisenberg, 1981), especially the smaller species. In only one species, Oryzomys trinitatis, could a size difference be found in this study. In this instance, however, females proved in average to be larger than males. This could be attributed to extra energetic demand placed on pregnant and lactating females, a cost non-existent for males who do not provide parental care (see Ralls, 19T6). Moreover, since rodent young do not undergo, unlike marsupials, a teat attachment phase, female placental manmals may increase litter size above that of the teats. The increase in rodent litter sizes may be achieved if the female is larger and better nourished. Marsupial litter sizes, on the other hand, due to the obligatory teat attachment phase, are constrained by the number of teats. Furthermore, marsupials can spontaneously terminate lactation, or more often in smaller species reduce litter size if energetically stressed (Lee and Cockburn, 1985). Young are also born after a very short gestation period, making the reproductive investment minimal. For these reasons no added




91
security would be achieved by genetically fixed propensity for larger marsupial females.
Population Turnover
Although the area sampled by traplines was not enough to
represent the home range of several species, especially the larger ones, the reduced persistence time of the average small mammal individual was nonetheless striking. Average persistence times ranged from 1.2 months for male Marmosa incana to a maximum of 3.8 months for Marmosa cinerea females, with most species remaining in traplines within the range of 1.7 to 2.7 months. While it is reasonable to assume that some of the disappearances can be attributed to home range shifts (see Nitikman and Mares, 1987) and/or juvenile migration and dispersal (Lidicker, 1975), predation might also play an important role in increasing turnover rates among tropical small mammals. Monthly turnover rate for the small cricetid rodent Akodon cursor approached 80 % in this study. Several authors have suggested that predators might cause variations in local abundances (August, 1983), and sometimes they do take a large percentage of small mammal standing biomass (Hershkovitz, 1969; Pearson, 1985; Emmons, 1987; Sunquist et al., 1987). Predation in traps was observed in this study on several occasions and it is felt that it may play a major role in the structure of these communities (Chapter 4). A large number of




92
mammalian predators, as well as owls, have been observed predating on several of the marsupials and rodents trapped in the present research, and I suspect that predators are responsible for a large fraction of the observed turnover rates on these small mammal
communities.




Full Text
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID EY6K09ZJ8_VVX0HH INGEST_TIME 2017-07-11T22:10:51Z PACKAGE AA00004970_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES



PAGE 1

PATTERNS OF SMALL MAMMAL SPECIES DIVERSITY IN THE BRAZILIAN ATLANTIC FOREST By GUSTAVO A. B. DA FONSECA A DISSERTATION 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 1988

PAGE 2

ACKNOWLEDGEMENTS First and foremost, I would like to thank Dr. John G. Robinson, my major advisor, for his dedication to my work way beyond duty. His advice during planning, field work, analysis and writing of this dissertation was invaluable, and I will be forever thankful. Drs. John Eisenberg, Larry Harris, Nigel Smith, Melvin Sunquist, and Charles Woods, members of my committee, provided support in the planning of this study and improvements in earlier drafts. Dr. Kent Redford also read and commented on the manuscript. I am grateful to all of them. My study also benefited from discussions with Drs. Thomas Lacher and Michael Mares. Once more. Dr. Russell A. Mittermeier supported my research efforts by advocating my cause with the World Wildlife Fund-US, which financed the most substantial part of the costs involved in this study. Small mammals are also people, just smaller. Additional financial support was provided by the Program for Studies in Tropical Conservation, of the University of Florida, and by the Research Council of the Federal University of Minas Gerais, Brazil. The National Research Council of Brazil (CNPq) awarded me with a doctoral fellowship. n

PAGE 3

Dr. Celio Valle encouraged the field work with his enthusiasm, vision, and eternal optimism. Several people helped during field work, too many, and in too many ways, that I apologize beforehand for failing to mention. Cecilia Kierulff was my most dedicated and frequent field assistant, and for that I will be eternally in debt. My small mammal data from the Rio Doce Park was partially collected by Jody Stal lings and his crew. This study greatly benefited from his friendship and discussions while I was in Brazil. I would also like to thank Ludmllla Aguiar, Ilmar Bastos, Sonia Rigueira, Carlos Alberto Pinto, Ederson Machado (sometimes), Silverio Machado, Ney Carnevalli, Luiz Fernando Mello, Gisela Herrmann, Jairo Vieira, Eduardo Veado, Eduardo Sabato, Luiz Paulo Pinto, and Maria Cristina Alves. Mr. Feliciano Abdalla and Dr. Antonio Cupertino kindly allowed me to work on their farms and provided housing. "Santinho" let us stay in his house at the expense of family problems, and I am thankful to him. The State Forest Institute of Minas Gerais (lEF) provided me with accommodations and gasoline at the Rio Doce State Park. I am very grateful to the staff of the Park. Eng. Ftal. Jose Lourenco Ladeira identified the trees in and outside of the Rio Doce Park. I am also indebted to him. Drs. M. Carleton, K. Creighton, L. Emmons, G. Musser, P. Myers, and J. Patton generously identified the small mammal voucher specimens. Even risking making her angry, I cannot help but to recognize that my wife's love was as helpful as her intellectual stimulation and iii

PAGE 4

support. My son Bruno tolerated my bad moods and I hope to be able to repay him in the future. My parents continued to provide me with love, help and support. The knowledge that I can count on them at all times is one of my most valuable strengths. To all of them, I emotionally dedicate this dissertation. I would also like to dedicate this work to Drs. Celio Valle and Russell Mittermeier. Together and unselfishly they managed, in less than 5 years, to train and put to work more Brazilian biologists than others had in the last three decades. The vast majority of young Brazilian ecologists, previously so prone to early infanticide, and who will be working in Brazil in the next decade, will certainly have crossed the paths of these two visionary people. The history of Brazilian biological conservation sciences will soon have to pay its dues to Celio and Russ. w

PAGE 5

TABLE OF CONTENTS Page ACKNOWLEDGEMENTS i i ABSTRACT vi i CHAPTER 1 INTRODUCTION General Background 1 Study Organization 4 2 BIOLOGY AND NATURAL HISTORY OF BRAZILIAN ATLANTIC FOREST SMALL MAMMALS Introduction 7 Mater i al s and Methods 9 Study Sites 9 Climate 12 Trapping 13 Results and Species Accounts 19 Discussion 83 Seasonality and Resource Use 83 Life History Patterns 89 Population Turnover 91 3 SMALL MAMMAL SPECIES DIVERSITY IN BRAZILIAN TROPICAL PRIMARY AND SECONDARY FORESTS OF DIFFERENT SIZES Introduction 93 Materials and Methods 97 The Region 97 Study Sites 98 Trapping 100 Habitat Variables 102 Statistical Methods and Data Analysis 104 Dependent Variables 104 Predictors of Community Structure 115 Primary Versus Secondary Forests 116 Area Size Effects 117 Results 118 Discussion 145

PAGE 6

Conservation and Management Implications 154 4 THE RELATIVE ROLE OF HABITAT SELECTION, COMPETITION AND PREDATOR PRESSURE ON THE STRUCTURE OF TROPICAL FOREST SMALL MAMMAL COMMUNITIES Introduction 157 Materials and Methods 161 Study Sites 161 Trapping 163 Habitat Variables 165 Statistical Methods and Data Analysis 166 Results 170 Species Composition and Abundance Patterns 170 Habitat Selection 176 Species Interactions 183 Predation 184 Discussion 192 5 CONCLUSIONS AND SYNTHESIS 210 LITERATURE CITED 216 BIOGRAPHICAL SKETCH 233 VT

PAGE 7

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 PATTERNS OF SMALL MAMMAL SPECIES DIVERSITY IN THE BRAZILIAN ATLANTIC FOREST By GUSTAVO A. B. DA FONSECA August 1988 Chairman: John G. Robinson Major Department: Wildlife and Range Sciences, School of Forest Resources and Conservation The influences that habitat structure, area size, and species interactions have on the species richness and diversity of small mammal communities in the Brazilian Atlantic forest were investigated in a 17-months study. Three primary and three secondary forest plots of different sizes were subjected to a capture/recapture program. Trapping effort totaled 57,120 trap nights. A total of 19 small mammal species (rodents and marsupials) were registered during the course of the study, with species richness and diversity achieving their highest at the second-growth forest plot of large size. Small mammal species diversity only responded to the increase in area size among secondary forests, with primary forests being fairly poor communities. Species vii

PAGE 8

richness, species diversity and total number of individuals in communities were found to be correlated with several habitat attributes characteristic of second-growth forests. The common opossum, Didelphis marsupialis is a dominant species in these small mjuronal communities. Densities of other small mammal species were found to be depressed by opossums. The higher predator pressure on opossums by mammalian carnivores in the larger forest patches may be the factor responsible for allowing plots of increasing size to be more spades rich and diverse. vi ii

PAGE 9

CHAPTER 1 INTRODUCTION General Background The Brazilian Atlantic forest is one of the most threatened ecosystems in the world, with habitat reduced to small islands and very little remaining in undisturbed state (Mittermeier et al., 1982; Fonseca, 1985a). These isolated areas are characterized mostly by second growth under different regeneration stages. The consequences for the vertebrate fauna inhabiting these habitat fragments is not known, except for a few primate species. This is aggravated by the fact that not much information is available on the natural history of Atlantic forest mammalian species. By investigating species composition, diversity and relative abundances of small mammal communities of these tropical forests, this study will address these questions. Small mammal community ecology research has revealed several parameters that, to variable degrees, explain underlying community characteristics and organization. Local species composition and relative abundances of small mammals and birds have been partially explained by several different measures of habitat structure and 1

PAGE 10

2 heterogeneity (Rosenzweig and Winakur, 1969; Johnson, 1975; M'Closkey, 1976; Dueser and Shugart, 1978; James and Wamer, 1982; August, 1983, 1984; Lynch and Whigham, 1984). As has been observed with plants (Horn, 1974), animal communities can sometimes be expected to decrease in diversity in late successional stages (Connell, 1978), possibly as a consequence of the lower degrees of spatial heterogeneity in climax forests. Therefore, the elimination of primary forests in the Brazilian Atlantic region may not have resulted in the decrease in small mammal species diversity. Within this framework, this study investigated the nature of small mammal species richness and diversity of second growth, which is increasingly a characteristic of the present day habitat patches of the Atlantic forest. The size of habitat fragments has also been used as a determinant of mammalian species richness, especially in insular communities (Dueser and Brown, 1980; Lomolino, 1984). MacArthur and Wilson (1967) proposed an immigration-extinction equilibrium model to explain number of species in islands. As the size of an island increases, the probability of species extinction decreases. If the island is close enough to a potential source of colonists, extinction is somewhat compensated for by the frequent addition of new colonists, but if the degree of isolation is the same for two islands of different sizes, the larger one will tend to support richer communities. Several studies on different taxa have shown that continental habitat patches can sometimes behave like islands, if the surrounding habitat provides

PAGE 11

3 effective isolation from the sources of colonists (e.g. Vuilleumeier, 1970; Seifert, 1975; Brown, 1978; Faeth and Kane, 1978; Harris, 1984; Blake and Karr, 1987). Notwithstanding the fact that a few attempts have been made (Malcolm, 1987), area-size relationships have yet to be demonstrated for tropical small mammal communities. The structure of the habitat may also influence community composition and relative abundances in islands of increasing size. As Williams (1964) pointed out, as the size of the area increases, the probability that additional habitats with their distinct species will be incorporated into samples also increases. Therefore, both factors have to be taken into consideration in investigating the mechanisms that promote species diversity. Recently, studies have began to incorporate both area-size relationships and habitat structure characteristics into models explaining community parameters. These have occasionally successfully separated the "pure" or dynamic effects of the size of the area from those related to environment structure (e.g., Simberloff, 1976; Weaver and Kellman, 1981; Buckley, 1982; 1985; Boecklen, 1986). In addition, both area and habitat structure can have an underlying effect in the determination of patterns of species interactions within communities, which in turn can potentially affect species composition and relative abundances of these assemblages. Competitive interplays are known to exert influence upon small mammal community characteristics (Price, 1978; Dueser and Porter, 1986; Lacher et al., in press), but mixed

PAGE 12

4 results have been obtained in the attempts to demonstrate competition (Murua et al. 1987). I believe that previous models, although successful to various degrees in their predictive power, were commonly too simplistic. The result is that significant amounts of variation in species richness and diversity between communities are left unexplained. Furthermore, single parameter models may conceal the underlying mechanisms by which species diversity is promoted and maintained (Connor and McCoy, 1979). Therefore, their application to conservation schemes can be potentially dangerous (see for example, Zimmerman and Bierregaard, 1986). The multivariate approach, which takes into account the many more potential influences arising from area-size relationships, habitat structure, and species interactions, was chosen for this study. This approach favors the information conveying value of the method rather than the quantitative models it can possibly generate. Although the method results in models that may have limited applicability to different habitats, they frequently have higher heuristic value and are more biologically realistic. Study Organization The core of this study is presented in three main chapters. They are meant to either stand alone as individual studies or constitute additive sections of a major work. Chapter 2 deals with the biology

PAGE 13

5 and natural history of the small mammal species that were recorded in this study. It presents the general aspects of reproduction and population dynamics, substrate use, diet, movements and persistence in traplines, for the species with large enough sample sizes to allow analysis. Morphometric data on all species are also presented. This section serves as a data base for the understanding of the following chapters. In Chapter 3, I deal mostly with the relation between habitat structure and area size, and the way in which they affect diversity and relative abundance of small mammals. Communities are studied as collective entities, with very little emphasis placed on the individual species. Six forest plots of different sizes and levels of disturbance were sampled for small mammals. In addition to species richness, the total number of individuals in the communities, and the Shannon-Wiener species diversity index, I also developed a second index of community structure that takes into account species abundances. I used a large number of vegetation parameters as independent variables indicative of habitat structure. Sizes of the forest plots sampled constituted the second most important variable under study. The third main chapter, Chapter 4, investigates the structure of small mammal communities in terms of requirements of individual species and the competitive interactions among these species. It also examines the possible influences that habitat requirements and size of the area can have on the outcome of community composition and relative

PAGE 14

6 abundances. It explores the mechanisms by which patterns observed in the previous chapter are determined. Finally, Chapter 5 offers a synthesis of the results, the conclusions, and provides a summary of overall major findings. It also explores possible conservation applications of the results of this research.

PAGE 15

CHAPTER 2 BIOLOGY AND NATURAL HISTORY OF BRAZILIAN ATLANTIC FOREST SMALL MAMMALS Introduction The neotropical region encompasses some of the most threatened ecosystems in the world, and yet the mammalian fauna of this area still remains very poorly known (Mares and Genoways, 1982). In South America, the Brazilian Atlantic forest has been the vegetation formation subjected to the highest rate of destruction, with less than 5 % of the region possessing some form of forest cover (Fonseca, 1985a), and probably less than 1 X in undisturbed state (Mittermeier et al., 1982). Nonetheless, the few natural history studies conducted on its fauna have indicated that species diversity is quite high, with many faunal elements being unique to the region (Mello-Leitao, 1946; Muller, 1973; Fonseca, unpublished data). The pioneer works of Moojen (1952), Vieira (1955), and Cabrera (1957; 1961) provide a general data base indicating that there are at least 129 species of non-volant mammals in the Atlantic forest region, about 40 % of which are endemic. Were the taxonomy of most of the groups better understood, other species would probably be described. There are at least 23 marsupial

PAGE 16

8 and 57 rodent species described for the region, of which 39 % and 53 %, respectively, are endemic to the region. Because of past and present habitat destruction, the fauna is increasingly isolated into small patches, and many members of this unique ecosystem are now endangered. This situation has been documented for the most conspicuous elements of the mammalian fauna, the primates (Coimbra-Filho and Mittermeier, 1977; Mittermeier et al., 1982; Fonseca, 1985b). Nonetheless, the rich and highly endemic small mammal fauna of the Brazilian Atlantic forest region has been the subject of very few long-term studies, especially when compared to the Amazon region (e.g.. Pine, 1973; Lovejoy et al., 1984; 1986; Terborgh et al., 1984; Malcolm, 1987) or the Cerrado (Alho, 1981; Alho et al., 1986; Fonseca and Redford, 1984; Lacher et al., in press; Nitikman and Mares, 1987). The only detailed study of Atlantic forest small mammals is now over 40-years-old (Davis, 1946). Historically, the Brazilian Atlantic forest extended from the coast to the eastern and portions of the western slopes of the coastal mountains (Hueck, 1972; Alonso, 1977), an area of approximately 700,000 square kilometers. Where rainfall permits the presence of tall, evergreens, this type of forest extends into the western slopes of the coastal formation. Due to a rain shadow, the vegetation of the western slopes, where this study was conducted, possesses a number of deciduous tree species, which lose their leaves during the approximately 6 months of the dry season. During the wet season, however, the forests of both

PAGE 17

9 the eastern and western regions are physiognomical! y undistinguishable. In addition, the faunal elements are mostly the same, generally belonging to the same biogeographical region (Muller, 1973). The objective of this study was to investigate several aspects of the biology and natural history of eastern Brazilian non-volant small mammals. During a period of 17 months, data on the general aspects of population dynamics, breeding, substrate use and movement patterns of small mammals (marsupials and rodents) were collected in six forest plots, at three main sites in the state of Minas Gerais, Brazil. Morphometric data were also collected. Species composition, relative abundances and other community structure patterns are presented elsewhere (Chapters 3 and 4). Materials and Methods Study Sites Small mammal communities of six forest plots, two at each of three sites in the state of Minas Gerais, were the subjects of this study (Figure 2-1). At each site, a primary and a secondary forest plots were selected. A set of three parallel transect trapping lines was established in each. The primary forest plots showed vertical stratification, with an average canopy height of 19 meters. The herbaceous stratum was somewhat sparse, while the midstory was

PAGE 18

10 Figure 2-1. Location of study sites: (1) Fazenda Esmeralda, Rio Casca county; (2) Parque Estadual Florestal do Rio Doce, Marileia county; (3) Fazenda Montes Claros, Caratinga and Ipanema counties.

PAGE 19

11 generally well developed. The secondary forests, on the other hand, were mostly in their m1d-stages of succession. Average canopy height was approximately 12 meters; herbaceous cover was frequently extensive, with masses of tangled vines being a common occurrence. Epiphytes and emergents were conspicuously absent from secondary forests. The first site was Fazenda Esmeralda, located on the county of Rio Casca. The farm is under extensive agricultural use, with very little remaining under forest cover. Located along the plains of the Rio Doce River, the farm was covered almost completely by pristine forest as recently as 1964. Wood extracting rights were then sold to the largest steel industry of the state of Minas Gerais and by 1970 most of the farm was deforested. Forest patches located on the top of two hills were selected for the study, one with a 60 ha. second growth, and the other with 80 ha. of primary forest, known locally as "Lagoa Fria". Fazenda Montes Claros was selected as the second site for this study. It is a coffee and cattle farm located within the counties of Ipanema and Caratinga. The total area of the farm is about 1,200 ha., 860 ha. of which remain under forest cover. A research station under the administration of the Brazilian Foundation for Conservation of Nature (FBCN) and the Federal University of Minas Gerais (UFMG) was established in the farm in 1983. A second growth forest patch at Fazenda Montes Claros, known as "Jao" was selected, and another under primary forest, "Matao," was also used.

PAGE 20

12 The third site was the Rio Doce State Park, which has portions of its area within the county of Marileia. The park, with its 35,973 ha., constitutes the largest continuous area under tropical forest in the state of Minas Gerais. It was created in 1944 and has been ever since under the jurisdiction of the Government of the state of Minas Gerais. Since the creation of the State of Minas Gerais Forest Institute (lEF) in the late 1960s, the park has been under its administration. It has recently become one of the best maintained and protected areas under the Brazilian Parks system. Because of several extensive fires in the 1960s, a considerable area of the park is second growth. One of these, "Mata do Hotel" was selected as a study area. A pristine primary forest, known locally as "Campolina," constituted the second patch selected for study within the Rio Doce Park. Climate CUmatological data were collected at all three sites, but because of the extreme similarity in temperature and rainfall regimes among study sites, only information collected at Fazenda Montes Claros is presented here. Hueck (1972) states that rainfall for the western slopes of the Atlantic forest region is always below 1,600 mm annually. In some areas, it can achieve a little over 1,000 mm. During the study period, the region experienced an unusually dry period, as we can observe in the climatogram of Walter (1971) constructed with data collected at

PAGE 21

13 Fazenda Montes Claros (Figure 2-2): total rainfall for the first 12 months of trapping was 850 mm, while for the last 12 months rainfall totaled 931 mm. Rainfall is highly seasonal, being concentrated between the months of September and February (Figure 2-2). Average monthly precipitation for the rainy season was 128 mm, while dry season rainfall averaged only 30 mm monthly. Mean minimum annual temperature during the study period was about 18 C, close to the average for the region (Hueck, 1972). Average daily differences between minimum and maximum temperatures are quite constant throughout the year. Rainfall maxima coincide with the warmest months of the year, while winters are usually very dry. For the purpose of this analysis, the dry season is considered to occur between the months of March and August, and the wet season between September and February. Trapping Three transect lines 300 meters long were established in each plot at each of the three sites. These transects were as parallel to each other as possible and separated by 100 meters. Each transect line possessed 16 trapping stations 20 meters apart. Traps were placed in suitable locations within a 3.5 meter radius measured around center of station. A squirrel-size Tomahawk live trap (Tomahawk Live Trap Co., one-door folding trap, size 203, Tomahawk, WI) was placed on the ground at each station. At every other station, a second Tomahawk trap of the

PAGE 22

14 MONTHLY RAINFALL (mm) 300 250 200 150 100 MEAN TEMPERATURE ( C) 150 135 120 RAINFALL SURPLUS 105 90 DROUGHT 75 60 45 30 15 Jun Jul AugSepOct NovDec Jan FebMar Apr May Jun Jul AugSepOct MONTHS Figure 2-2. Climate diagram of Walter (1971) for the forests of the western slopes of the Atlantic forest, built using data collected at Fazenda Montes Claros between June, 1985 and October, 1986. The diagram indicates periods of drought and of water surplus. The dry season can be defined between the months of March and August. Temperatures are given in oC, and rainfall in mm.

PAGE 23

15 same size was wired either to a branch or vine at heights from 1 to 4 meters high. In addition to these traps, every other station possessed a mouse-sized collapsible Sherman trap (H. B. Sherman Traps, Inc., Tallahassee, FL), with alternation of ground and tree traps. Moreover, the two outermost transect lines had, at every other station, a large 80 X 30 X 30 centimeters wire home-crafted live trap. Therefore, each outer line possessed 16 ground Tomahawk traps, 8 tree-bound Tomahawk traps, 4 ground and 4 tree-bound Sherman traps, and 8 ground large wire traps. The total for each outer line was 40 traps. The mid-line did not have large traps, but a total of 24 Tomahawk and 8 Sherman traps. In summary, each forest plot had 48 trap stations disposed into 3 transects of 16 stations each, and a total of 112 permanent based traps (Table 2-1). With the exception of Sherman traps, all traps were closed at the end of each five night trapping session and left in place. Sherman traps were removed, washed, and replaced each month. Trapping took place between June of 1985 and October of 1986, i.e., for 17 consecutive months. Each forest plot was trapped for five consecutive nights every month. During the course of the study, a total of 9,520 trap nights were accumulated for each plot. Therefore, trapping effort for all six forest plots together totaled 57,120 trap nights. Fresh pineapples, oatmeal and a cotton ball soaked with a commercial codfish oil solution were used as baits. Traps were checked

PAGE 24

16 Table 2-1. Schematic representation of the first four stations in transect trap lines. Additional stations are replicates of the first four. Each trapline is 300 meters long, separated from each other by 100 meters. Stations are disposed within each line at every 20 meters. LINE ONE — 100m — ii -Ground Tomahawk -Above-Ground Tomahawk LINE TWO ~100m~ LINE THREE 1 1 -Ground -Ground Tomahawk Tomahawk -Ground -Above-Ground Sherman Tomahawk — 20m — #2 -Ground Tomahawk -Large Trap -Above-Ground Sherman M2 -Ground Tomahawk -Above-Ground Tomahawk *2 -Ground Tomahawk -Large Trap -Above-Ground Sherman -20m~ 13 -Ground Tomahawk -Above-Ground Tomahawk M3 -Ground Tomahawk -Above-Ground Sherman -Ground Tomahawk -Above-Ground Tomahawk ~20m~ M -Ground Tomahawk -Large Trap -Ground Sherman -Ground Tomahawk -Above-Ground Tomahawk M -Ground Tomahawk -Large Trap -Ground Sherman #16 tt16 16

PAGE 25

17 every morning for captures and for adequacy of bait, which was replaced as needed. For each individual, the following information was recorded: 1. Species. If not readily identifiable, the individual was preserved for later taxonomic identification. The only individuals which were consistently prepared as skins and skulls were representative of Oryzomys trinitatis and Oryzomys nigripes The former shows a high degree of variability in skin color and pattern and was considered, during field work, as belonging to more than one species. Whenever possible, animals that died in the traps were also preserved. Moreover, live trapping and snap-trapping were conducted on other locations within the same forest type for preparation of voucher specimens and determination of stomach contents. Taxonomic identification of voucher specimens was provided through the kindness of the following specialists: Drs. M. Carleton, C. Creighton, L. Emmons, G. Musser, P. Myers, and J. Patton. 2. Location in grid. 3. Individual identification. If already tagged with metal fish tags (Fish and small animal tag, size 1, National Band and Tag Co., Newport, KY), the animal was promptly released; otherwise the animal was tagged. Individuals with positive taxonomic identification were released at the same station where captured. 4. Sex. 5. Weight.

PAGE 26

18 6. Body length. 7. Tail length. 8. Ear length. 8. Hind foot length. 9. Reproductive condition. Female rodents were checked for perforated vaginas. I also noticed whether individuals were pregnant (in late stages) or lactating. Male rodents were considered in breeding condition if testes were descended. Female marsupials were considered in breeding condition if neonates were found in either pouch or attached to nipples, or if the nipple area showed signs of recent nursing by previous litter. Male marsupials with testes of reduced size were considered pre-pubertal while for some species ( Metachirus nudicaudatus Marmosa incana Marmosa cinerea, Marmosa agilis and Philander opossum ) abdominal or sternal scent gland activity indicated breeding condition. We also recorded any pouch young old enough to be sexed. 10. Behavior upon release. Analysis of variance (Procedure GLM: PC-SAS, SAS Institute, Cary, NC) was used to test the null hypothesis of absence of sexual dimorphism, using the following morphological characters: weight, body length, tail length, hind foot length, and ear length. Kruskal-Wal lis tests were performed to test for differences in persistence times in traplines among sexes.

PAGE 27

19 Since the number of arboreal traps in each set of three traplines (36; 32 %) was different from that of terrestrial ones (76; 68 X), analysis of differential trapping success of arboreal versus terrestrial traps was conducted with corrections to account for the differential probability of trapping in the different strata. Where sample size permitted, chi-square tests were performed to test the null hypothesis of lack of difference in substrate use among sexes. For this analysis, all proportions were arc-sin transformed (Sokal and Rohlf, 1981). Results and Species Accounts During the course of this study, a total of 692 individuals of 8 marsupial and 11 rodent species were caught 1,366 times, in all six forest plots (Table 2-2). One individual of the species Monodelphis eimericana was trapped on one of the auxiliary lines at the primary forest of Fazenda Montes Claros. Mean trapping success, averaged over 17 months of trapping at all sites, was 2.4 %. For most species, trapping success was much higher during the cold, dry months, dropping sharply around October with the onset of rains. Caluromys philander Woolly opossums were only registered for the Rio Doce State Park. Although occasionally trapped on the ground, this species is mostly arboreal (Davis, 1947; Nowak and Paradiso,

PAGE 28

20 1983). In this study, only one individual was captured on the ground. Upon release, all individuals climbed into canopies, disappearing from sight at heights of about 10 meters. Both the small number of arboreal traps (32 %), relative to terrestrial ones, and the absence of traps in higher strata of the forest, may have accounted for the low trapping success for this species. Only three males and three females were captured during the course of the study. One female was recaptured once. Morphometric data on these individuals are presented in Table 2-3. Didelphis marsupialis Opossums occurred at all sites and represented the most common small mammal species captured in the majority of forest plots. This is a general ist species, found in all habitat types (Hunsaker, 1977; Miles et al., 1981; Alho et al., 1986). Adult males were significantly heavier than females (Table 2-4). Hind feet were also sexually dimorphic. Other biometric parameters did not vary between sexes. Opossums have been previously described as being mainly terrestrial, although they are able to climb opportunely while foraging (Miles et al 1981). These generalizations were confirmed here. Overall, 73 % of D^ marsupialis captures were from terrestrial traps, while 93 % of individuals released also remained on the ground. However, young opossums tend to use aerial

PAGE 29

21 Table 2-2. Trapping results for all months of trapping. six forest plots together during 17 SPECIES Number of Total Number of First Captures Captures Caluromys philander 6 7 Didelphis marsupialis 144 404 Marmosa aqilis 7 11 Marmosa cinerea 43 120 Marmosa incana 127 220 Marmosa microtarsus 1 1 Metachirus nudicaudatus 105 156 Phi lander opossum 14 29 Abrawavomvs ruschi 1 1 Akodon cursor 29 53 Nectomys sauamipes 10 14 Orvzomvs capito 1 2 Oryzomys niqripes 5 5 Oryzomys subflavus 1 1 Oryzomys trinitatis 84 99 Oxymycterus roberti 2 2 Rhipidomys mastacalis 1 1 Echymys sp. 1 1 Proechimys setosus 110 239 TOTAL 692 1366

PAGE 30

22 Table 2-3. Biometric data for three males and three females of Caluromys philander Weight is given in gms, and other body measurements in mm. MALE FEMALE Mean St.Dev. Mean St.Dev. Weight 171.3 57.6 180.5 40.5 Body length 210.7 30.0 206.8 11.4 Tail length 301.0 25.9 298.8 20.7 Hind Foot 36.3 2.5 36.3 2.6 Ear length 33.3 1.5 30.5 1.9

PAGE 31

23 Table 2-4. Biometric data for adult male and female Dldelphis marsup1al1s Weight is given in gms, and other body measurements in mm. N=number of individuals, St. Dev.=Standard Deviation. P refers to p-value associated with analysis of variance for sexual dimorphism (n.s.=p>0.05;*=p<0.05; **=p<0.01; ***=p<0.001 ) MALE FEMALE Mean St.Dev. N Mean St.Dev. N P Weight 1,024.0 361.4 63 946.0 157.0 34 ** Body length 354.0 40.4 63 354.0 29.7 34 n.s. Tail length 348.5 30.9 62 349.6 27.4 31 n.s. Hind Foot 58.1 3.9 62 56.1 4.3 34 t* Ear length 51.2 4.2 62 50.0 3.6 34 n.s.

PAGE 32

24 Frequency 3 4 5 6 7 8 9 Number of months in lines 10 12 Males "• — Females Figure 2-3. Frequency of Individuals of D^ marsuoialis persisting within traplines.

PAGE 33

25 substrate more often than adults. Approximately 23 % of the individuals released (11 out of 37) climbed into trees and/or onto vines. The same result was obtained by Davis (1947). There were no significant differences in substrate use between adult males and females. A total of 79 males and 64 females were captured 196 and 204 times, respectively, with sex ratios of both number of individuals and total captures deviating very little from 1:1. Based upon recaptures, males did not persist for long periods on grids, averaging 1.74 months. Females stayed within the traplines an average of 2.6 months, but this difference is not significant (p=0.14; Figure 2-3). Record persistence times were achieved by one female and one male, which were recorded for a total of 12 months. Fleming (1972) found similar low persistence times for Didelphis marsupialis in Panama. If the data from Sunquist et al. (1987) with Di marsupial is from Venezuela is extrapolated to the present study, the low persistence times can probably be attributed to high mortality rates rather than to dispersal. The distance travelled between successive captures were not significantly different between males and females. The overall mean was 143.4 meters (maximum=400 meters), slightly higher than the figure given by Fleming (1972). However, similar to grid trapping methods, traplines underestimate actual distances

PAGE 34

26 travelled in a night, which was found by Sunquist et al (1987), using radio transmitters, to average one kilometer. Even though trapping success during the first three months of study was very high, population levels, as reflected by the number of captures, did not appear to vary between dry and rainy seasons (Figure 2-4). Reproductive activity, on the other hand, was highly seasonal, with most females breeding just prior and well into the rainy season. The number of juveniles recorded also followed this same pattern (Figure 2-5). Two females were recorded as having two consecutive litters, the first in August, and the second in October. The same has been observed by Collins in Brazil (1973, in Nowak and Paradiso, 1983). The occurrence of two consecutive litters has also been reported elsewhere for D^. marsupial is in Colombia (Tyndale-Biscoe and Mackenzie, 1976) and southeastern Brazil (Davis, 1946). An interesting result was the decrease in opossum abundances during the second years's late dry season, especially in the small forest plots. Smaller plots are where D^ marsupialis was most found to be most abundant during the first year of trapping, and capture success was much lower than for the previous year (Figure 2-4). The reason for this apparent population crash is not yet clear, but may be due to a reduction in food resources that occurred as a result of a severe dry season. Although the reduction in abundance was observed across all plots, smaller

PAGE 35

27 100-f -I — — I 1 1 \ 1 1 1 1 1 — — I — — I 1 — — I 1 r Jun Jul AugSepOctNovDecJanFebMar AprMayJun Jul AugSepOct Number of Captures Rainy Season Limits Figure 2-4. Number of captures of D^ marsupialls according to monthly trapping period. Straight lines enclose boundaries of rainy season (September to February).

PAGE 36

28 14 12 10 8 6-\ 4 2 Jun Jul AugSepOct NovDec Jan Feb Mar Apr May Jun Jul AugSepOct CZ] Breeding Females ^^ Number of Juveniles — ^^Mean Rainfall (cm) Figure 2-5. Number of breeding females and juveniles of D^. marsupialis according to monthly trapping period. Line represents mean daily rainfall (mm).

PAGE 37

29 forests experienced the highest decrease, lending support to this hypothesis, since populations inhabiting smaller plots should be more susceptible to a decrease in food resources. Average litter size of pouch young was 8.6, with a M:F sex ratio of 1.7:1.0, highly skewed towards males (Table 2-5). Overall, two thirds of females produced male biased litters. This average litter size is much higher than those obtained in Colombia by Tyndale-Biscoe and Mackenzie (1976) or by Fleming (1973) in Panama. Both studies reported averages of approximately 6.5 young. In addition, these studies found that sex ratios of pouch young did not differ from a 1:1 ratio. Austad and Sunquist (1986) were able to experimentally induce male-biased sex ratios in D^. marsupialis by providing dietary supplementation. Therefore, relative to the sites studied by Fleming (1973) and Tydale-Biscoe and Mackenzie (1976), the forests sampled in the present research may be more productive for common opossums, which may have resulted in the observed skewed sex ratios. Marmosa agilis This small arboreal marsupial can be fairly common in gallery forests of the Cerrado region (Nitikman and Mares, 1987), but was infrequently caught in this study. The first individual of the species was recorded only after 12 months of trapping. In total, three males and four females were caught 11 times in two forest plots. Ten of the eleven captures (91 %) were

PAGE 38

30 Table 2-5. Litter sizes and sex ratios of Didelphis marsupialis pouch young. ID Number Number of Number of Total Sex Ratio Female Young Mai e Young Young M:F 0076 9 0088 9 0055 7 0072 9 0087 9 0126* 3 0156 7 0162* 4 0067 7 0192 7 0193 10 0209 3 6 9 2.0:1.0 0072** 9 0081 11 0220 8 0251 8 0257 8 0258 8 0274 3 5 8 1.7:1.0 0209** 2 7 9 3.5:1.0 0278 5 0281 3 4 7 1.3:1.0 0227 5 5 10 1.0:1.0 0480 5 5 10 1.0:1.0 14/10 2 7 9 3.5:1.0 42P 2 5 7 2.5:1.0 11/12 4 4 8 1.0:1.0 Means 8.6 1.7:1.0 Females were in poor condition, with multiple injuries. Therefore, their litter sizes were not used in calculating means. Second consecutive litter.

PAGE 39

31 on traps placed on trees or vines, which is a capture rate on arboreal traps similar to that obtained by Nitikman and Mares (1987). Alho et al. (1986) also found M^ agilis to be preferentially arboreal in gallery forest habitat. Except for two males and one female, which were recorded in traplines for two months, no other animal was recaptured in subsequent trapping sessions. The only individual for which data on movement between trapping stations were obtained travelled 40 meters in consecutive nights, a figure very close to that determined by Nitikman and Mares (1987). Biometric data on Marmosa agilis are presented in Table 2-6. Marmosa cinerea This species is one of the largest within the genus (Nowak and Paradise, 1983) and was fairly common at the Rio Doce Park, but was conspicuously absent from both Fazenda Montes Claros and Fazenda Esmeralda. Most individuals were captured during the cold, dry months (Figure 2-6), with trapping success during the rainy season being fairly low. Lactating females, on the other hand, were only trapped during the rainy season, indicating that the high trapping success of the following months reflected the addition of recently born young into the population. A total of 8 males and 35 females were caught 120 times (Table 2-2), making it the species with the highest recapture rate in

PAGE 40

32 Table 2-6. Biometric data for four males and one female Marmosa agiUs Weight is given in gms, and othe r body measurements in mm. MALE FEMALE Mean St.Dev. N Mean St.Dev. N Weight 29.5 11.3 4 25.0 1 Body length 109.3 13.2 4 99.0 1 Tail length 155.3 7.1 4 151.0 1 Hind Foot 18.3 1.5 4 17.0 1 Ear length 20.5 0.6 4 21.0 1

PAGE 41

33 20 15 10 ^ ^^^^ ^Z/^/////M/^^/f^f/Mf/ 77777Z. DU tfia M—MA M it — I \ — — I 1 1 — — I 'i — — I 1 1 — — I \ 1 1 — — I — — 1 r Jun Jul AugSepOctNovDecJanFebMar AprMayJun Jul AugSepOct Number of Captures Rainy Season Limits Figure 2-6. Number of captures of Ms. dnerea according to monthly trapping period. Straight lines enclose boundaries of rainy season (September to February).

PAGE 42

34 Differential Trapping Success (in %] Males Females Terrestrial Traps Arboreal Traps Figure 2-7. Differential trapping success (in percentages) of M^ cinerea in arboreal and terrestrial traps.

PAGE 43

35 this study (31 % for males and 37 % for females). While sex ratio for first captures was 1.6:1.0, total captures approximated a 1.0:1.0 ratio. This might indicate that females are captured more frequently than males, which may be due to the fact that males are more transient than females (chi-squarer5. 17; df=1, p<0.02), and averaged only about 1.8 months in traplines. Females, on the other hand, persisted in the areas for an average of 3.8 months. The longest tenancy in traplines was also by a female that was recorded present for 14 months. In addition, males also travelled farther between successive captures than females (males=142 meters, maximum: 380 meters; females=80 meters, maximum: 180 meters), which also supports the contention that transient males are a frequent occurrence. On average, adult males were found to be slightly heavier than females. However, the species did not show sexual dimorphism for any biometric parameter measured (Table 2-7). Although trapped frequently on the ground, M^ cinerea appeared to be primarily arboreal, especially the females (Figure 2-7). The behavior upon release also revealed that, while 56 % of the males remained on the ground, only 27 % of the females displayed a similar behavior. Terborgh et al. (1984) list M^ cinerea as being able to exploit several forest strata, from understory to canopy. Miles et al (1981) determined Mj. cinerea to be mostly arboreal, with all nests also located above ground.

PAGE 44

36 Table 2-7. Biometric data for adult male and female Marmosa dnerea Weight is given in gms, and other body measurements in mm. P refers to p-value associated with analysis of variance for sexual dimorphism (n.s.=p>0.05; *=p<0.05; **=p<0.01; **=p<0.001 ). MALE FEMALE Mean St Dev N 24 Mean St. Dev. N P Weight 117.3 29.4 99.7 16.7 11 n.s. Body length 176.1 20.0 24 177.1 20.5 11 n.s. Tail length 258.9 26.0 24 261.7 17.9 11 n.s. Hind Foot 28.9 3.9 24 28.0 3.1 11 n.s. Ear length 30.4 2.8 24 29.7 2.1 11 n.s.

PAGE 45

37 Frequency -r 2 3 4 5 6 Number of months in lines Males Females Figure 2-8. Frequency of individuals of Mj. incana persisting within traplines.

PAGE 46

38 Marmosa jncana This was the second most frequently caught species in this study. It was present in all six forest plots sampled. Eighty males and 44 females were captured a total of 220 times. Three individuals escaped before their sex could be determined. Sex ratio for first captures was 1.8:1.0, and 2.9:1.0 for all captures. Males also persisted on the traplines more than females (chi-square=8. 1 df=1, p<0.005), on average 1.9 and 1.2 months, respectively (Figure 2-8). Therefore, in contrast with M^ cinerea male M^ incana appear to be more sedentary than females. Mi incana males travelled an average of 64.7 meters between successive captures (maximum=200 meters), while the only female recaptured in the same trapping session travelled 40 meters. Marmosa incana adult males were sexually dimorphic for all morphometric characters measured, with males being, on average, 20 % heavier than adult females (Table 2-8). Body length of males was also 14 % larger than for females. Adult females also lack the sternal gland, which is functional in males during the breeding season. As with Mi. cinerea trapping success increased markedly with the end of the rainy season (Figure 2-9). This population growth is accounted for by the increase in the number of juveniles trapped (Figure 2-10). The hypothesis that increased trapping success is due to juvenile recruitment is further supported by the observation

PAGE 47

39 Table 2-8. Biometric data for adult male and female Marmosa incana Weight is given in gms, and other body measurements in mm, P refers to p-value associated with analysis of variance for sexual dimorphism (n.s.=p>0.05;*=p<0.05; **=p<0.01; ***=p<0.001). MALE FEMALE Mean St.Dev. N Mean St.Dev. N P Weight 82.3 23.6 62 68.4 23.1 23 t* Body length 150.5 18.0 62 141.0 15.5 23 *t* Tail length 203.9 18.2 62 191.2 18.2 23 *** Hind Foot 23.4 2.0 61 21.5 3.9 23 tt Ear length 28.1 2.4 61 26.5 3.0 23 *tt

PAGE 48

40 1 I r ~i I — — r T 1 i r Jun Jul AugSepOctNovDec JanFebMar AprMay Jun Jul AugSepOct Number of Captures Rainy Season Limits Figure 2-9. Number of captures of Mj. jncana according to monthly trapping period. Straight lines enclose boundaries of rainy season (September to February).

PAGE 49

41 I • I I • • I • • I — 'I I I I Jun Jul AugSepOct NovDec Jan Feb Mar Apr May Jun Jul AugSepOct Number of Juveniles ~^~ Mean Rainfall (cm) Figure 2-10. Number of juveniles of M^. incana according to monthly trapping period. Line represents mean daily rainfall (mm).

PAGE 50

42 Jun Jul AugSepOct NovDec Jan Feb Mar Apr May Jun Jul AugSepOct Breeding Males [ZZI Breeding Females -BMean Rainfall (cm) Figure 2-11. Number of breeding males and females of Mj. incana according to monthly trapping period. Line represents mean daily rainfall (mm).

PAGE 51

43 Differential Trapping Success (in %) Males Females Terrestrial Traps Arboreal Traps Figure 2-12. Differential trapping success (1n percentages) of M. incana in arboreal and terrestrial traps.

PAGE 52

44 that the occurrence of breeding males and especially breeding females is tightly tied to mid to late rainy season (Figure 2-11). Marmosa incana can be characterized as a scansorial species, as its use of arboreal and terrestrial traps was approximately evenly split (Figure 2-12). The same result was obtained using behavior upon release as measure of differential substrate use. Approximately 53 % of individuals released climbed trees or vines, while 47 % remained on the ground. There were no statistical differences in substrate use between males and females (chisquare= 3.19; p>0.05). Marmosa incana appears highly insectivorous. Three stomachs analyzed in this study contained only insects, mostly belonging to the orders Coleoptera and Orthoptera. Nowak and Paradise (1983) reported that most members of the genus Marmosa are insect and fruit eaters, although vertebrates are also occasionally consumed. Marmosa microtarsus During the last month of the survey, October of 1986, an adult male of this species was caught in an arboreal trap in the secondary forest of the Rio Doce Park. The species was previously described as being quite abundant in both secondary and primary forests of the Atlantic forest region (Davis, 1947). Marmosa microtarsus is similar in morphology to Marmosa aqilis and it is usually difficult to distinguish them. Morphometric data from this individual are as follows: weight=31

PAGE 53

45 grams; body 1ength=106 millimeters; tail length=148 millimeters; ear length=14 millimeters; hind foot=17 millimeters. Metachirus nudicaudatus This relatively large-bodied terrestrial didelphid was the third most common marsupial species trapped in this study (Table 2-2), The brown four-eyed opossum was present in all six forest plots surveyed. A total of 60 males and 45 females were caught, respectively, 88 and 68 times. The sex ratio for first captures was equal to that of recaptures (1.3:1.0). Adult males were, on average, larger than females, with most biometric parameters measured proving sexually dimorphic (Table 2-9). Males were, on average, 1/4 heavier than adult females. Even though the species was relatively common throughout the year, a trapping success peak was observed following the rainy season (Figure 2-13). This peak probably coincides with the onset of breeding in midrainy season and the beginning of the dry months. Both lactating females and males with functional abdominal glands were frequently caught at this time (Figure 2-14). Metachirus nudicaudatus may also be able to produce a second litter in the same year. One female had a litter in March and the second in October. The number of pouch young ranged from 5 to 9, with an average of 7.2 young (Table 2-10). For the two females in which

PAGE 54

46 Table 2-9. Biometric data for adult male and female Metachlrus nudicaudatus Weight is given in gms, and other body measurements in mm. P refers to p-value associated with analysis of variance for sexual dimorphism (n.s.=p>0.05; *=p<0.05; **=p<0.01; ***=p<0.001). MALE Mean St.Dev. N FEMALE Mean St.Dev. N Weight Body length Tail length Hind Foot Ear length 352.7 252.0 324.1 44.9 37.4 94.0 28.7 30.5 3.1 3.9 51 51 51 51 50 280.8 232.1 315.9 42.3 36.5 57.1 25.4 34.0 3.3 2.9 33 32 31 32 32 n.s. n.s.

PAGE 55

47 1 I I I i I I I I I I I I I r Jun Jul AugSepOctNovDecJanFebMar AprMay Jun Jul AugSepOct Number of Captures Rainy Season Limits Figure 2-13. Number of captures of M^ nudicaudatus according to monthly trapping period. Straight lines enclose boundaries of rainy season (September to February).

PAGE 56

48 Jun Jul AugSepOct NovDec Jan FebMar Apr May Jun Jul AugSepOct Breeding Males EZZI Breeding Females -BMean Rainfall (cm) Figure 2-14. Number of breeding males and females of M^ nudicaudatus according to trapping period. Line represents mean daily rainfall (mm),

PAGE 57

49 Table 2-10. Litter sizes and sex ratios of Metachirus nudicaudatus pouch young. ID Number Number of Number of Total Sex Ratio Female Young Male Young Young M:F 271 ^ 9 322 5 338 8 346 3 2 5 0.7:1.0 338* 6 3 9 0.5:1.0 Means 7.2 0.6:1.0 Second consecutive litter.

PAGE 58

50 young were old enough to be sexed, there was a biased sex ratio towards females. Nowak and Paradise (1983) regard brown four-eyed opossums as being arboreal, but the species was only once in 156 captures caught in an arboreal trap. Furthermore, only once was an individual observed to use aerial substrate upon being released. In fact, the large and non-graspable hind feet and clumsy behavior on above ground support do suggest a complete terrestrial life for the species. Using a spool-line device, Miles et al. (1981) also found Metachirus nudicaudatus to be completely terrestrial. Terborgh et al. (1984) also regard gray four-eyed opossums as a species confined to the ground. Male and female brown four-eyed opossums did not differ in persistence times in traplines, with both sex classes averaging 1.7 months. Two males and two females were also recorded in traplines over a nine month period (Figure 2-15). Distances travelled between successive trapping were among the lowest for marsupials, averaging 36.7 meters (maximum=120 meters; N=6) for males and 40 meters (maximum=80 meters; N=7) for females. Monodelphis americana Short-tailed opossums have a fairly wide distribution in Brazil (Streilein, 1982). Only one female was caught in an auxiliary line established in the primary forest at Fazenda Montes Claros, for the purpose of collecting voucher

PAGE 59

51 Frequency 4 6 Number of months in lines 10 Males -^— Females Figure 2-15. Frequency of individuals of M^ nudicaudatus persisting within traplines.

PAGE 60

52 specimens. The capture site did not differ physiognomically from the regular traplines, and therefore the species might also occur consistently at this site. However, it is felt that trapping methods were not proper to adequately represent the species. Its small body size and foraging habits on the forest litter may have accounted for its low representation in the sample (Davis, 1947). The biometric data on this individual are as follows: weight=19 grams; body length=92 millimeters; tail length=46 millimeters; hindfoot=16 millimeters; ear length=13 millimeters. Comparisons with data provided by Nowak and Paradise (1983) indicate that this individual was probably a juvenile. Philander opossum Gray four-eyed opossums were trapped in all forest plots, except at the Rio Doce Park. Fourteen individuals were caught during the course of this study. The occurrence of this large didelphid marsupial is apparently tied to the presence of standing or running water (Davis, 1947; Handley, 1976; Nowak and Paradise, 1983; Alho et al., 1986). As only a few transects occurred close to streams, this may explain the low trapping success for this species. As with other species, trapping success was much higher during the dry season (Figure 2-16). While only three individual females were recorded, these had a much higher recapture rate; 11 males were caught 17 times, while 3 females were captured in 12 different occasions. The longest

PAGE 61

53 12 10 4 2 H w/j///MMMmmMMyjM/ rm! i y~r^ ^^7=OsdIkLj. — I 1 — — I I I 1 1 i I I I I I I I I I Jun Jul AugSepOctNovDecJanFebMar AprMayJun Jul AugSepOct Number of Captures Rainy Season Limits Figure 2-16. Number of captures of P^ opossum according to monthly trapping period. Straight line encloses boundaries of rainy season (September to February).

PAGE 62

54 persistence times on the traplines were 4 and 5 months, achieved by two adult females. All females caught were lactating, and two still had pouch young. Nowak and Paradiso (1983), based on several studies, concluded that P^ opossum breeds aseasonally. However, the small sample size of the present study does not allow any conclusions as to seasonality of reproduction. One female had pouch young in February, while the remaining were caught lactating in August and September. Each had 5 young attached to the nipples. The figure given by Davis (1947) in the Atlantic forest was an average number of pouch young of 4.5, with a maximum of seven, while litter sizes in Nicaragua were found to be slightly larger (Phillips and Jones, 1965). For one litter which sex could be determined, there were four female and one male young. Even though the small sample size precluded statistical analysis of sexual dimorphism, adult males were on average over 30 % heavier than females (Table 2-11). As noticed by Nowak and Paradiso (1983), in this study the species also proved to be primarily terrestrial, with only 17 % of captures in arboreal traps, and 7 % of individuals climbing trees after being released. Based on field observations, however, it is felt that gray four-eyed opossums are able, if needed, to efficiently make use of arboreal substrate. The same was suggested by Miles et al. (1981) and Crespo (1982).

PAGE 63

55 Table 2-11. Biometric data for seven males and three females of gray four-eyed opossum, Philander opossum Weight is given in gms. and other body measurements in mm. MALE FEMALE Mean St.Dev. Mean St Dev Weight 394.9 98.7 295.0 39.7 Body length 282.0 51.3 285.0 48.2 Tail length 325.5 21.6 295.0 18.0 Hind Foot 45.3 2.6 40.3 4.0 Ear length 32.7 3.0 29.3 3.1

PAGE 64

56 Table 2-12. Biometric data for adult Akodon cursor Weight is given in gms, and other body measurements in mm. P refers to pvalue associated with analysis of variance for sexual dimorphism (n.s.=p>0.05; *=p<0.05; **=p<0.01; ***=p<0.001 ) MALE FEMALE Mean St.Dev. N Mean St.Dev. N P Weight 48.1 8.1 14 45.5 6.8 8 n.s Body length 107.8 10.1 13 103.0 14.3 8 n.s. Tail length 98.8 9.9 12 94.9 10.4 8 n.s. Hind Foot 26.9 1.3 13 25.9 1.3 8 n.s. Ear length 18.1 1.6 14 18.8 0.9 8 n.s.

PAGE 65

57 Abrawayomys ruschl This rare monotypic murid rodent, endemic to the Atlantic forest of eastern Brazil, is only known through its type specimen, collected in the state of Espirito Santo (Nowak and Paradiso, 1983). A single adult male, with descended testes, was collected at the secondary forest of the Rio Doce State Park in January of 1986, for which measurements are as follows: weight=63 grams; body=128 millimeters; tail length=146 millimeters; hind foot=31 millimeters; ear length=20 millimeters. Akodon cursor The genus Akodon comprises over 40 species, and A. cursor is among the largest. Sexual dimorphism is lacking in A. cursor (Table 2-12), although males weighted more than females in another study (Nitikman and Mares, 1987). This species occurred frequently in the Rio Doce Park secondary forest, but was also occasionally caught at other sites, especially in some auxiliary lines which were located in humid grasslands. Within the traplines, a total of 29 individuals were caught 53 times (Table 2-2), 27 of which were trapped at the Rio Doce Park. Sex ratios for first captures and all captures were, respectively, 2.3:1.0 and 3.4:1.0. Akodon cursor was previously described as being completely terrestrial (Crespo, 1982; Alho et al., 1986; Nitikman and Mares, 1987). However, the species demonstrated scansorial ability in this study, with approximately one-third of trapping success being obtained at arboreal traps.

PAGE 66

58 A distinct peak in population density was found between May and July. Very few individuals were recorded during the rainy season (Figure 2-17). This was attributed to recruitment of young into the population just at the end of the wet and in the dry season. This conclusion is supported by the observation that the number of males with descended testes closely follows that of trapping success, and coincides with mid dry season (Figure 2-18). Gestation and subsequent weaning are close to five weeks (Nowak and Paradise, 1983), and the influence of reproduction activity on population density was readily noticeable in terms of increased trapping success. Akodon cursor appears to be primarily insectivorous. Individuals trapped in both grasslands and forests turned out high amounts of insects in their stomachs, which possibly indicate that the species makes use of insects in their diets in all habitats. The species also makes use of seeds, fruits, and vegetative parts, especially those of the Graminae (Table 2-13). Akodon cursor turnover rates appear very high, with 78 % of individuals only being recorded during one trapping session. Maximum persistence was achieved by two individuals, but for only three and four months. The two individuals for which data on travel distances were available, moved 20 and 40 meters between successive captures.

PAGE 67

59 T 1 1 1 1 1 1 1 I 1 \ 1 — — T r Jun Jul AugSepOctNovDecJanFebMar AprMay Jun Jul AugSepOct Number of Captures Rainy Season Limits Figure 2-17. Number of captures of A,, cursor according to monthly trapping period. Straight line encloses boundaries of rainy season (September to February).

PAGE 68

60 II — ^ ^ I 1 I I i I V I '1' I Jun Jul AugSepOct NovDec Jan Feb Mar Apr May Jun Jul AugSepOct Breeding Males ^h~ Mean Rainfall (cm) Figure 2-18. Number of males of A^. cursor with descended testes, according to trapping period. Line represents mean daily rainfall (mm),

PAGE 69

61 Table 2-13. Stomach contents of four Akodon cursor. PERCENTAGE OF ID Number Insects Leafy Material Seeds Fruits CEM1/10* 47P** 101P* 106P* 33 100 100*** 96*** 33 4 33 Forest trapl Grasslands. *** Coleoptera ines. and ants

PAGE 70

62 Echymys sp Members of this genus are entirely arboreal (Hershkovitz, 1969). This may explain the low success in recording this species in the forests sampled. It is possible that it preferentially occupies higher canopy strata. Davis (1947) never trapped Echymys below 5 meters, while Miles et al. (1981) found nests of Echymys chrysurus in tree cavities located at the canopy level. Only one adult male was caught on the primary forest of Fazenda Montes Claros in July of 1986, in an arboreal trap. The voucher specimen could not yet be identified beyond genus, but may possibly either be Echymys brasiliensis (sensu Moojen, 1952) or a species which has not been described (L. Emmons, pers. comm. 1988). This specimen had the following measurements: weight=225 grams; body length=215 millimeters; tail length=205 millimeters; hind foot=38 millimeters; ear length=14 millimeters. Nectomys squamipes This semi-aquatic rat was occasionally trapped in the vicinities of small streams or flooded areas within forests. Alho et al (1986) obtained 93 % of all the captures of Nectomys squamipes in the Cerrado region on flooded areas. A few individuals in the present study were found over 500 meters from any source of water, suggesting that the species may occasionally exploit non-aquatic habitats. The three stomachs analyzed contained only vegetative material, two exclusively fruit pulp.

PAGE 71

63 Nectomys squamlpes is a fairly large-bodied rodent, with males attaining over 250 grams (Table 2-14). A total of eight males and one female were recorded in tree plots, one at each of the study sites. Another individual escaped before sex could be determined. All but one individual was caught during the mid-dry season, i.e., between May and August. Although adapted for semi -aquatic life, N^ squamipes were trapped 38 % of the times in arboreal traps. Nests which were assumed to belong to N^ squamipes were found one meter above ground. Individuals were also captured at stations away from water sources. Two males persisted in the areas for three and four months, respectively. Oryzomvs capito A widespread habitat general ist, this rodent is extensively distributed in South America (Handley, 1976). It is possibly mostly terrestrial, and in the Cerrado region was found more commonly at dense forests (Alho et al., 1986; Nitikman and Mares, 1987). However, it was a rare species in the forests sampled in this study and was represented by only one male and one female at the Rio Doce Park, both caught in September of 1986. All captures were on ground traps. The measurements of these individuals are, respectively: weight=60 and 63 grams; bodylength=122 and 132 millimeters; tail length=115 and 129 millimeters; hind foot= 32 and 34 millimeters; ear length=22 and 21 millimeters.

PAGE 72

64 Table 2-14. Biometric data for five males and one female Nectomys sauamipes. Weight is given in gms, and other body measurements i MALE FEMALE Mean St.Dev. N Mean St.Dev. N Weight 197.0 67.0 5 165 1 Body length 188.4 19.9 5 190 1 Tail length 207.8 24.9 5 210 1 Hind Foot 50.0 2.8 5 50 1 Ear length 22.0 2.2 5 22 1

PAGE 73

65 Oryzomys n1qr1pes This small cricetid rodent commonly occurred at auxiliary traplines located in grasslands and it has wide distribution among Cerrado habitats (Alho et al., 1986; Nitikman and Mares, 1987). It was, however, relatively uncommon in forests. Only 5 individuals, two males and three females, were caught in forest traplines. All trappings were done on midto late rainy season, i.e., between January and April. Two of these captures were on arboreal traps, although it is felt that the species is certainly more terrestrial and/or scansorial, as it was also observed by Crespo (1982). Alho and Pereira (1985) in Cerrado gallery forests and Veiga-Borgeaud (1982) in the Atlantic forest region determined that Oryzomys nigripes (^ eliurus ) makes extensive use of low shrubs, with nests located about 1 meter from ground. Stomach contents revealed high frequency of insects, complemented by fruit, seeds and leafy material (Table 2-15). Barlow (1969, in Dalby, 1975), and Crespo (1982) also found insects as part of 0. nigripes diet, even though in lower proportions. Biometric data are presented in Table 2-16. Reproduction occurs throughout the year, albeit it may increase in frequency at some periods. Two females, one caught in February and the other in August, both had 4 fetuses, coinciding with two major reproductive peaks observed by Veiga-Borgeaud (1982). A third female, trapped in April, had 5 implanted fetuses.

PAGE 74

66 Table 2-15, Stomach contents of Oryzomys n1qr1pes PERCENTAGE OF ID Number Insects Fruit Leafy Material Seeds 105P 100 56P 20 80 CEM3 6 90 4 CEM2 30 12 58

PAGE 75

67 Table 2-16. Biometric data for two males and three females of Oryzomys nlqripes Weight is given in gms, and other body measurements in mm. MALE FEMALE Mean St.Dev. N 2 Mean St.Dev. N Weight 19.5 0.7 18.3 2.9 3 Body length 87.5 0.7 2 83.0 5.3 3 Tail length 126.0 4.2 2 116.8 6.1 3 Hind Foot 23.5 0.7 2 24.0 1.0 3 Ear length 16.0 0.0 2 17.3 0.6 3

PAGE 76

68 Oryzomys subflavus Only one female of this otherwise fairly common Cerrado species (Melo, 1977; Alho and Pereira, 1985; Alho et al., 1986) was caught in an arboreal trap in the forest traplines. The measurements of this individual are as follows: weight=92 grams; body length=165 millimeters; tail length=173 millimeters; hind foot=34 millimeters; ear length=25 millimeters. Oryzomys trinitatis (=concolor) This cricetid was the most frequent rodent found in the forests sampled, being present at all sites. It is widely distributed, from Costa Rica to Paraguay (Nowak and Paradiso, 1983), and found mostly in forested habitats (Alho and Pereira, 1985; Nitikman and Mares, 1987). A total of 84 individuals were captured 99 times (Table 2-2). Due to coat color variation, most individuals were prepared as skins. Therefore, recapture figures underestimate true recapture rates. Sex ratios at first captures were 1.35:1.00. In contrast to other small mammal species observed in this study, females were slightly heavier and longer than males. Other biometric parameters were found not to be significantly dimorphic (Table 2-17). The species appears to be mostly arboreal. Approximately 74 % of males and 62 % of females were captured in arboreal traps (Figure 2-19). This figure is very close to that observed by Nitikman and Mares (1987) in a gallery forest of Central Brazil.

PAGE 77

69 Table 2-17. Biometric data for adult male and female Oryzomys tr1n1tat1s Weight is given in gms, and other body measurements in mm. P refers to dimorphism ( Pn -value associated s.=p>0.05; *=p<0. with analysis 05; **=p<0.01; of variance for ***=p<0.001). sexual MALE FEMALE Mean St.Dev N Mean St.Dev. N P Weight 63.2 8.2 35 70.0 10.9 28 ** Body length 129.1 8.8 35 136.4 8.1 28 *** Tail length 152.3 10.2 35 159.4 18.2 27 n.s. Hind Foot 30.4 4.1 35 29.1 2.0 28 n.s. Ear length 18.6 3.0 35 19.0 1.8 28 n.s.

PAGE 78

70 Differential Trapping Success (in %) Males Terrestrial Traps Females Arboreal Traps tMnitat^l'in llZZlT'^l T'''^'"^ ^^^^^^^ ^ '"^l^ ^"^ female 0. T^rimtatis in arboreal and terrestrial traps. —

PAGE 79

71 Males and females did not differ in degree of arboreal ity (chi-square=2.40; df=1; p>0.05). Since most individuals were prepared as voucher specimens, sample size was too small to evaluate the differential use of substrate by analyzing behavior upon release. However, the few individuals that were released demonstrated climbing ability. Even though relatively uncommon in the forests of Manu National Park in Peru, Oryzomys trinitatis (=concolor) was listed by Terborgh et al. (1984) as being able to use most all forest strata, from the understory to canopy. Although there were variations in trapping success at a monthly basis, there does not seem to exist any season characterized by high population, in contrast to most other species (Figure 2-20). This might be a consequence of the absence of seasonal breeding. Reproduction seems to take place throughout the year, as breeding males and females were captured in 14 out of 17 months of study (Figure 2-21). The two stomachs available for analysis indicated a high degree of insectivory. The stomach contents of an individual were 100 % insects, while a second had 48 % unidentified insects and coleoptera larvae, and 52 % fruit pulp, Oxymycterus sp. This genus includes semi-fossorial cricetids. The genus is represented in these forests by only one juvenile male and one adult female of an yet undetermined species

PAGE 80

72 16 14 12 10 8 6 4 -\ 2 "1 1 I I 1 I I 1 I I I i I [ I ', 1 Jun Jul AugSepOctNovDecJanFebMar AprMayJun Jul AugSspOct NumDer of Captures Rainy Season Limits Figure 2-20. Number of captures of 0,. tr1n1tat1s according to monthly trapping period. Straight line encloses boundaries of rainy season (September to February).

PAGE 81

73 Jun Jul AugSepOct NovDec Jan FeoMar Apr May Jun Jul AugSepOct Breeding Males CZH Breeaing Females -SMean Rainfall (cm) Figure 2-21. Number of breeding male and female Oj. trinitatis according to monthly trapping period. Line represents mean daily rainfall (mm).

PAGE 82

74 (Table 2-2). Members of this genus are mostly insectivorous (Borchert and Hansen, 1983; Redford, 1984); two stomachs analyzed in the present study yielded 100 % insects, especially Coleoptera larvae and ants. Species of the genus Oxymycterus are more commonly found in grasslands and inundated savannas (Borchert and Hansen, 1983; Fonseca and Redford, 1984; Redford, 1984), a habit which can account for the rarity of the species in the forests of this study. Biometric data for the male and female are, respectively: weight=57 and 85 grams; body length=124 and 245 millimeters; tail length= 111 and 112 millimeters; hind foot=27 and 30 millimeters. Proechimvs setosus Spiny rats were the third most abundant small mammal in this study (Table 2-2). Only opossums surpassed Proechimys setosus in number of recaptures. Even though species of the genus Proechimys were also found to be quite common in other studies in the Atlantic forest (Davis, 1947; Carvalho, 1965; Avila-Pires and Gouvea, 1977; Botelho and Linardi, 1980; Miles et al., 1981; Fonseca et al., 1987), in the Cerrado (Fonseca and Redford, 1984; Alho et al., 1986) and in the Amazon region (Bishop, 1974; Emmons, 1982; 1984; Terborgh et al., 1984; Malcolm, 1987), spiny rats were surprisingly absent from the Rio Doce Park traplines. This can possibly be attributed to higher predation rate by mammalian carnivores, owls and other predators at the Rio

PAGE 83

75 Doce Park (see Chapter 4). The park has a larger and richer carnivore fauna which exert a larger impact upon P^. setosus populations than the comparatively depauperated predator community of the smaller forest plots. Sex ratios of first captures deviated little from a 1:1 sex ratio, with 56 male and 49 female individuals being trapped. Five individuals escaped before their sex could be determined. Females were recaptured at a slightly higher rate than males, yielding a sex ratio for all captures of 0.83:1.00. Although spiny rats were caught in all 17 months of the study, a distinct mid to late dry season peak in trapping success was notable (Figure 2-22). This may result from the increase in the recruitment rate of juveniles into the population at this time (Figure 2-23). However, reproduction does not seem to be strictly seasonal, as it was also observed by Bishop (1974) in Mato Grosso, Brazil. Pregnant and lactating females were present in all trapping sessions, although an increase in the frequency of breeding individuals could be observed both in mid-rainy season and in mid-dry season (Figure 2-24). A few females had two litters in the same year, coinciding with the above mentioned breeding peaks. There is no sexual dimorphism in this species of spiny rat, as none of the morphological characters noeasured significantly differ between sexes (Table 2-18). Proechimys setosus is entirely

PAGE 84

76 Table 2-18. Biometric data for adult male and female Proechimys setosus Weight is given in gms, and other body measurements in mm. N=number of individuals, St. Dev.=Standard Deviation. P refers to pvalue associated with analysis of variance for sexual dimorphism (n.s.=p>0.05;*=p<0.05; **=p<0.01; ***=p<0.001 ) MALE FEMALE Mean St.Dev. N Mean St.Dev. N P Weight 270.7 36.2 43 259.6 43.4 41 n.s. Body length 200.2 14.2 42 196.5 18.9 40 n.s. Tail length 212.3 13.9 39 208.3 15.4 39 n.s. Hind Foot 51.2 2.1 43 50.7 4.6 40 n.s. Ear length 29.2 2.1 43 30.6 2.8 40 n.s.

PAGE 85

77 Table 2-19, Analysis of seven stomachs of Proechitnys setosus. PERCENTAGE OF ID Number Fruit Insects Seeds 39P 81 19 93P 64 36 95P 100 90P 100* 91P 100 92P 100 79P 47 53 Termites and Coleoptera larvae.

PAGE 86

78 "1 — ^ — I 1 1 — — t 1 1 1 1 1 — — I r Jun Jul AugSepOctNovDecJanFeoMar AprMay Jun Jul AugSepOct Number of Captures Rainy Season Limits Figure 2-22. Number of captures of P^ setosus according to monthly trapping period. Straigh line enclose boundaries of rainy season (September to February).

PAGE 87

79 1 I I II I r Jun Jul AugSepOct NovDec Jan FebMar Apr May Jun Jul AugSepOct I I Number of Juveniles ~^~ Mean Rainfall (cm) Figure 2-23. Number of P^ setosus juveniles according to monthly trapping period. Line represents mean daily rainfall (mm).

PAGE 88

80 Jun Jul AugSepOct NovDec Jan Feb Mar Apr May Jun Jul AugSepCct iBreeOing Males IZZlBreeOing Females -3Mean Rainfall (cm) Figure 2-24. Number of breeding male and female P^ setosus according to monthly trapping period. Line represents mean daily rainfall (mm).

PAGE 89

81 Hreauen ij 3 4 5 6 7 8 9 Mumter of months in iines lU n 12 Males hemates Figure 2-25. Frequency of individuals of P^ setosus persisting within traplines.

PAGE 90

82 terrestrial, lacking any scansorial ability. All 239 spiny rat captures were conducted in terrestrial traps. Analysis of stomach contents indicates that spiny rats are primarily frugivorous, but also make opportune use of insects and seeds (Table 2-19). Emmons (1982) found Proechimys to be highly frugivorous, with insects being frequent in stomach samples. Although, in average, females persisted in traplines more than males (respectively, 2.7 and 2.0 months), there were no significant differences between sexes (chi-square=0.58, df=1, p>0.05). The records for persistence times were obtained by two females who stayed, respectively, 10 and 12 months within traplines (Figure 2-25). Males and females also did not differ in mean travelled distances between successive captures with, respectively, 98.7 (maximum=220 meters) and 87.5 meters (maximum=240 meters). Rhypidomvs mastacalis This arboreal rat (Hershkovitz, 1969) was only caught once in this study, in the month of September. It is found mostly in moist, forested habitats (Davis, 1947; Dietz, 1983; Fonseca and Redford, 1984; Alho et al., 1986), even though it can also invade households (Nowak and Paradiso, 1983; J. Stal lings, pers. comm., 1988). The adult male measurements are as follows: weight=72 grams; body length=137 millimeters; tail length=161 millimeters; hind foot=27 millimeters; ear length=18 millimeters.

PAGE 91

83 Discussion Seasonality and Resource Use One of the most striking phenomena cormnon to several small mammals of the western slopes of the Atlantic forest is the occurrence of seasonal reproduction. With the exception of the two most abundant rodents ( Oryzomys trinitatis and Proechimys setosus ) all four species with sufficient sample size to allow analysis proved to concentrate breeding in the late dry season and into the early to mid-wet season. These were all marsupials. The evidence for these patterns derives from both the increase in trapping success starting in late wet season, represented mostly by the addition of juveniles, as well as the observation of adults in breeding condition just previous to that period. Furthermore, the number of first-time captures of all small meimmals is significantly and negatively correlated with average monthly precipitation (r=0.55; p<0.02), which indicates that recently weaned juveniles are recruited into the population at the end of the rains. Akodon cursor is also a seasonal breeder, but unlike marsupials it concentrated its reproduction in the dry season. Davis (1946) and Laemmert et al (1946) observed a pattern of cyclic reproduction among several Brazilian Atlantic forest small mammal species. Breeding was concentrated in the late dry winter

PAGE 92

84 and in the wet summer months, albeit less marked than in the present study. The possible reason for this may be that both these studies were in the eastern side of the Brazilian coastal mountains, where the dry season is shorter and less severe (Hueck, 1972). These two studies have also revealed the lack of seasonality for some rodent species, especially ones from the genus Proechimys, which probably breed year-round throughout its range. A large number of African rodents seem also to closely tie the onset of reproductive activities to just before the end of the wet season (Delany, 1986). Cerrado species appear to be equally divided between seasonal and year-round breeders (Dietz, 1983; Alho, 1982; Alho and Pereira, 1985). In Panama, over 50 % of mammals are seasonal breeders (Fleming, 1973), although reproduction is concentrated in the three months of the dry season, which results in most young being weaned at the beginning of the rains. Other things equal, reproductive output and juvenile survival of small mammals would be maximized if reproduction occurs when the environment is best in terms of resource availability (Fleming, 1975b). In fluctuating environments the energetic costs of pregnancy and lactation, coupled with the needs of adequate resources for newly weaned young, should pose constraints on the timing of reproduction. Lee and Cockburn (1985) state that all tropical didelphids are seasonal breeders and the timing of

PAGE 93

85 reproduction is linked with availability of food. Marsupial food resources, in turn, are tied to the fluctuation in rainfall regimes of seasonal environments (Charles-Dominique, 1983). Information on small mammal diets in this study is not sufficient to provide an accurate and longitudinal picture of diet composition. However, it is important to notice that all but the semi-aquatic Nectomys squamipes were observed to consume insects in variable quantities. All didelphid marsupials appear to be predominantly insectivorous (Nowak and Paradiso, 1983), and even large-bodied species such as Didelphis marsupialis and Philander opossum depend heavily on insect prey (Charles-Dominque, 1983). Metachirus nudicaudatus Marmosa spp. and Caluromys philander have also been listed as being consumers of ants and termites (Redford, 1987), although the latter species may rely more heavily on fruits (Atramentowicz, 1982). Results of studies on insect population fluctuations in the neotropics vary (see Elton, 1975; Bigger, 1976; Janzen and Schoener, 1968; Janzen, 1973; Wolda, 1978), but most authors agree that samples obtained in wet season months are larger than comparable ones collected during the dry months (Davis, 1946; Wolda, 1978; Smythe, 1982; Charles-Dominique, 1983). Therefore, the trends of this study certainly support the notion that small mammal reproduction, especially that of marsupials, is influenced by availability of insect prey. However, until the variation in

PAGE 94

86 the cycles of prey populations across the year is described, explaining timing of reproduction of small mammals as a consequence of abundance of insect prey in the late wet to early rainy season will remain speculative. Fruits and seeds were the other frequent items present in the stomachs of small mammals in this study, although it is felt that rodents may depend more heavily on these resources than the more insectivorous marsupials. If the number of trees with fruits can serve as a measure of food resource availability for small mammals, the timing of reproduction of most small mammal species seems to track that of resources. In the Brazilian Atlantic forest asynchroniously fruiting trees may be found at every month of the year (Davis, 1946). Nonetheless, fruit production of several plant species of seasonal neotropical environments has been shown to be influenced by precipitation, usually peaking just prior to the start of the rainy season (Foster, 1982; Smythe et al 1982; Charles-Dominique, 1983). A second peak in fruit productivity can also be present at the end of the rainy season (Davis, 1946). A preliminary study conducted in the Rio Doce Park (CETEC, 1981) also indicated that there are tree species flowering and fruiting throughout the year, and also that there is a slight peak in the number of fruiting species in September and October, i.e., in the early rainy season.

PAGE 95

87 I expected that marsupials, which use invertebrates to a larger extent than rodents, should display a higher degree of seasonality due the more fluctuating nature of their food resources. This was observed in the present study. Rodents, on the other hand, can probably rely on resources which are more seasonably stable in the Atlantic forest, such as fruits, seeds and leaf tissue, and thus can reproduce throughout the year. The two most common rodents, Proechimys setosus and Oryzomys trinitatis do not show seasonality in their reproductive patterns. Since they have also shown to be able to use and at some periods heavily rely on insects, the consequent larger resource spectrum should place lower limits on the reproduction of rodent species. While some small mammals may exploit resources on the forest floor, most of the tropical forest productivity is located above ground (Eisenberg and Thorington, 1973). Therefore, even for primarily terrestrial species, some level of scansorial ability should prove advantageous. There was a high degree of overlap in substrate use among small meimmal species in this study. Twelve species are shown to use both aerial and ground substrate regularly, even though their relative degrees of arboreal ity varied. Among marsupials, only Metachirus nudicaudatus was confined to the ground level of the forest. Only four of eleven rodent species can be safely classified as terrestrial, while the remaining are at least marginally arboreal. This is consistent

PAGE 96

88 with descriptions of other neotropical small mammal fauna, where the majority of species show some level of climbing ability (see August, 1984; Fonseca et al 1987; Nitikman and Mares, 1987). It should be stressed that these findings do not imply that arboreal ity is the dominant mode of life among Atlantic forest small mammals. Of the species caught in the present study, only Caluromys philander Marmosa cinerea Marmosa agilis Oryzomys trinitatis Rhipidomys mastacalis and Echymys sp. can be regarded as predominantly arboreal. It does, however, indicate the widespread ability of a greater fraction of the community in exploring a tri-dimensional environment. This may be especially important during certain periods of the year. Charles-Dominique (1983) provided data indicating that the insect faunas of the canopy and of the undergrowth can fluctuate asynchroniously with each other. Therefore, if a particular food resource is undergoing a period of seasonal shortage, those species with versatile habits would be at an advantage. It has been demonstrated before that in the highly seasonal savanna region of central Brazil gallery forests play an important role in maintaining overall mammalian species diversity (Fonseca and Redford, 1984), and become crucial during periods of stress. Part of the reason for this may be linked to the higher productivity of the tri-dimensional gallery forest environment, when compared, during periods of moisture deficit, to that of the cerrado savannas.

PAGE 97

89 Life History Patterns While the occurrence of pouch litters with skewed sex ratios In neotropical marsupial species has not often been reported, it was interesting to find Didelphis marsupial is with litters having a predominance of males. Skewed sex ratios have only been achieved experimentally under a regime of diet supplementation, a procedure which does not always produce the predicted male bias outcome (Austad and Sunquist, 1986). Therefore, it is important that future investigations address this question. The female-biased sex ratio obtained for Metachirus nudicaudatus pouch young is not based on a large enough sample size and therefore remains inconclusive. Three out of four marsupial species, for which sample size was large enough to allow statistical inference, proved sexually dimorphic. Male Didelphis marsupialis has also been found to be larger than females elsewhere in South America (O'Connell, 1979). Other Marmosa species also have larger and heavier males (Nowak and Paradiso, 1983), the same being true for Philander opossum No references have been found to Indicate size differences for Metachirus nudicaudatus except for the present study. While larger males have been usually associated with polygynous species in which males strongly compete for females (Ralls, 1976), there was no evidence of territorial defense by any of the marsupials studied here. However, this does not necessarily preclude males

PAGE 98

90 from actively competing for females at overlapping ranges. Fierce fighting for estrous females has been observed elsewhere in Didelphis marsupial is (Austad and Sunquist, 1986), Since marsupial breeding in this study was usually confined to a certain season, the potential for male-male competition for estrous females is likely. Rodents, on the other hand, are seldom dimorphic (see Eisenberg, 1981), especially the smaller species. In only one species, Oryzomys trinitatis could a size difference be found in this study. In this instance, however, females proved in average to be larger than males. This could be attributed to extra energetic demand placed on pregnant and lactating females, a cost non-existent for males who do not provide parental care (see Ralls, 1976). Moreover, since rodent young do not undergo, unlike marsupials, a teat attachment phase, female placental mammals may increase litter size above that of the teats. The increase in rodent litter sizes may be achieved if the female is larger and better nourished. Marsupial litter sizes, on the other hand, due to the obligatory teat attachment phase, are constrained by the number of teats. Furthermore, marsupials can spontaneously terminate lactation, or more often in smaller species reduce litter size if energetically stressed (Lee and Cockburn, 1985). Young are also born after a very short gestation period, making the reproductive investment minimal. For these reasons no added

PAGE 99

91 security would be achieved by genetically fixed propensity for larger marsupial females. Population Turnover Although the area sampled by traplines was not enough to represent the home range of several species, especially the larger ones, the reduced persistence time of the average small mammal individual was nonetheless striking. Average persistence times ranged from 1.2 months for male Marmosa incana to a maximum of 3.8 months for Marmosa cinerea females, with most species remaining in traplines within the range of 1.7 to 2.7 months. While it is reasonable to assume that some of the disappearances can be attributed to home range shifts (see Nitikman and Mares, 1987) and/or juvenile migration and dispersal (Lidicker, 1975), predation might also play an important role in increasing turnover rates among tropical small mammals. Monthly turnover rate for the small cricetid rodent Akodon cursor approached 80 % in this study. Several authors have suggested that predators might cause variations in local abundances (August, 1983), and sometimes they do take a large percentage of small mammal standing biomass (Hershkovitz, 1969; Pearson, 1985; Emmons, 1987; Sunquist et al., 1987). Predation in traps was observed in this study on several occasions and it is felt that it may play a major role in the structure of these communities (Chapter 4). A large number of

PAGE 100

92 mammalian predators, as well as owls, have been observed predating on several of the marsupials and rodents trapped in the present research, and I suspect that predators are responsible for a large fraction of the observed turnover rates on these small mammal communities.

PAGE 101

CHAPTER 3 SMALL MAMMAL SPECIES DIVERSITY IN BRAZILIAN TROPICAL PRIMARY AND SECONDARY FORESTS OF DIFFERENT SIZES Introduction During recent years considerable attention has been given to the Brazilian Atlantic forest ecosystem, one of the most threatened in the world (Mittermeier et al., 1982; Fonseca, 1985a). Estimates of remaining forest area indicate that less than 5 % of the region has some form of forest cover (Victor, 1975, in Almeida and Rocha, 1977; Fonseca, 1985a). It is possible that less than 1 % of the Atlantic forest region of Brazil remains in a relatively undisturbed form (Mittermeier et al., 1982). Previous studies have focused mainly on the highly endemic and diverse primate fauna (Coimbra-Filho and Mittermeier, 1977) and on birds (Willis, 1979). On the other hand, the bulk of mammalian diversity, represented principally by rodents, marsupials and bats, remains largely unknown. This omission must be rectified because deforestation and other forms of habitat modification, such as lumbering and fuelwood extraction, continue to alter the mammalian communities. The vast majority of forest patches left in the region are small and disturbed. Under these circumstances second 93

PAGE 102

94 growth forests and their animal communities acquire a fundamental importance in any conservation scheme for the region and should be actively researched. The present study will examine the effects of area size and habitat structure on the diversity of non-volant small mammals (rodents and marsupials) in several forested areas of the Atlantic region of Brazil. Rodents and marsupials are assumed to represent an identifiable major guild within the larger mammalian community. They are all relatively short-lived, have small body sizes (usually less than 1.5 kg), are mostly nocturnal, have generalized diets and are not confined to higher strata of the forest. The objective of this study was to examine how (a) modifications in habitat structure and heterogeneity and (b) the size of a forested area affect small mammal species richness and diversity. Species richness is defined only by the absolute number of species in a community. Species diversity is defined by both the number of species and the evenness of the contribution of each to the total numbers within the community. Diversity is shown to increase with the decrease in the probability that two individuals drawn from the community at random will belong to the same species. Said differently, diversity is a measure of average rarity (Pielou, 1974; Patil and Taillie, 1983). Local species richness and diversity in the tropics have been attributed to various factors (Leigh, 1975; Prance, 1982;

PAGE 103

95 Bourliere, 1983). Leaving aside questions of historical biogeography (e.g., continental drift, patterns of speciation, paleoclimatic cycles, etc.), several explanations have been offered as to the proximate mechanisms that lead to the maintenance of species diversity or to the achievement of lower rates of extinction in localized areas. Among these are investigations of the effects of environmental stability or moderate levels of disturbance on the coexistence of species (Connell and Slatyer, 1977; Connell, 1978). Habitat structure and heterogeneity have been proposed as primary factors promoting small meuranal species diversity in tropical areas (August, 1983, 1984; Fonseca and Redford, 1984). The theory of island biogeography had, previous to that, also called attention to the effect that area size has on the number of species of equilibrium communities (MacArthur and Wilson, 1967; MacArthur, 1972). According to this framework, the number of species in an area would be the result of extinction/colonization dynamics. The importance of the theory of island biogeography lies in that it was the first quantitative framework formulated to explain area-size relationships which were long-known through empirical evidence (for a review, see Darlington, 1957). Williams (1964) has offered an alternative explanation for the area-size relationship by hypothesizing that the observed increase in

PAGE 104

96 diversity with the size of the area reflects the addition of distinct habitats and its species components to the samples. It is also possible to survey for the causes for the presence or absence of particular species from local areas basing the investigations on the microhabitat characteristics of the species and their competitive interactions within the communities (Dueser and Shugart, 1978; Hallett et al 1983; Dueser and Porter, 1986; see also Chapter 4). Although that is a valid question, the main purpose of the present study was to find predictor variables that can be instrumental in explaining variation in species diversity of small mammal communities as a function of habitat structure and area size, as well as at the same time conveying biological meaning. Only one study has dealt quantitatively with the effects of the reduction of size and the isolation of patches on small mammal corranunity characteristics in evergreen neotropical forests (Malcolm, 1987). Other studies have provided data on small mammal ecology of several tropical ecosystems, from savanna type vegetation to evergreen forests (Fleming, 1971; 1975a; Alho, 1981; Fonseca and Redford, 1984; Alho et al., 1986; Emmons et al., 1983; August 1983, 1984; Delany, 1986; Nitikman and Mares, 1987; Isabirye-Basuta and Kasenene, 1987; Lacher and Alho, in press; Lacher et al., in press). These areas, however, were mostly characterized by comparatively lower small mammal diversity.

PAGE 105

97 Communities of small mammals in the Brazilian Atlantic forest, on the other hand, are unusually rich in number of species (Davis, 1946; 1947; Moojen, 1952), thus allowing comparisons of diversity among qualitatively different areas. Materials and Methods The Region This study was conducted at three sites, all located on the western slopes of the Brazilian Atlantic forest. Even though the western slopes of the Atlantic coastal region of Brazil are widely recognized in the literature as belonging to the major Atlantic forest formation primarily because of faunistical similarities (Muller, 1973; Mittermeier et al., 1982; Fonseca, 1983), some botanists draw a distinction between the eastern and the western slopes (Hueck, 1972; Eiten, 1974; Alonso, 1977). Hueck (1972) refers to the area that contains the study sites of the present research as the subtropical forest of eastern and southern Brazil. The major distinction between the eastern and western slopes is due to a rain shadow which causes the western slopes to receive much less rainfall than the coastal slopes. The climate is characterized by a more distinct dry season, and the vegetation possesses a greater percentage of tree species showing partial deciduity during the dry season. During the rainy season, however,

PAGE 106

98 the forest is physiognomical ly very similar that found in the more wet areas. Primary forest in this region will generally have its canopy at about 20 meters and is botanical ly very rich (Hueck, 1972), with a well developed mid-story. Study Sites All three study sites are within a maximum of 300 km from each other and are located in the state of Minas Gerais, Brazil (See Figure 2-1). All sites were of different sizes. In each, two forest patches were selected: one primary, or at least in its late successional stages, and another secondary, in all cases approximately 20 years old following clearcuting or an extensive fire. Each of the forest patches was then monitored by a 17-month live-trapping regime for small mammals. The first site is Fazenda Esmeralda, a cattle and sugar cane farm in the county of Rio Casca, Minas Gerais (see Figure 2-1, in Chapter 2). The property is bounded by the right bank of the Rio Doce river and is composed of a series of valleys between several hills with a maximum of 150 meters in altitude. The farm is 4,800 ha., of which very little remain under forest. The bulk of the deforestation took place between 1964 and 1974. One approximately 80 ha. patch, known as "Lagoa Fria," was in late succession and was selected as the primary forest site of the small size category. It is designated in this study as SM-PR. There are other small areas

PAGE 107

99 mainly covered by secondary vegetation scattered on the crowns of the several hills that occur within the farm. One of these, a 60 ha. patch approximately 15-20 years old (according to the farm owner and by estimates made through physiognomic comparisons with other patches of known age) was selected. It will be hereon known as SM-SC Fazenda Montes Claros is located between the counties of Ipanema and Caratinga, Minas Gerais. This 1,200 ha. farm has been owned by Mr. Feliciano Abdalla since 1944. Forest covers approximately 860 ha. of the farm. The terrain is mountainous, varying between 320 and 580 meters above sea level. There are two main sections with vegetation considered as primary forest, one of which was selected, MD-PR Another location which was under coffee and cattle pasture a few decades ago was selected as the secondary site of the middle size category, MD-SC As the primary and secondary forests shared a common boundary, have mostly the same small mammal fauna and can potentially exchange species and individuals, all samples are considered as coming from a forest patch of similar size, or 860 ha. The Rio Doce State Park is under the protection of the State of Minas Gerais Forest Institute (lEF). Created in 1944, the park has an area of 35,973 ha. Due to its size the park is certainly the most important conservation unit for the Atlantic forest formation in the state of Minas Gerais. A major characteristic of

PAGE 108

100 the park is its network of 42 bog lakes. The park is almost completely covered with forest, but due to frequent fires during the 1960s perhaps only about one-third is primary. The Rio Doce Park still contains the largest tract of primary forest left in the state. One section with tall primary forest, known locally as "Campolina," was selected and will be referred to as LG-PR A second site that burned extensively in 1967 (in the "Hotel" region of the park), LG-SC was chosen to represent the secondary forest of the large size category. For the same reasons as at Fazenda Montes Claros, the trapping samples from each of the forest types are assumed as coming from a 35,973 ha. forest. Given its large size, I assume that the park represents a true sample of the original fauna! assemblage of the region. Trapping In each of the six forest plots selected for the study three 300 meters long parallel lines were cut. Lines were separated by 100 meters. Therefore, traplines sampled an area of 6 ha. (300 meters x 200 meters). In each line 16 trapping stations located 20 meters apart were established. Traps were all confined within a circle of 3.5 meter radius measured from the center of station. Each trap station had one 48 x 15 x 15 centimeter Tomahawk squirrel-size live trap (Tomahawk Live Trap Co., Tomahawk, Wisconsin) located on the ground. At every other station another

PAGE 109

101 Tomahawk of the same size was wired above ground on branches or vines between 1 and 4 meters high. In addition to these traps, every other station possessed a mouse-sized collapsible Sherman trap (H. B. Sherman Traps, Inc., Tallahassee, FL), with alternation of ground and tree traps. Moreover, the two outermost transect lines had, at every other station, a large 80 x 30 x 30 centimeter wire home-crafted live trap. Therefore, each outer line possessed 16 ground Tomahawk traps, 8 tree-bound Tomahawk traps, 4 ground and 4 tree-bound Sherman traps and 8 ground large wire traps. The total for each outer line was 40 traps. The midline did not have large traps, but a total of 24 Tomahawk traps and 8 Sherman traps. In summary, every forest site had 48 trap stations disposed into 3 transects of 16 stations each, and a total of 112 permanent based traps. With the exception of Shermans, all traps were closed at the end of each five night trapping session and left in place. Sherman traps were removed, washed, and replaced each month. Trapping took place between June of 1985 and October of 1986. Each forest site was trapped for five consecutive nights a month. During 17 months of consecutive trapping each forest plot accumulated a total of 9,520 trap nights, or a total of 57,120 trap nights for all the six forest plots together. Fresh pineapples, oatmeal and a cotton ball soaked with a commercial codfish oil solution were used as baits. Traps were checked every morning for captures and for adequacy of bait, which was replaced as needed.

PAGE 110

102 For each individual captured, the following information was recorded: 1. Species. 2. Location in grid. 3. Individual identification, if already tagged with metal fish tags (Fish and small animal tag, size 1, National Band and Tag Co., Newport, Kentucky). Individuals with positive taxonomic identification were released at station where captured. If a trapped individual was not readily identifiable, it would be preserved for later taxonomic identification. The only individuals which were consistently preserved as skins and skulls were representatives of Oryzomys trinitiatis and, to a lesser extent, Orvzomys fornesi The former show a high degree of variability in skin color and pattern and were considered, during field work, as belonging to more than one species. Whenever possible animals that died in the traps were also preserved. Habitat Variables The characterization of the habitat was conducted in an area of 3.5 meter of radius around the center of each trap station. The variables which were used are slightly modified versions of measurements that have been successfully used in other studies in the description of small mammal habitat (M'Closkey, 1976; Dueser and Shugart, 1978; August, 1983). At each station, information

PAGE 111

103 was recorded on a qualitative sliding scale between (minimum score for a parameter) and 3 (maximum score for a parameter) for the following variables: 1. Percent canopy cover (CC) above trap station. 2. Percent herbaceous cover (HC) around center of trap station. 3. Interconnectedness of canopy (CXC), represented by branch to branch contact or canopy to canopy link through vines. 4. Interconnectedness of midstory (CXM). 5. Volume at canopy height (DVC), representing sparse vegetation, 3 dense vegetation. 6. Volume at midstory height (DVM). 7. Volume of herbaceous vegetation (DVH). 8. Vine density (DV). 9. Epiphyte density (DE). 10. Fallen logs within or in close vicinity of trap station (FL). 11. Litter volume at soil surface (HUM). In addition to these, other variables were quantitatively measured: 12. Overstory height (OH). 13. Herbaceous height, determined as the most representative height category at a trap station (HH).

PAGE 112

104 14. DBH of all woody plants with DBH greater than 3.2 cm. The composite variable SDI is represented by the sum of all DBHs at a trap station. 15. HT, or the average height of all trees within trap station with DBH greater than 3,2 cm. 16. NTR, or number of trees within station with DBH greater than 3.2 cm. 17. NSP, the number of woody plant species at trap station. The composite variable XSP represents the average number of plant species per station at each transect trapline. Statistical Methods and Data Analysis Dependent Variables There were four indices of community structure under study. These reflect different but interrelated measurements of community attributes: 1. Species richness (S), or number of species recorded for each particular trapline at a forest plot during the whole course of the study. 2. Species diversity, as defined by the Shannon-Wiener diversity index H' 3. Total number of individuals (NI) observed at a particular site during the 17 months of trapping, excluding recaptures.

PAGE 113

105 4. Species diversity weighted by the number of individuals trapped (WH) and defined as WH = :^Pi In pi VnT where WH=weighted H' and ni=number of individuals belonging to the ith species in sample. In order to detect significant statistical differences between species diversity index values, analysis of variance was performed, with each of the three traplines at a forest patch constituting a replicate of that particular forest. Analyses for differences between primary and secondary forests, as well as for the different size categories, were performed. Differences between diversity indices with pooled samples which included all three traplines were tested with pair-wise t-tests (Poole, 1974). Species diversity is one of the most commonly measured community structure parameters and has important ecological (Peet, 1974; Pielou, 1974; May, 1975) and management implications (Margules and Usher, 1981; Hair, 1982; Conant et al., 1983). Several indices have been suggested in the measurement of diversity (for a discussion, see Peet, 1974). The most commonly used index in wildlife studies is the Shannon-Wiener H' (Hair, 1982), which is calculated as

PAGE 114

106 n H' = 2 (Pi In Pi), i where pi represents the proportion of individuals captured belonging to the ^^^ species. Given that evenness measures are not sensitive to absolute population sizes of species, they cannot always be used as indicators of underlying community structure or as Indices useful to guide conservation and management decisions. For this reason I developed a weighted version of the index (WH) that takes into consideration the population sizes of the individual species. The rationale for developing this index was to provide for the effect of population size on the probability of local extinctions. In addition to being less vulnerable to demographic/stochastic extinction events (MacArthur and Wilson, 1967; Shaffer, 1981), larger populations also tend to be less susceptible to inbreeding depression and loss of genetic variability (Soule, 1980; 1983). Therefore, large populations are less liable to go locally extinct and this variable should be included in a species diversity index. Measures of species diversity that are composed of number of species and evenness, such as H' and others, carry not only ecological implications but are increasingly used in conservation and management applications (e.g., "Preservation of biological diversity" concept; Usher, 1986). The use of an index that takes into account the number of individuals trapped, in addition to the

PAGE 115

107 relative contribution of each to the total population (such as H'"), is certainly desirable. WH distinguishes between communities that differ in the overall population density, an impossibility with H'. In addition, H' has always the potential to underrate highly productive and rich communities in which a small degree of imbalance in relative proportions of species occurs. The use of a square-root function of number of individuals trapped in the index, instead of its absolute value, serves some useful purposes. The use of an exponentially decreasing function reflects the reduction in the importance that number of individuals should have on the index when species' population numbers approach high levels. As population size increases, the index becomes more sensitive to the number of species and to their overall contribution to the community and less sensitive to population numbers. A basic characteristic of WH is that it is an index highly but not solely influenced by number of individuals. WH increases faster when the increase in total number of individuals is not very skewed toward one or a few species within the community. If this is the case, the index will hopefully preserve some of the evenness component inherited from H'. There are some caveats that should be stressed before interpreting WH values. Ideally, we should be able to maintain the maximum information about H' as possible, while simultaneously assessing population sizes of species. There are some cases,

PAGE 116

108 however, in which the WH values are not useful. A community with a few species, but with extremely high population densities, will yield WHs that are higher than communities which may have many more species, but in which population densities are low across all members of the community. Therefore, it should never be used disassociated from H' and S. Several simulations were performed with the index to examine its behavior under different population size conditions. All simulations that were conducted with WH were also conducted with H' so as to compare them under different circumstances. In addition, I observed the fluctuation of the index values in successive samples as a further effort to estimate the reliability of the diversity measures obtained after 17 months of trapping. Examination of Figure 3-1 reveals that for the study period, H' only becomes stable after 9-10 months of consecutive trapping. Therefore, if the study had been of shorter duration estimates of species diversity would constitute unreliable measures. WH is, on the other hand, sample dependent, i.e., will almost always increase with the addition of new samples. However, the index, as applied to the present data, showed a much more stable statistical behavior (Figure 3-2), for after 8 months of successive samples it kept the existing relationships constant. H*, on the other hand, demonstrated some statistical shifts even after 10 months of

PAGE 117

109 SHANNON-WIENER H' 0.75 5 6 7 8 9 10 11 12 13 14 15 16 17 CUMULATIVE SAMPLES LG-SC MD-PR MD-SC LG-PR SM-SC SM-PR Figure 3-1. Evolution in the behavior of the Shannon-Wiener H' index with the increase in the number of cumulative samples.

PAGE 118

110 WH INDEX 5 6 7 8 9 10 11 12 13 14 15 16 17 CUMULATIVE SAMPLES LG-SC MD-PR MD-SC LG-PR SM-SC SM-PR Figure 3-2. Evolution in the behavior of the weighted version of the Shannon-Wiener index (WH), with the increase in the number of cumulative samples.

PAGE 119

111 sampling, even though their relational trend was kept consistent throughout the last 7 months of the study. Figures 3-3 and 3-4 illustrate a similar comparison examining the proportional change of indices from a previous saimple to the next sample throughout the study period. After 9 months of consecutive trapping, H' tended to stabilize, generally changing less than 5 % from the previous to the next index value. With WH, even though it takes roughly the same time to reach a point of steady fluctuation, the changes can reach up to 10 % from one sample to another, or twice the rate of H'. An additional difference is that, except with a very small number of samples, WH will always increase with further samples, while H' can grow negatively from a former to a latter sample. Another simulation that demonstrates the stable relationships of WH and its value as a measurable feature of these small mammal communities can be found in Figure 3-5. The graph shows the behavior of the index if we insert actual species richness and proportions to variable hypothetical population sizes. What the method provides is a heuristic way of looking at these different communities were their overall population sizes different for whatever reason than the actual values. We can then, to a certain degree, isolate the desirable information provided by the Shannon-Wiener diversity index, without losing the "population size" perspective.

PAGE 120

112 ri>_'i^i^'i injiiai ^-viiai (j'3 U.M. 4 O 6 o ^ iu 11 lo i^ io lo iv.u-rr, .-— PP 1 o o Q cj Q I wp -^ o rr. ri i Q Q SM-PR Figure 3-3. Rate of change in Shannon-Wiener H' index with the number of successive samples.

PAGE 121

113 r .,^.^ -3 1 ""l-. 5 \^ 0.4 l-\ \ r, ^ i. .\ \ / J. I I ^ iO \— V-4 -^ '^ *^ O O I V ^ ^> Ol 1 I I ^-' i \_/ w ivi (_; -r r. Sfvi-oC L'3-SC Figure 3-4. Rate of change in WH diversity index with the number of successive samples.

PAGE 122

114 v'vH iMDb^ 10 20 30 ^0 50 60 70 80 90 100 110 120130140150160170180190200 POPULATION Sizes (x io) — LG-SC — ^ MD-SG ^^ 3M-SC -aMD-PR — LG-PR ^^ SM-PR Figure 3-5. Hypothetical behavior of the WH diversity index with constant increase in total population sizes of communities. The pi values (see formula in methods section) were generated using real data from this study.

PAGE 123

115 Predictors of Community Structure The first step to Identify predictor variables that explain variation in species richness or number of species (S), number of Individuals (NI) and species diversity (H'), as well as its weighted version (WH), is by the use of correlation analysis and regression analysis. By using correlations I was able to both determine significant associations between dependent and independent variables, as well as intercorrelations between habitat predictor variables. This Is Important in order to avoid problems of colinearity in subsequent multiple regression analysis (Sokal and Rohlf, 1981; Kachigan, 1982). I expected that a large number of habitat variables would be intercorrelated because many similar measures were recorded. After correlation coefficients were determined, multiple regression was applied to the reduced data in an effort to find the predictor variables that would explain significant amounts of variation in the dependent variables (S, NI, H', WH). All habitat variables that were collected as percentages were arcsin transformed and the remaining variables log-transformed to approach homoscedasticity (Sokal and Rohlf, 1981). A SAS-PC (SAS Institute, Cary, NC) statistical package was used in the analyses on MS-DOS (Microsoft Corporation) based machines. Multiple regressions were carried out with the RSQUARE method ("maximum r^ improvement").

PAGE 124

116 Primary Versus Secondary Forests The first step on this analysis was to determine if the habitat variables collected supported the classification of the forests, conducted previously in the field, as primary and secondary. Discriminant analysis was used to classify trapping stations as belonging to either primary or secondary forests. The accuracy with which the resulting discriminant function separates stations in either type of forest provides an estimate of its validity (Kachigan, 1982). Validity was measured by examining both the magnitude of correct classifications and the chi-square values associated with Wilks' lambda at a p-level of 0.05. The analyses were conducted with all data points of dependent variables taken as a whole, as well as separating individual analysis between primary and secondary forests. The objective with the first method was to determine if any existing relationships between community parameters and the independent variables area size and habitat structure would hold regardless of forest structure. The latter method would indicate if the community parameters measured on primary and secondary forests, respectively, respond differently to independent variables.

PAGE 125

117 Area Size Effects One of the main objectives of this study was to determine the effect of area size on species richness, species diversity and total number of small mammal individuals. However, it is often difficult to separate the effects of area size from those arising from habitat structure because of problems of intercorrelation. I approached this problem by eliminating from the analysis habitat variables that were correlated with size. The next step consisted of entering the remaining set of independent variables, as well as area size, as regressors in other models. The following method was emulated from Lacher et al. (in press), a study in which the objective was to separate habitat structural effects from those of species interactions within small mammal communities of the Brazilian flooded savannas (Pantanal). First, I noted the change in r^ when the habitat structure variables were entered in a model that already contained area size as an independent variable. Secondly, the same process was repeated, with the difference that this time area size entered a model already containing the independent habitat variables. The r^ for the full model is equal for either equation regardless of entry order. Habitat structure r^ is the improvement in overall r^ for the first equation, and conversely, area size r^ is represented by the change in the full model r2 This analysis provides an estimate of the percentage of variation in the full regression model that can be attributed to either area size or habitat structure variables.

PAGE 126

118 Results During the course of the study I captured a total of 692 individuals (excluding recaptures) belonging to 19 species, 8 of them marsupials and 11 rodents (Table 3-1). Marsupials were by far the most common small mammals in all sites examined. Marsupials were responsible for 65 % and 64 % of all individuals in primary and secondary forests, and 74 % and 75 % of all individuals in small and large plots, respectively. The only sites in which rodents appear to be almost as abundant as marsupials are in the forests of intermediate size. Approximately 47 % of all individuals in the two forests taken together were rodents. The most abundant rodent species were Proechimys setosus and Oryzomys trinitatis. Four species, Didelphis marsupial is Metachirus nudicaudatus Marmosa incana and Oryzomys trinitatis occurred at all six forest plots sampled. Two others, Proechimys setosus and Philander opossum were found in all sites of intermediate and small sizes but were conspicuously absent from the large sites. Eight of the nineteen small mammal species were only registered in one forest site, usually with two or one individuals. Marmosa cinerea even though it only occurred in the two large plots, was quite common at these sites. Similarly, Akodon cursor was very common in the large

PAGE 127

119 Table 3-1. Number of individuals of each species captured (no recaptures included) in the six forest plots sampled. For description of the sites, see Materials and Methods section. Forest Plot SPECIES Small Medium Large Prim Sec Prim Sec Prim Sec Total SM-PR SM-SC MD-PR MD-SC LG-PR LG-SC Caluromys philander 06 06 Didelphis marsupialis 41 18 40 30 11 04 144 Marmosa aqilis 03 04 07 Marmosa cinerea 08 35 43 Marmosa incana 05 05 15 39 07 56 127 Marmosa microtarsus 01 01 Metachirus nudicaudatus 13 01 16 18 26 31 105 Philander opossum 01 01 09 03 14 Abrawayomys ruschi 01 01 Akodon cursor 01 01 27 29 Nectomys squamipes 02 02 06 10 Oryzomys capito 01 01 Oryzomys niqripes 01 02 02 05 Oryzomys subflavus 01 01 Oryzomys trinitatis 03 15 04 40 04 18 84 Oxymycterus sp. 02 02 Rhipidomys mastacalis 01 01 Echymys sp. 01 01 Proechimys setosus 07 03 80 20 110 Total No. of Individual s 70 49 169 157 57 190 692 Total No. of Species 06 09 09 09 06 14 19* *Total is the number of different species pooled for all sites.

PAGE 128

120 secondary forest, only appearing again in the sample at the large primary and medium secondary plots, with one individual in each. The richest site in number of species and number of individuals was the largest secondary forest, with 190 individuals belonging to 14 species. The poorer forest types in number of species were represented by the small primary forest category and by the large primary category (both with 6 species). The intermediate sizes and the small secondary forest all yielded nine species each. All primary forests had higher H' values than secondaries, and within each class larger plots were more diverse than medium-sized, and these more diverse than the small forest sites. In Table 3-2, species diversity, as measured by H' and WH, is shown for each trapline, as well as the overall score for each particular forest type. The order of Shannon-Wiener H' indices, from lowest to highest, was as follows: Small Primary, Medium Primary, Large Primary, Small Secondary, Medium Secondary, Large Secondary (Figure 3-6). Of these, the first three (all secondary) were significantly different from each other, and each from all primaries (pairwise t-tests; p<0.05). However, for the primary forests, regardless of size, differences were not statistically significant. WH values generally demonstrated the same trend (Figure 3-6): the two forest plots with the highest indices were secondary forests of medium and large sizes. The lowest score was also obtained at the smallest primary forest. However, WH ranked the

PAGE 129

121 Table 3-2. Shannon-Wiener diversity indices (H) and weighted diversity indices (WH) for each of the forest plots and for each transect line within forest plot. T=indices for all samples taken together. SAMPLES FOREST PLOTS AND TRANSECT LINES (TRAPLINES) Small Medium Large Primary Secondary Primary Secondary Primary Secondary (SM-PR) (SM-SC) (MD-PR) (MD-SC) (LG-PR) (LG-SC) H WH H WH H WH H WH H WH H WH tt1 1.09 1.67 1.42 2.59 1.61 5.24 1.80 5.67 1.52 3.21 1.87 5.72 #2 1.50 2.77 1.47 2.51 1.20 4.40 1.68 5.15 1.13 1.81 1.87 5.26 3 1.34 2.96 1.43 2.72 1.30 3.61 1.50 4.18 1.49 3.02 1.93 5.44 T 1.24 4.46 1.67 4.50 1.51 7.87 1.78 8.75 1.53 4.84 1.95 9.48

PAGE 130

122 medium primary forest above the small secondary. In addition, significant differences between WH values (pair-wise t-tests, p<0.05) were only found between the three highest ranking forests (LG-SC, MDSC, MD-PR) and the lowest ranking plots (SM-SC, LG-PR, SM-PR). If we scrutinize the data (see Table 3-1) we see that the results are consistent with the sensitivity of WH. Both plots have 9 species, 8 of which are shared between the two. However, for all but one species ( Oryzomvs trinitiatis ) the medium primary site supports much larger population sizes for the individual species. In Figure 3-5 we can observe that if species populations were allowed to increase at SM-SC (small secondary forest) its value would surpass that of MD-PR (medium primary) by virtue of its higher evenness between species' contribution to the whole community. It is also important to notice that the small secondary forest has three species represented by only one individual. The resulting discriminant function varied in its power of correctly discriminating primary and secondary forests trapping stations, from a low of 75 % (small primary) to a high of 98 % (small secondary, Table 3-3). Given that the initial probability of erring was 50 %, the forest which had the highest percentage of incorrectly classified stations still had the majority of them within the correct category. The small primary forest had the highest proportion of misclassif ied stations because of the large edge area relative to other forests. In summary, the discriminant function was

PAGE 131

123 SPECIES DIVERSITY RANK T 1 1 r SM-PR MD-PR LG-PR SM-SC MD-SC LG-SC FOREST PLOT SPECIES DIVERSITY ^^ SHANNON-WIENER H' EZl WH INDEX Figure 3-6. Comparison of ranks of diversity indices for both the Shannon-Wiener H' diversity index and its weighted version (WH). The y axis values represent ranks and not the actual values of the diversity indices.

PAGE 132

124 generally successful in discriminating trap stations across all forest sites, and its Wilks' lambda value was highly significant (p<0.0001). Table 3-4 summarizes the means of all the habitat variables in the different forest types and traplines. In general, primary forests tend to possess higher trees than do secondaries, with a high canopy and a more developed emergent connection. A well developed midstory, as measured by the connectedness at that stratum, is a characteristic of secondary forests. They also have a higher density of woody trees per unit area than primary forests because of the presence of a large number of trees with small basal areas. Although tree species richness tends to be higher in secondary forests, the average number of species per trapping station does not appear to differ between primary and secondary forests. Correlation analyses between community parameters and habitat variables, lumped together across all traplines, indicate that there were no significant correlations between any of the habitat variables and the Shannon-Wiener index, or between these variables and species diversity as measured by the WH index. The number of species (S) showed a negative association with volume at the herbaceous stratum DVH (r=0.17, p<0.05), and a positive one with volume of midstory vegetation DVM (r=0.17, p<0.05). These two habitat variables are probably not independent and may be measuring the same relationship. A negative association was found between herbaceous cover (HC) and both number of individuals (NI) and the weighted diversity index (WH) (respectively,

PAGE 133

125 Table 3-3. Results of the discriminant analysis separating trapping stations into primary and secondary forests (Wilk's lambda=0.514; p<0.0001). Only two classes were recognized (primary and secondary forest stations), with no discrimination for size. Therefore, initial probability of classification in either category was 0.5. Small Medium Large Primary Secondary Primary Secondary Primary Secondary % Correct 75.0 98.0 87.5 81.3 93.4 83.3

PAGE 134

126 Table 3-4. Summary statistics of habitat variables measured in the six different forest plots, A total description of the variables can be found in text. For all variables, the total number of observations is equal to the total number of trap stations in each site (N=48), except where noted. All measurements (except sliding scales) are in meters; SDI is in centimeters. VARIABLES FOREST PLOTS Fazenda Esmeralda Rio Fazenda Montes Claros Parque Estadual Casca Caratinga Florestal do Rio Doce Small (80 ha / 60 ha) Medium (860 ha) Large (35,980 ha) Primary Secondary Primary Secondary Primary Secondary (SM Mean -PR) (SM-SC) (MDPR) (MDSO SD (LG-PR) (LGMean SC) SD Mean SD Mean SD Mean Mean SD SD HT 5.6 1.8 4.9 2.0 7.4 2.1 6.70 1.6 7.2 2.2 6.5 1.6 OH 13.8 5.8 6.3 4.1 20.4 6.7 13.7 5.6 21.7 6.9 14.9 5.0 HH 2.0 0.8 2.2 0.9 1.4 0.8 2.41 0.9 1.1 0.8 1.0 0.6 CXC 0.8 1.0 0.1 0.4 0.9 0.8 0.44 0.7 1.4 0.9 0.7 0.8 CXM 1.6 1.0 0.6 0.8 1.5 0.7 1.42 1.0 2.2 0.8 2.3 0.8 CC 52.4 12.9 31.3 19.1 43.0 17.0 48.0 13.5 52.7 7.4 51.6 10.0 HC 51.8 12.0 48.5 12.9 38.3 15.0 51.9 11.2 56.9 7.1 40.3 17.3 DVC 1.2 0.9 0.3 0.5 0.9 0.8 0.79 0.9 1.5 0.9 1.2 0.8 DVM 1.7 0.9 0.9 1.0 1.9 0.8 1.71 0.9 2.2 0.7 2.2 0.8 DVH 2.1 0.9 2.2 0.8 1.5 0.9 2.06 0.9 1.9 0.7 1.4 1.0 DV 1.8 0.8 1.0 1.1 1.6 1.1 1.77 1.1 1.2 0.4 1.8 0.7 DE 0.4 0.9 0.0 0.3 0.4 0.6 0.08 0.4 2.0 0.9 0.2 0.5 FL 1.2 1.1 0.3 0.6 1.9 1.5 0.85 1.2 0.7 0.8 0.7 0.8 HUM 1.7 1.0 0.7 0.9 1.9 0.6 1.42 0.7 1.4 0.7 1.3 0.7 XSP 4.7 2.0 3.1 1.7 6.3 2.0 5.96 2.7 5.5 2.5 4.5 1.8 NSP 78 65 51 68 60 70 NTR 367 429 200 397 284 323 SDI 32.6* 37.1 19.3** 41.7 30.7 34.8 *N=47. **N= 44.

PAGE 135

127 r=0.57, p<0.007; r=0.52, p<0.01). The toal number of individuals was also positively associated with average height of trees (r=0.41, p<0. 05). weighted diversity index (WH) (respectively, r=0.57, p<0.007; r=0.52, p<0.01)The total number of individuals was also positively associated with average height of trees (r=0.41, p<0.05). The second type of analysis investigated the effect of the habitat variables on the variation of H, WH, S, and NI, but this time primary and secondary forests were considered as separate data sets. Figure 3-7 illustrates how the analysis of both primary and secondary forests together can be misleading: the correlation between the number of fallen logs and H' is non-significant when all sites are taken together. However, among secondary forest traplines, species diversity is significantly and positively associated with fallen logs. In contrast, the species diversity of primary forest traplines is not significantly correlated with fallen logs. Therefore, the importance of specific habitat variables varies between primary and secondary forests. In Figure 3-8, the plot of WH against mean number of plant species (XSP) shows that even though both primary and secondary forests' WH indices tend to increase with the mean number of plant species present in the habitat, the regression for primary forests is not significant.

PAGE 136

128 Shannori-Wiener index -11 1,5SECOMDARY FORESTS H' = 1.4 + 0.38 FL(p<0,04) 1 ^^.^--^^^^ ^^^-^-"^ -1+ 4; -^ 0.54,_--^' + i f'^ + 1 1 1 1 1.3 1.4 1.5 1.6 1.7 1.8 Number of Fallen Logs 1.9 SECONDARY FORESTS Figure 3-7. Plot of H' against FL for secondary forests, with each individual data point representing averages of a forest trapline. The regression equation is also shown.

PAGE 137

Figure 3-8. Plots of WH against XSP (mean number of tree species), with individual data points representing forest traplines. Figure 3. 8. a shows the relationship for secondary forests, and Figure 3.8.b. for primary forests only.

PAGE 138

Weighted Shannon-Wiener Index (WH) 130 5421 SECONDARY FORESTS WH = 0.10 0.61 XSP(p<0.01) 1 r 1.3 1. a. 2,3 2.8 3.3 3.3 4,3 4,8 5,3 ,5,. Mean Number of Tree Species ^ SECONDARY FORESTS 6,3 6, Weighted Shannon-Wiener Index (WH) • PRIMARY FORESTS (N.S.) 3^^^^---^ 2^^^.^.^--^'^''^ • 1 1 n1 1 1 1 4 b. 4,5 5 5.5 6 6.5 7 Mean Number of Tree Species ^PRIMARY FORESTS 7.5

PAGE 139

Figure 3-9. Plots of WH against DVC (canopy volume), with each individual data point representing a forest trapline. Figure 3. 9. a. shows the regression for secondary forests, and 3.9.b. for primary forests. Note the opposite trends in WH between primary and secondary forests.

PAGE 140

132 Weighted Shannon-Wiener Index (WH) 3SECONDARY FORESTS WH = 1.61 + 1.64 DVC (p<0.008) a. 0.2 0.4 0.6 0.8 1 Canopy Volume (DVC) — SECONDARY FORESTS 1.4 1.6 Weighted Shannon-Wiener index (WH) 1 PRIMARY FORESTS WH = 4.712.23 DVC (p<0.C3) 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1,5 Canopy Volume (DVC) D. ^—PRIMARY FORESTS 1.6 1.8

PAGE 141

133 The plot of WH against canopy volume (DVC) yields regressions with slopes of different signs for primary and secondary sites (Figure 3-9). The differential response of diversity of primary and secondary forests was not found, however, for all habitat variables. The weighted diversity index (WH) for both primary and secondary forests tends to decrease with the increase in herbaceous volume (DVH), both yielding significant regressions (Figure 3-10). Species diversity, as measured by the Shannon-Wiener index H', appears to increase in secondary forests with higher trees and canopies (HT, OH), with well developed connections in midand upper-stories (CXM, CXC). Diversity is also higher in well shaded areas (CC, DVC, DVM), and in trap stations close to fallen logs (FL). H' was also found to decrease with the amount herbaceous cover present (HH, DVH). The same general trends were obtained for species diversity as measured by WH, but which also demonstrated to increase in areas with high vine density (DV). Diversity of primary forests, on the other hand, appears insensitive to most habitat variables measured, only a few of which were found to be correlated (Table 3-5). The above relationship also held the same trend for the component variables of the diversity indices, i.e., number of species and number of individuals (Table 3-5). Species richness (S) was associated with all variables that H' was for secondary forests.

PAGE 142

Figure 3-10. Plots of WH against DVH (herbaceous volume), with regressions for secondary (Figure 3. 10. a) and primary forests (Figure 3.10.b) shown separately.

PAGE 143

135 Weighted Shannon-Wiener Index (WH) 3.532.5 21.5SECONDARY FORESTS WH = 5.20 1,27 DVH (p<0.05) — \ 1 1 [ 1 1 1 1 — 1.2 1.4 1,6 1.8 2 2.2 2.4 2.6 Herbaceous Volume (DVH) ^ SECONDARY FORESTS a. 2.3 Weighted Shannon-Wiener index (WH) b. PRIMARY FORESTS WH = 5.00 1.63 DVH (p<0.C5) ^2 1.4 1.6 1.8 2 2.2 Herbaceous Volume (DVH) -^PRIMARY FORESTS 2,4

PAGE 144

136 Table 3-5. Summary of correlation analyses, showing variables with significant correlation coefficients with Shannon-Wiener (H), Weighted Shannon-Wiener (WH) indices, Total number of individuals in community (NI), and Number of species (S) (*=p<0.05; **=p<0.01; ***=p<0.001). Secondary Primary H WH H WH VARIABLE r p r p r p r p HT 0.72 0.85 ** OH 0.89 ** 0.92 ** HH -0.76 cxc 0.90 ** 0.69 CXM 0.95 *** 0.88 ** cc 0.93 ** 0.95 *** -0.68 HC -0.83 ** DVC 0.88 ** 0.73 -0.74 DVM 0.88 ** 0.88 ** DVH -0.82 ** -0.69 DV 0.78 ** FL 0.68 0.78 NSP 0.79 ** HUM -0.62

PAGE 145

137 Table 3-5. continued Secondary P rimary S NI S NI VARIABLE r p r P r P r P HT 0.70 0.86 ** OH 0.84 ** 0.96 *** HH -0.80 ** CXC 0.90 ** 0.79 ** CXM 0.93 ** 0.87 ** CC 0.85 ** 0.95 *** -0.68 HC -0.84 ** DVC 0.83 ** 0.81 ** -0.70 DVM 0.88 ** 0.84 ** DVH -0.81 ** -0.68 -0.66 FL 0.75 0.80 ** NSP 0.74

PAGE 146

138 with exception of number of logs (FL). Total number of individuals (NI) of secondaries was associated with all habitat variables that WH was plus herbaceous volume (DVH). In contrast, the number of species (S) in primary forests was not associated with any of the habitat variables measured. Number of small mammal individuals of primary communities (NI) was significantly associated with all variables that WH was plus number of fallen logs (FL). We investigated the possible association between area size and the community parameters number of species (S), number of individuals (NI) and species diversity, as measured by H' and WH. There were three main category sizes, each represented by two forests, one primary and another secondary: small (60 and 80 ha.), medium (860 ha.) and large (35,983 ha.). In all cases loge(area size) provided a better fit. There is no significant association between area size and any of the community variables if primary and secondary forests are lumped together. However, area size yielded significant correlations with H', WH, S and NI when secondary forests were examined separately (respectively, r=0.92, p<0.004; r=0.84, p<0.003; r=0.92, p<0.004; r=0.84, p<0.003). On the other hand, no significant trends were observed for primary forests alone. Figure 3-11 illustrates the relationship found for the separate analysis between primary and secondary forests.

PAGE 147

139 Weighted Shannon-Wiener Index (WH) 3SECONDARY FORESTS WH = 0.64 + 0.32 log AREA (p<0.004) 3i + + PRIMARY FORESTS (N.S.) 1 \ 1 4 6 8 Log Area of Forest Plot 10 12 -+PRIMARY FORESTS SECONDARY FORESTS Figure 3-11. Plots of WH and log of Area Size for secondary and primary forests. Note that the regression is significant for secondary forests, while the one for primary forests is not. WH of primary forests do tend to increase from the small to the medium size category, but no change can be noticed between medium and large size forest trapl ines.

PAGE 148

140 Primary forest species diversity (as measured by WH) does tend to increase with area from the small size to the medium size category (Table 3-2), but species diversity of large primary traplines dropped to levels comparable with the ones obtained for the smaller forests plots. Based on all of the above results, we then eliminated those independent habitat variables that were either intercorrelated or that on their own did not explain any significant amount of variation in the mammal species diversity variables. In addition, those habitat variables that were significantly associated with area size were also dropped. It was therefore possible to construct multivariate models that included both the effect of area size and habitat characteristics. The best models for H' at secondary forests were obtained by using number of fallen logs (FL) and area size, which was able to explain 92 % of the variation in species diversity (Table 3-6). For WH, the mean number of plant species in trapping station (XSP) and area size together explained 97 % of the variation. Area size alone explained 84 % of the variation in species richness, even though connectedness at the midstory (CXM), a variable correlated with size, accounted for the approximately the same amount of variation in the number of species (86 %). No multivariate model for the total number of individuals (NI) generated with the above method was able to obtain a significantly better fit than overstory height (for OH alone: r2=.92, p<0.0001; for XSP and area: r2=.91, p<0.0003). Therefore, for both

PAGE 149

141 Table 3-6. Multiple regression analyses of Shannon-Wiener (H) and Weighted Shannon-Wiener (WH) indices, and its components, against habitat variables. S=number of species of small mammals; NI=total number of individuals of small mammals trapped at each transect line. PREDICTOR VARIABLES IN MODEL FL A FL, A XSP A XP, A Secondary Forests r2 P< Dependent 0.39 0.84 0.92 Variable=H 0.04 0.0004 0.0002 Dependent Variable=WH 0.57 0.71 0.97 0.01 0.003 0.0001 Dependent Variable=S No Habitat predictor variables recognized A 0.84 0.0004 XSP A XSP, A Dependent Variable=NI 0.49 0.72 0.91 0.02 0.003 0.0003 Primary Forests r2 P< Dependent Variable=H No Habitat predictor variables recognized A n.s.

PAGE 150

Table 3-6 continued. 142 PREDICTOR VARIABLES IN MODEL Secondary Forests P< DVH, DVC A Dependent Var1able=WH 0.65 0.01 n.s. No Habitat predictor variables recognized A Dependent Var1ab1e=S n.s. Dependent Var1able=NI FL, HC A 0.82 0.008 n.s.

PAGE 151

143 Table 3-7. Multiple correlation coefficients for habitat and area variables in regression models of species diversity for secondary forests. The % column refers to percentage of variance attributable to area size, and is equal to the ratio (Area r^-Habitat r2)/0verall r^ DEPENDENT Secondary Forests VARIABLE Overall r^ Habitat r^ Area r^ Area-Habitat % H 0.92 0.08 WH 0.97 0.26 0.53 0.45 0.40 0.14 48.9 14.4

PAGE 152

144 species richness and number of individuals, the analysis did not proceed further. On the other hand, the multivariate regression analysis for species diversity of secondary forests can be further separated into its components, as shown in Table 3-7. Shannon-Wiener species diversity is better explained by both the habitat variable used in the model and area size. However, area size appears to contribute to approximately 48.9 % of the variation in H' in the complete model, or a little more than does the habitat structure variable. In the WH model, on the other hand, area appears to contribute less than the habitat variable (XSP) in the variation of the dependent variable (14.4 %). For primary forests area size did not function as a predictor in any of the models (Table 3-7). No multivariate regression model that included area size was better able to explain variation in H', WH and S than either of the previous habitat structure variables that were significantly correlated with species diversity. There was, however, a model for NI that included more than one non-correlated habitat structure variable that provided a better fit to the data. This included number of fallen logs (FL) and herbaceous cover (HC), the former with a positive parameter and the later with a negative one.

PAGE 153

145 Discussion This study indicated that marsupials predominate both in terms of number of individuals and in biomass in the small mammal forest communities of the Brazilian Atlantic forest region. Rodent species richness was comparable to that of marsupials in the larger plots, while some rodent species were quite common in the medium sized plots. Previous studies in neotropical forests have yielded contrasting results: Laemmert et al (1946) in the northern part of the Atlantic forest, and Emmons (1984) and Malcolm (1987) in the Amazon region, have found rodents to outnumber marsupials in species diversity and abundance. Stallings (in prep.), using the same trapline methods of this study in the Rio Doce State Park, found that marsupials outnumbered rodents. No definite explanation is yet possible for these different results, although marsupial overabundance may be related to the present-day high frequency of occurrence of second growth habitat in the Atlantic forest region. Second growth has been shown to be more productive in terms to the food resources used by marsupials (Charles-Dominique et al., 1981). This study has shown that primary forests tend to have lower small mammal species diversity. Taking H' as a measure of species diversity, no primary forest plot, regardless of size, was more diverse than any of the secondary sites. The lowest species richness and diversity was recorded for a site which, in spite of its large size,

PAGE 154

146 was mostly primary in nature. The results also indicated that higher species richness and diversity are found in secondary forest that are of large size. I should stress that these are relatively older-growth secondary forests, averaging about 20 years old. Charles-Dominique et al (1981) found that the marsupial part of a community in French Guyana is much more diverse and abundant in secondary growth. Isabirye-Basuta and Kasenene (1987) also found rodent species diversity and abundances to be higher in selectively felled forests, when compared with relatively undisturbed primary vegetation. Recent studies in younger forests (less than 5-8 years) indicate that species richness and diversity also decrease in very early growth forests (Fonseca et al., 1987). These forests are characterized by even-aged tree stands, which tend to be structurally homogeneous. These results are in accord with other studies of diversity and of plant community succession. They all have indicated that cUmax communities are found to be less diverse than some younger stage, provided the nature of the disturbance is not of extreme frequency or intensity (Connell, 1978; Souza, 1984; Brokaw, 1985; Glitzenstein et al., 1986). As Horn (1974; page 30) has stated, Only one pattern of diversity in succession should be nearly universal. In a succession that is not subject to extensive and chronic disturbance, the diversity of the climax must be lower than that of some preceding stage.

PAGE 155

147 The habitat variables that were associated with significant amounts of variation in the small mammal community variables indicate that for both primary and secondary forests species richness and diversity will be higher in habitats with a well developed midstory. This is reflected by those variables related to midstory volume and vine density. Species richness and also the total number of individuals in the community are negatively associated with presence of well developed herbaceous vegetation. Moreover, population sizes are also larger in forests with higher trees (but excluding emergents). While well developed herbaceous vegetation is usually associated with second growth, higher trees are not. However, on a station to station basis, average height and volume of herbaceous vegetation of secondary forests are not necessarily higher than in primary ones (see Table 3-4). Even though herbaceous cover in very young forests can be attributable to lack of heavy tree cover, a well developed midstory, characteristic of many older-growth secondary forests, is much more efficient in buffering light than canopies dominated by a few emergents, a characteristic of Atlantic region climax primary forests. These observations thus lend further support to the contention that small mammal communities will be richer in midto older stages of secondary growth than either very young or very old forest communities. The importance of mid-growth as a determinant of small mammal diversity is also evident if we examine primary forest regression models separately. While H' and number of species in primary forests

PAGE 156

148 do not relate strongly to most any habitat variables measured, the exceptions are notwithstanding meaningful. The habitat variables included in the regression model for primary forests indicate that diversity and number of individuals increase with the reduction in volume of vegetation at the canopy strata and increase with the amount of fallen logs. Therefore, communities of secondary forests tend to be more diverse than climax forests, while traplines within primary forests crossing gaps or areas with some form of disturbance also yield higher species diversity. Diversity and richness among secondary forests are usually associated with vegetation possessing several characteristics, such as well developed midstories, taller trees, and as importantly, the number of fallen logs. Number of individuals in the primary communities was also positively correlated with this last variable. Fallen logs potentially constitute an important resource for terrestrial species. Malcolm (pers. comm. 1987) found Proechimys in central Amazonia highly associated with fallen logs, in which they look for refuge and use extensively for locomotion (see also Chapter 4). Moreover, Miles et al. (1981) also found Proechimys using fallen logs as substrate for locomotion and taking refuge in hollow logs. The behavioral response upon release of most of the small mammal species in this study indicated that gaps under fallen trunks are a preferred location for an escaping individual. But since knowledge of the natural history of these species is still superficial, the exact role of fallen logs

PAGE 157

149 remains an open question. Nonetheless, it is felt that the presence of logs on forest gaps may be crucial for the maintenance of Proechimys Several studies have shown that species compositions and relative abundances of small mammal communities can be partially explained by habitat structure and heterogeneity (Rosenzweig and Winakur, 1969; M'Closkey, 1976; Dueser and Shugart, 1978; August, 1983, 1984; Lacher et al., in press; Lacher and Alho, in press). Bird community studies have also revealed these correlations (Johnson, 1975; James and Wamer, 1982; Lynch and Whigham, 1984). Other studies have focussed on area size as a determinant of mammalian insular species richness (Dueser and Brown, 1980; Lomolino, 1984; Malcolm, 1987). A problem for both approaches is that several habitat variables covary with area size, and it is often difficult to separate the "pure" or dynamic effects of area sensu MacArthur and Wilson (1967) from the habitat structure question. Zimmerman and Bierregaard (1986) using tree frog data from Amazon forest fragments. Weaver and Kellman (1981) with birds, and Westman (1983) plant associations, all have empirically warned against the unqualified use of island biogeographical derivations. Some studies have been successful in separating, or at least controlling for area size and habitat structure (Buckley, 1982; 1985; Boecklen, 1986). In his mangrove island experiments with arthropod communities, Simberloff (1976) was able to identify variables related to habitat structural diversity that were able to explain some of the variation in the number of species, and that were independent of island size.

PAGE 158

150 In this study I separated the effects on small mammal diversity of a few habitat variables from those of forest fragment size. Some of the best models for species diversity of secondary forests included both the effect of "area size" alone and one or two habitat variables non-correlated with size. The effects of area and habitat for both species diversity indices were identified and their relative importance for the determination of species diversity established. Area was responsible for a little over half of the variation in H' in secondary forests, with habitat being responsible for the remaining. There are several possible ways in which small mammal diversity would be increased in secondary forests. Unfortunately, a large number of habitat variables were not used in regression models due to colinearity problems. Thus, the interpretation of the underlying mechanisms by which the specific habitat variables used in regression models affect variation in small mammal community patterns was not totally clear. As it has been demonstrated in another circumstance (Weaver and Kellman, 1981), differential rates of extinction can sometimes be directly linked to final stages of succession, more so than area size, with pioneer species being unable to persist in mature forests. Forests that provide niche space in their intermediate stages of succession for both pioneer and secondary invading species are much better able to maintain a more diverse fauna. Mid-stage second growth communities would be then characterized by species with higher evenness within a particular guild than structurally simple mature forest

PAGE 159

151 communities dominated by a few species. Furthermore, it is suspected that very few small mammal species of the Brazilian Atlantic region, and perhaps elsewhere in the tropics, depend on climax forests. Corroborating this hypothesis, with the exception of one species represented by one individual, all species of the primary communities examined here were also recorded elsewhere in secondaries. Moreover, seven species could only be found in secondary settings, a further indication that mid-level succession stages are better able to maintain a larger diversity at the local level. Isabirye-Basuta and Kasenene (1987) determined that all but one species present in mature African forests were also represented in disturbed forests. Similarly, no vertebrate species of the Western Cascades of the United States has been shown to dwell exclusively on mature forests (Harris, 1984), although some species, such as the spotted owl and Arborimus may be more dependent on old forest stands (J. Eisenberg, pers. comm. 1988). An alternative explanation to the results obtained here would be that the arboreal trapping methods were not able to adequately represent a possible unique primary forest fauna that could inhabit higher vegetation strata. This does not seem to be the case. Stal lings (in prep.) has trapped extensively in higher strata of two forests of the Rio Doce Park. His sample did not yield any additional species than the ones reported for this study, even though some of the more arboreal marsupials ( Marmosa cinerea and Caluromys philander ) were overly represented in relation to ground trapping. An exception may be

PAGE 160

152 the arboreal spiny rat Echvmys sp. which according to Miles et al. (1981), is never observed on the ground. However, the species was only represented in this study by one individual in one forest plot, and therefore did not have a major effect in altering the results of this analysis. Species diversity of primary forests, on the other hand, does not seem to be affected by area size. Why is it that species diversity of secondary forests responds to area size and primary forests show no response? Vertical packing of species has been originally described for avian communities (MacArthur, 1957). Since 11 out of the 19 small mammal species demonstrate variable degrees of climbing ability (see Chapter 2), the addition of a well developed 3-dimensional environment can similarly potentially provide additional and broader niche space. A growth in species richness and diversity of secondary forests with a corresponding increase in area size might conceivably be achieved by several means, described as follows: 1. The growth in the individual sizes of species populations, possible because of the increase in the width of the individual niches of 3-dimens1onal environments. 2. The increase in alpha-diversity, represented by midstory structural components such as vines and canopies of mid-story second-growth trees. 3. The increase in beta-diversity. The first factor certainly lowers the probabilities of local

PAGE 161

153 extinctions from particular communities. The second factor increases habitat heterogeneity both in relation to primary forest structure and also by, as gaps other openings tend to differ from one another (Souza, 1984), the increase in the number of distinct habitats that become available. This line of reasoning links both the habitat structure and diversity theory of Williams (1964) and the MacArthur & Wilson extinction-colonization dynamic model, without the need to invalidate either one of them. Species diversity of primary forests communities, on the other hand, due to their relatively larger homogeneity per unit area and the absence of developed midstory, is less capable of locally responding to changes in area size. Probably very large changes in area size (or sampling effort) would be necessary to notice these influences. Although productivity of small mammal resources was not measured, it is felt that it can also exert a certain degree of influence on community parameters. Second growth vegetation can possibly be more productive for marsupials (see Charles-Dominique, 1983), as well as for rodents. If this is the case, one might expect that the consequent response to area size would be much more noticeable in secondary than in primary forests. Some primary productivity measures have been found to increase species diversity in plant communities (Horn, 1974). Recently, Huntly and Inouye (1987) suggested that, for a relatively simple small mammal community of central Minnesota, measures of primary productivity, such as nitrogen content, may play a role in increasing

PAGE 162

154 species diversity and abundances. Similarly, Charles-Dominique et al. (1981) have implicated the low diversity and biomasses of didelphid marsupials at French Guyanan primary forests as the result of lower productivity, relative to secondary vegetation. Apparently, the frugivore portion of the marsupial diet is overly represented by items characteristic of pioneer plant species. Conservation and Management Implications The traditional view of tropical forest animal and plant communities is that of the inexorable march towards a stable climax characterized by the highest possible assemblage of tightly packed species. The notion that disturbance regimes that are higher in magnitude than presumed "historical" ones always tend towards a reduction in diversity (Denslow, 1985), especially in the tropics, is still quite common in the literature. Even though it is not wise to extrapolate the results of this study, which focused on only one guild, to other mammals of the tropical region, it is possible to say that small mammal species richness and diversity of the forests examined here are at their highest in mid-growth secondary forests. The role of edge effects on temperate communities (Weaver and Kellman, 1981) and the suggestion that some level of disturbance is necessary for maintaining high species diversity, even in tropical communities (Connell, 1978; Karr,

PAGE 163

155 1982), are now receiving increasing support from field data. However, Lovejoy et al. (1986) have found that the bird fauna of the edges of small isolated reserved in the Amazon basin tend towards a decrease in species diversity and relative abundances, while Malcolm (1987) did not find any conclusive evidence for the small mammal communities of these same reserves. Furthermore, some primate species may not survive if the mode of habitat alteration is such as to drastically reduce important food resources (Johns and Skorupa, 1987), although moderately disturbed forests are nonetheless capable of maintaining most rain forest primates. These contrasting results should make us careful about generalizations, specially between different guilds and taxa. The findings of this study are also pertinent to the decade long SLOSS (Single Large or Several Small Reserves) debate on reserve planning (Diamond, 1975; Simberloff and Abele, 1976; Simberloff, 1982; Quinn and Hastings, 1987). I have determined that all four forest patches of small and medium size (60, 80, and 860 ha.) together yielded 11 species, while the single large forest of secondary nature alone supported 14 small mammal species (Table 3-1). Therefore, the evidence with small mammals supports the contention that larger patches do support higher species diversity per unit area than a collection of smaller reserves. I should stress that the argument for multiple reserves is usually conducted with the idea that the added area of these reserves is similar in size to that of the large unit being compared, which is clearly not the case here. However, since all

PAGE 164

156 reserves were sampled with the same trapping effort and over sampling areas of the same size (6 ha.), the comparison is still meaningful. We have to consider that smaller reserves together turned out 11 species using four sets of traplines (24 ha.), in contrast with 14 species in a single set (6 ha.) in the large reserve. Nevertheless, it is necessary to realize that the habitat structure of a reserve is also crucial, since the large primary forest of this study proved one of the most species poor communities studied. Thus, even though forests of larger sizes usually support more resilient communities, the pure application of island biogeography to conservation and research design in the tropics is not advisable. Habitat evaluation and studies of the life histories of individual species are of utmost importance if we are to establish sound planning methods.

PAGE 165

CHAPTER 4 THE RELATIVE ROLE OF HABITAT SELECTION, COMPETITION AND PREDATOR PRESSURE ON THE STRUCTURE OF TROPICAL FOREST SMALL MAMMAL COMMUNITIES Introduction Some theoretical ecologists have explained that the high species diversity of tropical forest communities is related to reduced level of interspecific competition brought about, among other things, by habitat or spatial segregation (MacArthur, 1972; Pianka, 1983). As tropical forests are more spatially heterogeneous and display a higher level of stability, communities would consist mostly of specialist, K-selected species, displaying very limited niche overlap (for a review, see Bourliere, 1983). Rosenzweig and Winakur (1969), in a widely quoted study, have implicated the reduction in habitat overlap in desert rodents as means of promoting coexistence in competitive environments. Other small mammal studies have arrived at similar conclusions in several non-arid habitats (M'Closkey, 1976; Dueser and Shugart, 1978; Price, 1978; Dueser and Hallett, 1980; Hallett, 1982; Hallett et al., 1983; Lacher et al., in press). However, none were conducted in a tropical forest setting. Studies of tropical small mammal communities have yielded mixed results. Malcolm (1987) compared small mammal community structure of 157

PAGE 166

158 different sites in neotropical forests of central Amazonia but did not recognize habitat correlates of differential species abundances. On the other hand, differential use of habitat by tropical didelphid marsupials is apparently implicated in allowing sympatric species to coexist (O'Connell ,1979; Charles-Dominique, 1983). August (1983; 1984) has also shown that, for the relatively poor small mammal tropical community of the Venezuelan llanos, certain habitat variables were correlated with the relative abundance of some small mammal species. Sympatric species found in a gallery forest of the Brazilian cerrado have also shown differences in habitat preference (Nitikman and Mares, 1987), interpreted as mechanisms for minimizing competition. In the previous chapter, I indicated that small mammal communities of mid-stage secondary forests of the Brazilian atlantic coastal region tend to display higher species richness and diversity than climax forests. Secondary forests, because of their structure, may allow a higher degree of habitat separation to occur and thus increase local diversity. Species richness and diversity in Atlantic forest small mammal communities have also been observed to be positively correlated with area size (Chapter 3). Given that species richness and diversity increase in secondary forests and also in forests of increasing size, the present study will focus on the following questions: 1. As diversity is both defined by the total number of species and their relative contributions to the community (Hair, 1982), can the

PAGE 167

159 variation in the composition and the relative abundances of small mammal species between different forest plots be explained by habitat requirement distinctions? 2. If tropical forest small mammal species select microhabitat from available habitat, how segregated are their microhabitats? Habitat selectivity and partitioning would suggest that competition may play a role in the structure of these small mammal tropical forest communities. 3. Can patterns of species interactions further suggest that competition constitutes a significant variable affecting small mammal community characteristics? 4. Finally, we investigate the possible mechanisms involved in allowing larger forests of similar structure to display higher species diversity per unit area. The explanation for this trend may relate less to the classical island biogeographical theory (MacArthur and Wilson, 1967) and instead be more closely linked to the higher predator pressure occurring in larger forests. For both plants and animals, selective predation of more abundant species has been suggested as a possible mechanism of promoting community diversity at a local level (Paine, 1966; Harper, 1969; MacArthur, 1972; Connell, 1975). The hypothesis was originally put forward by Paine (1966, page 65): "Local species diversity is directly related with the efficiency with which predators prevent the monopolization of the major environmental requisites by one species."

PAGE 168

160 This hypothesis, however, has never been tested in mammalian communities, even though there have been suggestions that predation may be responsible for the extinction of species within some avian guilds whne area size restriction exists (Karr, 1982; Karr and Freemark, 1985; Loiselle and Hoppes, 1985). Very few studies have been conducted on the ecology of eastern Brazilian mammals. The most complete work on small mammal communities of the Atlantic forest is now 40 years old (Davis, 1946; 1947), while the majority of studies were confined to very general natural history or taxonomic notes on individual species (e.g., Laemmert, 1946; Moojen, 1952; Avila-Pires and Gouvea, 1977). These studies have all indicated that communities of small mammals in the Atlantic forest are unusually rich in number of species, and the level of endemism is quite high. A 6 ha. sample in a Rio Doce Park secondary forest yielded as much as 14 small mammal species (see Chapter 3). This richness certainly propitiates comparisons of different small mammal communities in that region. Small mammals are defined here as marsupial and rodent species with average adult body size of less than 1,000 grams. Rodents and marsupials are assumed to represent an identifiable guild within the broader mammalian community. They are all relatively short-lived, are mostly nocturnal, have wide diet overlaps and are not confined to the canopy.

PAGE 169

161 Materials and Methods Study Sites The eastern region of the South American continent encomoasses the main vegetation formation known as the Brazilian Atlantic forest. Within the South American continent, the area is recognized as having zoogeographic identity (Mullen, 1973). In its primeval state the vegetation of this coastal mountainous region was mostly composed of tall non-deciduous trees characteristic of tropical evergreen forests (Hueck, 1972). On the western slopes of the coastal region, where this study was conducted, forests may possess a variable number of deciduous trees which lose their leaves during the approximately five months of the dry season. There are no major differences in the mammalian fauna of eastern and western slopes of the Atlantic forest and during the rainy season the forests of either side are physiognomical ly very similar. The study was conducted at three main sites, each represented by two forest plots, one primary and the other secondary, all located within the Atlantic forest region of the state of Minas Gerais, Brazil. All sites are within a maximum distance of 300 kilometers and represent the same vegetation formation of that biogeographical region. The first site is a private farm, Fazenda Esmeralda, bounded by the Rio Doce river in the county of Rio Casca. The farm is 4,800 ha. in size, but not much of it remains under forest cover. Deforestation started

PAGE 170

162 in 1964 and was mostly completed by the early 1970s. One approximately 80 ha. patch (known as "Lagoa Fria"), which was never extensively exploited and that was in its latest succession stage, was selected as the primary forest site of the small size category. It is designated in this study as SM-PR. A 15-20 year old 60 ha. patch was selected as the secondary forest representative. It will be hereon known as SM-SC Fazenda Montes Claros, located in the counties of Ipanema and Caratinga, represents the medium size forest category. This farm is approximately 1,200 ha. in size, 860 ha. of which remain under forest. The terrain is mountainous, varying between 320 and 580 meters above sea level. There are two main sections with old primary forests, one of which, locally known as "Matao," was selected ( MD-PR ). Another location which was under coffee and pasture a few decades ago was selected as the secondary site of the middle size category, hereon known as MD-SC As both forests have common boundaries, have mostly the same small mammal fauna and can potentially exchange species and individuals, all samples are considered as coming from a forest patch of similar size, or 860 ha. The Rio Doce State Park, with its 35,973 ha., constituted the large forest sample or the control area. Due to the presence of extensive populations of top predators, such as jaguars, pumas and other carnivores, I assumed that the park represents a true sample of the original faunal assemblage of the region. The park belongs to the state of Minas Gerais Forest Institute (lEF). Except for its extensive

PAGE 171

163 network of bog lakes, the park is almost completely covered with forest, but due to frequent fires during the 1960s, perhaps only about less than one-third is primary. Nonetheless, the largest tract of primary forest left in the state is located within its boundaries. One section of primary forest, known locally as "Campolina," was selected and will be referred to as LG-PR An area that burned extensively in 1967, LG-SC was chosen to represent the secondary forest of large size. For the same reasons as in Fazenda Montes Glares, the trapping samples from each of the forest types were assumed to come from a 35,973 ha. forest. A typical primary forest in this study had an average height of approximately 20 meters, with relatively well developed canopies. Secondary forests ranged from an average of 7 to 14 meters in height. The understory may be quite dense, and tangled vines at ground and mid-levels are a common occurrence. Trapping A complete description of trapping methods can be found in Chapter 3. A summary is also provided here. All forest sites had 48 trap stations disposed into 3 parallel transects of 16 stations each, with a total of 112 permanent based traps. Lines were 300 meters long separated by 100 meters between lines. Each trapline is assumed to represent an independent sample and therefore each forest has three

PAGE 172

164 replicates of itself. Stations were separated by 20 meters within each line. Each station had at least one squirrel-size Tomahawk live trap, with alternation, at every other station, of an additional trap wired to a tree, branch, or vine between 1 and 4 meters high. In addition to these traps, every other station possessed a mouse-sized collapsible Sherman trap, with alternation of ground and tree traps. Moreover, the two outermost transect lines had, at every other station, a large 80 X 30 X 30 centimeter wire home-crafted live trap. Therefore, each outer line possessed 16 ground Tomahawk traps, 8 tree-bound Tomahawk traps, 4 ground and 4 tree-bound Sherman traps and 8 ground large wire traps. The total for each outer line was 40 traps. The midline did not have large traps, but a total of 24 Tomahawk traps and 8 Sherman traps. Trapping took place between June of 1985 and October of 1986, i.e., for 17 months of continuous sampling. Each forest plot was trapped for five consecutive nights a month. During the course of the study each forest site accumulated a total of 9,520 trap nights or a total of 57,120 trap nights for all the six forest plots together. Fresh pineapples, oatmeal and a cotton ball soaked with a commercial codfish oil solution were used as baits. Traps were checked every morning for captures and for adequacy of bait, which was replaced as needed.

PAGE 173

165 Habitat Variables A detailed description of the methods used in collecting the habitat variables is presented in Chapter 3. At each trapping station, the following variables were recorded: 1. Percent canopy cover (CC) above trap station. 2. Percent herbaceous cover (HC) around center of trap station. 3. Interconnectedness of canopies (CXC), represented by branch to branch contact or canopy to canopy link through vines. 4. Interconnectedness of midstory (CXM). 5. Volume at canopy height (DVC), representing sparse vegetation, 3 dense vegetation. 6. Volume at midstory height (DVM). 7. Volume of herbaceous vegetation (DVH). 8. Vine density (DV). 9. Epiphyte density (DE). 10. Fallen logs within or in close vicinity of trap station (FL). 12. Litter volume at soil surface (HUM). In addition to these, other habitat variables were quantitatively measured: 13. Overstory height (OH), to the nearest one-tenth of meter. 14. Herbaceous height (HH), determined as the most representative height category at a trap station, also to the nearest one-tenth of meter.

PAGE 174

166 15. DBH of all woody plants with DBH greater than 3.2 cm to the nearest centimeter. The composite variable SDIAM is represented by the sum of all DBHs at a trap station. 16. HT, or the average height of all trees at trap station with DBH greater than 3.2 cm to the nearest meter. 17. NTR, or number of trees at station with DBH greater than 3.2 cm. 18. NSP, the number of woody plant species at trap station. 19. BP, or presence (1) or absence (0) of palms. 20. TB, or presence (1) or absence (0) of bamboo. Statistical Methods and Data Analysis Habitat selection Habitat selectivity is frequently measured using either multivariate descriptions of a species' habitat (Boecklen, 1986; for an analogous discussion using systematic data, see also Willig et al., 1986) or emphasizing univariate correlates which sometimes may serve as predictors of species occurrences and abundances (Lynch and Whigham, 1984). Some recent studies have conducted both analyses (August, 1983; Blake and Karr, 1987), an approach adopted in the present paper. Both univariate (ANOVA; SAS, procedure GLM) and multivariate (Discriminant Analysis; SAS, procedures GLM, subroutine MANOVA, and procedure

PAGE 175

167 DISCRIM) methods were used in assessing habitat selectivity of a particular species. Only species with 20 or more captures were used. The univariate analysis of variance was used to test the null hypothesis that the mean value for each habitat measure at trap stations where a species was caught did not significantly differ from the mean of that variable obtained over all trapping stations. Thus, I considered that the overall means of each habitat variable represented available habitat. Discriminant analysis was used to test the null hypothesis that habitat profile for each species did not differ from profile of available habitat. This type of analysis is equivalent to MANOVA (August, 1983). The probability associated with Wilks' lambda was used to assess differences between habitat and species profiles. Available habitat is represented by means of all trapping stations of all forest plots where species was recorded. The discriminant functions generated with the above methods were also used to classify trapping stations in forests plots where species were and were not trapped. The objective here was to assess the possible absence of a species from a particular site as a result of habitat requirements. In addition to Wilks' lambda values, the ability of the discriminant functions in correctly classifying trapping stations provided a further indication of the accuracy of the analysis (Kachigan, 1982). ANOVA assumes homogeneity of variances (Sokal and Rohlf, 1981). To examine this assumption I applied Hartley's Fmax-test. Most variables proved to be heteroscedastic. Therefore, I transformed all

PAGE 176

168 habitat variables to logarithms, which stabilized the variances of nearly all variables used. For the multivariate models, both variance and covariance matrices are required to be homogeneous (Kachigan, 1982). The violation of these assumptions is frequent with ecological data, but they do not seem to affect the reliability of the tests to a great extent (August, 1983), especially when compared to non-parametric alternatives (Dueser and Shugart, 1978). All statistical analyses were conducted using the PC-SAS (SAS Institute, Gary, NO software package on MS-DOS (Microsoft Corporation) based machines. Species interactions In order to detect possible competitive interactions I used univariate analysis of variance (ANOVA). I tested the null hypothesis that the mean trapping success of a species did not differ between trapping stations where it only was captured and those at which a second species was also recorded. Species with less than 20 captures were not used in this analysis. In addition, Lloyd's (1967) mean crowding index was calculated for each species under analysis. The index reflects the mean number, per individual of a species, of individuals of a second species in the same station where it was recorded (Lloyd, 1967; Pielou, 1974). The complement was also carried out for the second species. The difference between the mean crowding indices between two species provides an estimate of the possible competitive superiority of a particular species. The assumption of

PAGE 177

169 this analysis is that a species which, on average, crowds a second species more than it is crowded by it, possesses superior competitive ability. Because it is not possible to obtain a measure of sampling variance, and thus of error terms, interpretation of these estimates is limited (Pielou, 1974). Chi-square analyses for presence and absence of species at trapping stations were also performed. However, it is felt that unless competitive effects are very strong, or are relative fine-grained in relation to the sampling methods used, the probability of a type I error of not rejecting a non-interaction null hypothesis is increased. This is due to the fact that overlap analyses do not take into account trapping success, but only presence (which in reality is represented by values ranging anywhere from 1 to the maximum number of captures at a particular trapping station) and absence of species. If we consider that perhaps the distance separating trapping stations in this study (20 meters) is not enough to allow for a fine discrimination of possible spatial segregation between species, the results of this type of analysis should be carefully interpreted. Predator pressure Evidence for the determination of the effects of mammalian predator pressure on small mammal community structure was collected both directly in the study sites and indirectly from data available in

PAGE 178

170 the literature. Direct sighting records were collected in the forest plots while running traplines. Additional information on carnivores was provided by employees of the farms, park personnel and local hunters. No direct censuses were conducted. The predator species that were considered as having a direct influence on small mammal community composition were the ocelot ( Felis pardalis ) the jaguarundi ( Pel is yaqouaroundi ) the margay cat ( Fells wieldii ) the jaguar (Felis onca ) and the tayra ( Eira barbara ). Data on predator diets and predatory behavior came from the following sources: Bisbal (1986), Bisbal and Ojasti (1980), Emmons (1987), Enders (1935), Hunsaker (1977), Konecny (in press), Ludlow and Sunquist (1987), Mondolfi (1986), and Rabinowitz and Nottingham (1986). Results Species Composition and Abundance Patterns Over a period of 17 months, 19 small mammal species were recorded in the six forest plots taken together (Table 4-1), Species richness ranged from six, in the small and in the large primary forest plots, to 14 in the large secondary forest. In total I obtained 1,366 captures of 692 individuals (Table 4-1).

PAGE 179

171 Table 4-1. Total biomass (in grams), number of individuals (N), and total number of captures (TC) recorded for each species during 17 consecutive months of trapping at the six different forest plots. FOREST PLOTS RC FMC RD Small (80 / 60 ha.) Medium (860 ha.) Large (35980 ha. ) Primary Secondary Primary Secondary Primary Secondary Caluromys 1080 philander N=06 TC=07 Didelphis 41514 13915 19961 24344 8970 3046 marsupialis N=41 N=18 N=40 N=30 N=11 N=04 TC=115 TC=56 TC=130 TC=66 TC=32 TC=05 Marmosa 52 122 aqilis N=03 TC=04 N=04 TC=07 Marmosa 582 3636 cinerea N=08 TC=26 N=35 TC=94 Marmosa 431 289 1020 2371 547 3469 incana N=05 N=05 N=15 N=39 N=07 N=56 TC=06 TC=09 TC=33 TC=79 TC=07 TC=86 Marmosa 31 microtarsus N=01 TC=01 Metachirus 3300 260 6390 4454 7334 7451 nudicaudatus N=13 N=01 N=16 N=18 N=26 N=31 TC=16 TC=2 TC=26 TC=26 TC=33 TC=53 Philander 360 320 3072 515 opossum N=01 N=01 N=09 N=03 TC=02 TC=01 TC=23 TC=03 Abrawayomys ruschi 63 N=01 TC=01

PAGE 180

172 Table 4-1 continued. RC FMC RD Small (80 / 60 ha.) Medium i [860 ha.) Large (35980 ha.) Primary Secondary Primary Secondary Primary Secondary Akodon cursor 66 N=01 TC=01 55 N=01 TC=01 1428 N=27 TC=51 Nectomys sauamipes 345 N=02 TC=02 420 N=02 TC=03 746 N=06 TC=09 Oryzomys capito 60 N=01 TC=02 Oryzomys niqripes 20 N=01 TC=01 35 N=02 TC=02 39 N=02 TC=02 Oryzomys subflavus 92 N=01 TC=01 Oryzomys trinitatis 152 N=03 TC=03 913 N=15 TC=15 255 N=04 TC=04 2186 N=40 TC=46 288 N=04 TC=04 992 N=18 TC=27 Oxymycterus sp. 142 N=02 TC=02 Rhipidomys mastacalis 72 N=01 TC=01 Echymys sp. 225 N=01 TC=01 Proechimys setosus 1573 N=07 TC=07 546 N=03 TC=03 16997 N=80 TC=176 5330 N=20 TC=53 TOTAL 47330 N=70 TC=149 16660 N=49 TC=93 48375 N=169 TC=398 39427 N=157 TC=283 17776 N=57 TC=103 22308 N=190 TC=339

PAGE 181

173 Marsupials were by far the most common small mammals in each of the forests studied, both in terms of proportion of individuals in different sites (ranging from 65 % to 75 %) and in proportion of captures (ranging from 54 % to 96 %) If these proportions are converted to absolute biomass, the difference in favor of marsupials is overwhelming (63 % to 93 %). The two intermediate size forests were responsible for the largest number of rodent individuals, mostly belonging to two species: Proechimys setosus and Oryzomys trinitatis The largest secondary forest had the most number of rodent species (eight), while marsupial species ranged from four (small primary) to six species (large secondary). Four species, three marsupials ( Didelphis marsupial is Marmosa incana and Metachirus nudicaudatus ) and one rodent ( Oryzomys trinitatis ) were present in all six forests. Philander opossum and Proechimys setosus were present in four forests, but were absent from the Rio Doce Park, where the two large forest patches were located. Conversely, Marmosa cinerea was common on the Rio Doce Park, but conspicuously absent from all of the smaller plots. These seven species were responsible for 91 % of all individuals trapped over the whole period of the study. The other members of the communities were not frequently caught and eight species were only present in one forest plot, usually with only one or two individuals.

PAGE 182

174 In secondary forests, Oryzomys tr1n1tat1s and Marmosa incana achieved their highest densities (One-way ANOVA, p<0.001), while Didelphis marsupialis and Proechimys setosus populations were more dense in primary forests (One-way ANOVA, p<0.01). On the other hand, Metachirus nudicaudatus and Philander opossum 's abundances did not seem to be affected by any measured structural or botanical difference existing between primary and secondary forests (One-way ANOVA, p>0.10). The largest secondary forest was the richest site both in number of species (14), and in total number of individuals (190) observed over the course of 17 months of trapping. Both the small and the large primary forests had few individuals (70 and 57, respectively) and fewer species (six each). The forests of intermediate sizes and the small secondary forest all yielded nine small mammal species, even though the small primary forest supported the least number of individuals of all plots (Table 4-1). Metachirus nudicaudatus Marmosa incana and Proechimys setosus captures were higher in forest plots of increasing size (One-way ANOVAS, p<0.001). Didelphis marsupialis captures demonstrated a reverse trend, with smaller plots displaying the highest number of captures, followed by the mid-size and the large plots, respectively (One-way ANOVA, p<0.001). For other species, the small sample size did not allow statistical analyses.

PAGE 183

175 Table 4-2. Results of the discriminant analyses using habitat variables for each species. The discriminant function separates groups of trap stations where species was caught in any forest plot against all available habitat represented by all trap stations in all forest plots. Only species with sample sizes greater than 20 trapping records were used in the analysis. Wilk's lambda values and associated p-values refer to the manova model as measured by lambda (*=p<0.05; **=p<0.01; ***=p<0.001) Species % Correct Wilk's lambda Didelphis marsupial is 55 Metachirus nudicaudatus 58 Marmosa incana 62 Marmosa cinerea 58 Philander opossum 87 Proechimys setosus 71 Oryzomys trinitatis 58 Akodon cursor 77 0.969 n.s 0.952 n.s 0.923 n.s 0.980 n.s 0.817 *** 0.880 0.954 n.s 0.878 n.s

PAGE 184

176 Habitat Selection Despite the fact that two species (Marmosa incana and Oryzomys trinitatis ) achieved their higher abundances in secondary forests, while another two (D^ marsupial is and P^ setosus ) were more commonly found in primary forests, all of these species, with the exception of Proechimys setosus appear to use their habitat randomly with respect to available habitat across all forest plots (Table 4-2). In other words, by using aggregate data which did not separate primary and secondary forest trapping stations, habitat profiles for species did not significantly differ from habitat profiles of available habitat, as indicated by the p-values associated with Wilks' lambda scores. The discriminant functions, furthermore, had a very low power of discriminating between trapping stations where species were and were not recorded (Table 4-2). The exceptions were P^ setosus and P. opossum whose habitat profiles differed from available habitat, even though their functions did not have a very high discriminating ability (Table 4-2). At forests where Proechimys setosus was caught, the preferred trapping stations were those where the habitat consisted of taller than average trees, of a well developed canopy cover, sparse herbaceous growth and where fallen logs occurred frequently (Table 4-3). Philander opossum also seems to prefer tall primary forest areas, but was also commonly associated with moist situations. The small sample

PAGE 185

177 Table 4-3. Results of univariate analysis of variance (ANOVA) of habitat variables for species against habitat availability (P=probabilities associated with F-tests, where *=p<0,05; **=p<0.01; ***=p<0.001). Dm. Mn. Mi. Mc. Po. Ps. Ot. Ac. HT ***(+) ***(+) OH ***(+) HH *(-) **(-) CXC CXM **(+) *(+) *(+) CC HC *(-) ***(-) *(-) DVC DVM *(+) *(+) DVH *(-) ***(-) DV *(+) DE *(-) **(-) *(-) FL ***(+) HUM *(+) NSP ***(+) NTR *(+) BP **( -) *( + ) *( + ) TB *(-) *(-) *(-) ***(-) Ns 138 108 104 55 21 76 57 20 Nh 288 288 288 96 144 192 288 144 Note: The + ( Dr signs in the vari ables that were significantly different from available habitat refer to the direction of difference as determined using pairwise comparison of means with the GT2 method. For species that occurred in all forest plots ( Didelphis marsupial is Marmosa incana Metachirus nudicaudatus and Oryzomys trinitatis ) available habitat is measured by means of all trapping stations pooled. For other species, available habitat was measured by means of habitat variables only from forests where species was recorded. Ns=number of stations where species was caught; Nh=number of stations used in describing available habitat. Variable abbreviations can be found in methods section, and species abbreviations are as follows: Dm= Didelphis marsupialis : Mn^ Metachirus nudicaudatus : Mi ^ Marmosa incana : Mc=Marmosa cinerea : Po= Phi lander opossum : Ps= Proechimys setosus; Ot= 0ryzomys trinitatis : Ac=Akodon cursor

PAGE 186

178 size of P^ opossum captures, however, hinders further conclusions regarding habitat preferences. Although for most species the multivariate model was not significantly different from random use of available habitat, some trends were observed when the habitat profiles were analyzed univariatly. Didelphis marsupialis despite being a generalist, avoided habitats with a well developed herbaceous understory and and also absent from bamboo stands (Table 4-3). Marmosa incana occurs strongly associated with habitat features characteristic of mid-stage secondary forests, with a well developed and viny mid-story, but also avoided areas dominated by bamboo. Metachirus nudicaudatus Oryzomys trinitatis and Akodon cursor to a lesser extent, were also associated with secondary forests. Marmosa cinerea was the most thoroughly habitat generalist in the forests where this study was conducted. Its discriminant function was not significant and none of the habitat variables measured proved significantly different from the overall forest means. For both Proechimys setosus and Phi lander opossum whose multivariate models indicated habitat selectivity, the discriminant functions were applied to forest patches where they were not recorded, in an effort to explain their absences in terms of habitat differences. Proechimys setosus is absent from the two largest plots, but nonetheless the discriminant function classified approximately 45 % of the trapping stations in both forest plots as

PAGE 187

179 Table 4-4. Results of discriminant analysis for classification of trap stations for Proechimys setosus, Marmosa cinerea and Philander opossum for forests where species was not recorded. Discriminant function was built using habitat variables for stations where species were caught against all available trap stations. Results are presented in table in percentage of trap stations classified as having species (Presence=Yes) or not having species (Presence=No). Species abbreviations are found in Table 4-3. rcTl rcTm fmc/m fmc/j rdTc rdTh Presence Yes No Yes No Yes No Yes No Yes No Yes No Ps. 43.8 56.2 45.8 54.2 Mc. 58.3 41.7 93.0 07.0 58.3 41.7 85.4 14.6 Po. 10.4 89.6 22.9 77.1 31.3 68.7

PAGE 188

180 potentially being able to support the species (Table 4-4). Conversely, the discriminant functions of Philander opossum when applied to other forests where it was absent, mostly confirmed this distribution pattern as a result of habitat differences, in general only classifying about 20 % of the stations as species' habitat (Table 4-4). The same analysis was also attempted for the general ist Marmosa cinerea very common on the large plots but absent from all others. The function classified over 85 % of the stations on the secondary forests of medium and small size as Marmosa cinerea habitat. With this species, however, the discriminant function did not reliably classify trap stations in forests where the species was indeed caught, and so the applicability of this analysis is not strong. Nevertheless, it indicates that the species is a general ist, and is probably absent from the smaller plots due to reasons other than available habitat. In summary, Proechimys setosus does not occur in the large forest plots for reasons other than habitat availability. Similarly to P^ setosus small and medium size forests are probably well suited for Marmosa cinerea and the reasons for its absence must be searched elsewhere. Phi lander opossum on the other hand, is not present in the large forests probably because of the lack of running or standing water close to traplines and therefore these forests do not provide suitable habitat (Nowak and Paradiso, 1983).

PAGE 189

181 Table 4-5. Mean number of captures of species at trapping stations where Didelphis marsupialis was also captured ( Didelphis stations) and mean number of captures of species at stations where Didelphis marsupial is was never captured (NonDide1phis stations). Also shown the is the mean trapping success for rodents in stations where marsupials were and were not captured. P-values indicate significance of univariate F-tests, where *=p<0.05, **=p<0.01, ***=p<0.001. Sd=standard deviation of untransformed variables; r=range of untransformed variables; N=sample size for Didelphis and NonDidelphis stations, and marsupial and non-marsupial stations, respectively. N Didelphis stations NonDidelphis stations M. nudicaudatus 170/92 0.41(sd=0.74;r=0-5) 0.93(sd=1 .00; r=0-5) *** M. incana 170/92 0.69(sd=1 .22;r=0-6) 1 .07(sd=2.03; r=0-16) 29/61 0.11(sd=0.45;r=0-4) 1 10(sd=1 70; r=0-10) ** 21/31 0.13(sd=0.62;r=0-5) 0.08(sd=0.37; r=0-3) n.s. 170/92 0.32(sd=0.74;r=0-4) 0.45(sd=1 14; r=0-9) n.s. 77/95 1.23(sd=2.04;r=0-9) 0.32(sd=1 19; r=0-8) n.s. A. cursor 29/61 0,02(sd=0.19;r=0-2) 0.52(sd=1 62; r=0-1 1 ) M. cinerea P. opossum 0^ trinitatis P. setosus N Marsupial stations Non-Marsupial stations ALL RODENTS ALL RODENTS EXCEPT P. setosus 253/19 1.48(sd=2.23;r=0-11) 1 1 1 (sd=0.60; r=0-3) n.s. 253/19 0.55(sd=1.37;r=0-11) 1 .00(sd=0.71 ; r=0-3)

PAGE 190

182 Table 4-6. Lloyd mean crowding indices averaged over all trapping stations. Indices on top of matrix refer to the mean crowding of row species on column species; the bottom part of the matrix refers to the reciprocal effect. Abbreviations of species names can be found in Table 4-3. Dm. Mn. Mi. Mc. Po. Ps. Ot. Ac. 1.01 1.29 0.16 0.81 0.43 0.91 0.08 1.20 1.00 0.18 0.00 0.88 0.91 0.35 0.91 Dm. — Mn. 0.40 Mi. 0.67 Mc. 0.30 Po. 0.08 Ps. 0.65 Ot. 0.24 Ac. 0.11

PAGE 191

183 Species Interactions Patterns of species interactions were analyzed for eight species for which sample size was sufficient ( Didelphis marsupial is Metachirus nudicaudatus Marmosa incana Marmosa cinerea Phi lander opossum Proechimys setosus, Oryzomys trinitatis and Akodon cursor ) Differences in mean trapping success of a species when a second one was present indicated that only one species, Didelphis marsupial is depressed the trapping success of another. Therefore, all subsequent analysis investigated the effect of common opossums on trapping success of other species. Mean trapping success of three marsupial and one rodent species were reduced when Didelphis marsupial is was present on the same trapping station (Table 4-5). Opossums also negatively depressed the capture success of aggregate trappings of Oryzomys trinitatis and Akodon cursor (Table 4-5). The same results can be observed by examining Lloyd mean crowding indices for Didelphis marsupialis (Table 4-6). Opossums depress trapping success of three marsupial and one rodent species more than they are, on a station by station basis, depressed by any of the second species. Overlap analysis indicated negative interactions between opossums and Metachirus nudicaudatus (Chi-square=5.91 p<0.02), Marmosa cinerea (Chi-square=4.30, p<0.05) and Akodon cursor (Chi-square=15.30, p<0.001), and no statistically significant negative interactions between any other pairs of species. These

PAGE 192

184 results are conservative because they do not take into account trapping success, but presence or absence of species from trapping stations. The frequency with which Didelphis marsupial is interacts with other species appear to be dependent on forest size (Table 4-7). Five species which had at least one forest plot of two size categories in common with opossums showed that their overlap increase from small to medium size and then subsequently decrease in large plots. These results might indicate possible effects of species interactions or alternatively reflect the number of possible pair-wise cases. Nonetheless, they also indicate that opportunities for interactions are reduced as one goes from either a small or medium sized forest to a large one (Figure 4-1). Predation Table 4-8 lists the mammalian predator species recorded at each of the study sites which are known to prey on some of the small mammals trapped in those forests (Table 4-9). Even though this study did not include direct census, I was able to estimate potential predator populations by extrapolating known densities at other tropical areas (Table 4-10). As I expected, the small size forests had very few predator species and these were not commonly

PAGE 193

185 PERCENT OVERLAP MN Ml SMALL OT SPECIES MEDIUM PS LARGE PO Figure 4-1. Percent overlap of trap stations between D^ marsupial 1s and M^ incana M. nudicaudatus 0. trinitatis and P^ opossum and its variation in forest plots of different sizes. For each size category, both primary and secondary forest data were lumped.

PAGE 194

136 Table 4-7. Proportion of stations with overlap between Didelphis marsupialis and a second species. Only trapping stations where at least one capture of one of the species was recorded were included in the analysis. Overlap indices are presented at each of the forest plots, as well as for all stations lumped at each size category irrespective of habitat structure (Overall). Species abbreviations as presented in Table 4-3. PR=primary forest; SC=secondary forests. Didelphis marsupialis Small Medium Large PR SC OVERALL PR SC OVERALL PR SC OVERALL Mn. 0.50 0.04 0.19 0.65 0.56 0.60 0.31 0.07 0.17 Ml. 0.25 0.19 0.21 0.70 0.69 0.70 0.14 0.07 0.09 Ot. 0.25 0.19 0.21 0.11 0.65 0.53 0.08 0.02 0.04 Po. 0.10 0.00 0.03 0.50 0.14 0.33 0.00 0.00 0.00 Ps. 0.31 0.12 0.18 0.83 0.63 0.75 0.00 0.00 0.00 Mc. 0.00 0.00 0.00 0.00 0.00 0.00 0.23 0.10 0.09 Ac. 0.00 0.00 0.00 0.00 0.08 0.08 0.04 0.02 0.03

PAGE 195

187 Table 4-8. Carnivores recorded for each of the forest sites studied. Fazenda Esmeralda Fazenda Montes Rio Doce Park Rio Casca Claros (FMC) (RD) (RC) Small Patches Medium Patches Large Patches Felis pardalis Felis yaqouaroundi Felis wieldii Felis concolor Panthera onca Eira t Darbara Galictis vittata Cerdocyon thous t t *

PAGE 196

188 Table 4-9. Small mammals of this study which are prey species of neotropical carnivores. Ocelot Jaguarundi Margay Jaguar Tayra Grison Fox Studies Didelphis marsupial is Philander opossum Marmosa spp. Metachirus t nudicaudatus Marmosa cinerea Proechimys spp. Oryzomys capito 1.2.4.5.6.7.8.9 6.9 4.6.7 4 1.3.4.7 References: 1 Bisbal (1986), 2 Bisbal and Ojasti (1980), 3 Enders (1935), 4 Emmons (1987), 5 Hunsaker (1977), 6 Konecny (in press), 7 Ludlow and Sunquist (1987), 8 Mondolfi (1986), 9 Rabinowitz and Nottingham (1986).

PAGE 197

189 Table 4-10. Densities of ocelots, jaguars, pumas, jaguarundis and tayras, and potential population sizes of forest sites of this study. Data on ocelots, jaguars and pumas were taken from Emmons (1987), jaguarundis from Konecny (in press) and tayra from Sunquist et al. (in press). Body Weight Ocelots 7-12 kg Jaguars 31-37 kg Pumas 29 kg Jaguarundis 4-6 kg Tayras 5 kg Recorded Density per km2 0.4 0.05 0.07 0.05 0.11 Fazenda Esmeralda 0.6/0.8 km2 Potential Potential Potential Potential Potential Number of Number of Number of Number of Number of Ocelots Jaguars Pumas Jaguarundis Tayras Fazenda Montes Claros 8.6 km2 11 0.4 0.6 0.4 Rio Doce Park 360 km2 280 18 25 18 39

PAGE 198

190 Table 4-11. Relative importance of Didelphis marsupial is ( Dm ) Proechimvs sp. ( Psp. ) and Phi lander opossum ( Po. ) in the diets of ocelot and jaguarundi. Only mammals, that comprise the bulk of these predators' diets, are considered here. The first number between parentheses refers to rank of importance in diet, and the second refers to the total number of mammalian items in diet. Ocelot Studies Ludlow and Konecny Emmons (1987) Sunquist (1986) (in press) % Frequency % Biomass % Frequency % Biomass % Frequency % Biomass in scats in diet in scats in diet in scats in diet Dm. 2.3(7/14) 16.8(4/10) 38.7(1/8) 16.9(3/8) 1.7(8/16) 3.4(7/16) Psp. 46.7(1/16) 26.6(1/16) Pg^ 30.6(2/8) 3.9(5/8) Jaguarundi Study Konecny (in press) % Frequency in scats % Biomass in diet Dm. 13.0(4/6) 58.0(1/6)

PAGE 199

191 sighted at the study sites. The medium-sized forests contained both ocelots ( Felis pardalis ) and jaguaroundis ( Fe 1 i s yaqouaroundi ) felids which are known to prey heavily on both Didelphis marsupial is and Phi lander opossum (Table 4-11). The large plots possessed a quite rich and large population of mammalian predators. Although not quantified, sights and tracks of ocelots (and possible other smaller cats) and jaguars were recorded at the Rio Doce Park more frequently than in the medium size plots. The importance of the small mammals present in the communities on the diets of these mammalian predators can be seen in Table 4-11. Data from Ludlow and Sunquist (1986 and unpublished data), Konecny (in press) and Emmons (1987) indicate that both Didelphis marsupial is and Proechimys spp. are very often included in the diets of ocelots, both in terms of frequency of items and in total volume or biomass. Information on jaguaroundis is scanty, but in the only study available, D^ marsupialis remains were recorded in 13 % of all scats collected. If this figure is converted in total biomass in diet, it would make opossums the first ranking prey species (Table 4-11). Additional evidence of mammalian predation was found during trapping. In the small-sized forests no animal ever died in a trap as a result of predation. In both the medium and large plots several traps either completely disappeared from stations or were found over 30 meters away, frequently containing the remains of the

PAGE 200

192 partially eaten small mammals. In one of these instances a trap with an 1800 grams male opossum was found 50 meters away from the station. The animal had its leg pulled through a gap on the trap door, which was then partially eaten. Proechimys are known to lose their tails easily (Hershkovitz, 1969; Nowak and Paradiso, 1983), and this is commonly interpreted as an anti-predator mechanism. It was thus interesting to find that about 12 % of the animals trapped at the medium size forests had their tails broken, while all individuals trapped in the smaller plots had theirs intact. No spiny rats were trapped on the larger plots. Discussion Patterns of relative abundances among small mammals have been previously linked to microhabitat requirements, spatial segregation and competitive interplays (Rosenzweig and Winakur, 1969; M'Closkey, 1976; Dueser and Shuggart, 1978; August, 1983; Nitikman and Mares, 1987). Results of this study have indicated that the four species common to all six forest plots, regardless of their succession stage, did not select microhabitat from available habitat across all forests investigated. A fifth species, Proechimys setosus, present in all plots of medium and small size, did appear to be more selective, as its discriminant function was

PAGE 201

193 marginally significant. However, no habitat reason seems to prevent the species from being present in larger plots. Therefore, differences in species occurrences and relative abundances between forest plots cannot be successfully explained by microhabitat requirement distinctions. The one exception appears to be the gray two-eyed opossum ( Phi lander opossum ) which may depend on availability of water. However, some of the species were found to be either more or less abundant at different forest sites. If most species are not selecting habitat from available space, what accounts for differences in relative abundances? The answer could consistently lie within the framework of competitive interactions. The present analysis failed to detect strong negative effects between pairs of species, but with the noted exception of Didelphis marsupial is While most species' populations increase in number with the area size of the forest, D^ marsupial is relative abundances experience a reverse dynamic, especially among secondary forests. If we take absolute biomass as a measure of species dominance, this trend is even more pronounced (Figures 4-2 and 4-3). This suggests that the common opossum, a clear general ist, negatively affects the carrying capacities of other species in the forests where they coexist. On a microhabitat basis, D^ marsupialis depressed the densities of most other species for which sample size was sufficient to allow

PAGE 202

194 Proportion SMALL PLOT D. marsuDialis i:iil::J Q, tfinitatIS MEDIUM PLOT I i M. nuOicaudatus I I Other Species LARGE PLOT ^ M incana Figure 4-2. Proportion of relative biomasses of D^ marsupialis M^ incana M^. nudicaudatus 0^ trinitatis and remaining species together, as function of size category, for secondary forests only.

PAGE 203

195 1 0.80,60.40.20SMALL PLOT Q. marsuDialis Omn] Q, trinitatis vmmi > MEDIUM PLOT di M. nunirauriatus I I Other Species LARGE PLOT M inc3na Figure 4-3. Proportion of relative biomasses of D^. marsupialis M^. incana \A^ nudicaudatus 0^ trinitatis and remaining species together, as function of size category, for primary forests only.

PAGE 204

196 analysis, and was the only species which was found to negatively affect other species' relative abundances. Where opossums are frequently caught, Metachirus nudicaudatus Marmosa incana Marmosa cinerea and Oryzomys trinitatis occur in significantly lower densities. The potential depressing effect of common opossums on other small mammal species is possibly present in most south American tropical communities. Didelphis are the most general ist and versatile animals among neotropical small mammals. Didelphis marsupial is according to Hershkovitz (1969; page 35), "surpasses all other living didelphids, and perhaps all other living mammals, in the combination of adaptability, tenacity, viability, vagility and phyletic longevity." The genus Didelphis occurs in almost all vegetation formations (Hunsaker, 1977; Nowak and Paradiso, 1983; Streilein, 1982), from southern Canada to central Argentina, including even suburban settings. The range of Dj_ marsupialis is also very extensive, from Central America to southern Brazil and Paraguay (Hunsaker, 1977; Streilein, 1982). Opossums can affect other small mammal species in two ways. First, the diet and spatial use of opossums overlaps with almost all small mammals of the communities where it exists (e.g., Charles-Dominique et al., 1981; Miles et al., 1981), and in particular with other sympatric didelphids. All didelphids are regarded as insectivore/omnivores or f rugivore/omnivores (Lee and

PAGE 205

197 Cockburn, 1985). Charles-Dominique (1983), in a French Guyanan forest, found Didelphis marsupialis diet to be the broadest amongst five sympatric didelphids, and to utilize all major food items of other marsupial species, with the exception of gums and flowers. A diet composition study of D^. marsupialis in Venezuela indicated the use of both animal (63.5 % in volume) and plant foods (12.8 % in volume). About one fourth of the volume of opossum diet consisted of insects and fruits (Cordero and Nicolas, 1987). Another remarkable feature of opossums is that they shift substrate utilization preference and diet composition with age (see Chapter 2). Young eat almost exclusively insect and plant material, while adults start relying more on vertebrate prey and carrion (Cordero and Nicolas, 1987). Juveniles show a higher degree of arboreal ity than adults, while the latter become increasingly more terrestrial in older age classes. This places the species in potential spatial competition with several mostly arboreal and/or scansorial species, which also frequently consume insects and fruits (e.g., Marmosa cinerea Marmosa incana Caluromys philander Oryzomys trinitatis ) and with exclusively terrestrial ones ( Metachirus nudicaudatus Philander opossum Proechimys setosus ). Nests and refuges can be found in tree cavities, tangled vines and tree forks, and burrows (Miles et al., 1981), resources which are shared with other small mammals. Miles et al. (1981) determined that Didelphis marsupialis can wander

PAGE 206

198 through the canopy with remarkable agility, being able to use even the smaller vines as paths between different trees. These are substrates frequently used by other small mammals. Since weaning takes place when young are around 70 to 100 grams, their phase as independent foraging individuals overlaps in size with all coexisting adults of other species in the community. Marmosa cinerea a species in which adults are very similar to young opossums in morphology and behavior, was found to be negatively affected by Didelphis marsupial is despite the fact that adults of the latter species are 10 times larger than the former. Interestingly, M^ cinerea was common only in forest plots with very low Di. marsupial is densities. Secondly, opossums act as predators of small mammals and other vertebrates, especially nestlings (Fleming, 1972; Eisenberg and Thorington, 1973; Dalby, 1975; Nowak and Paradise, 1983; Charles-Dominique, 1983; Hunsaker, 1977; Cordero and Nicolas, 1987), although hard evidence of the magnitude of predation is still lacking. Vertebrates are consumed by Didelphis most probably as carrion, but occasional and opportunistic attacks on live prey have also been recorded. In this way, D^ marsupial is may affect other species through both direct competition for food resources and by predation on small mammals, the former being probably more important.

PAGE 207

199 The variation in abundances of small mammal species found to be affected by common opossums can be explained by differential predator pressure in larger plots. Didelphis marsupial is trapping success decreases with size of the forest plots, and this variation is apparently not a consequence of variation in habitat. Instead, predator pressure alters community structure patterns of these forests. Since mammalian predators and other top carnivores do require large areas to maintain demographically healthy populations, it was expected that the smaller forest plots would possess fewer species and smaller carnivore populations. This prediction was supported by the results obtained in this study. Furthermore, data compiled from available literature indicate Didelphis marsupial is and Proechimys spp. as major prey items for mammalian carnivores in neotropical forests (see Hershkovitz, 1969; Emmons, 1987; Ludlow and Sunquist, 1987; Konecny, in press). Some studies have shown that predators do take quite a large percentage of some small mammal populations (Pearson, 1985; Emmons, 1987). August (1983) has implicated mammalian predators as the agents responsible for low densities of small mammals in a site in the Venezuelan llanos. Common opossums as large as 600 grams and other small mammals have been observed falling prey to great horned owls ( Bubo virqineanus ) in Paraguay (Wright, 1985). Further evidence can be found in data recently collected by Heideman et al. (1987) in Philippine forests. Small mammal species richness was

PAGE 208

200 determined to be higher in forests where carnivores were also more abundant, although the authors did not suggest any possible connection. Predation has been implicated as being a major factor structuring communities (Connell, 1961; Paine, 1966), by either preventing any one species of becoming extremely abundant or by not permitting dominant species to outcompete and locally drive other species to extinction (Emlen, 1973; Bourliere, 1983). Connell (1975) went even further to suggest that predators should be more effective in reducing competition between prey species in more benign environments. This would suggest that the effect of predators on competing prey should perhaps be more prominent in tropical forest communities. Karr (1982), and Loiselle and Hoppes (1985), for different reasons, have partially implicated predation in the loss of bird species at Barro Colorado Island. Small predators, which are mostly responsible for nest predation, would have increased in number as a response to the loss of larger predators on the small and isolated island. The Barro Colorado Island community has very few large mammalian predators, and the distribution of relative biomasses of small mammals is very similar to that of the intermediate size forest of this study. The BCI data on small mammals (Glanz, 1982) provided a remarkable fit with the patterns found here, suggesting

PAGE 209

201 that the lack of predators may have resulted in common opossums dominating the small mammal community (Figure 4-4). The data from Eisenberg et al. (1979) suggests that didelphids, and particularly Didelphis marsupialis are not well represented in terms of biomass, relative to rodents, in the Guatopo National Park in Venezuela. The park has the full complement of predators (Eisenberg, pers. comm.), which could have exerted an impact on D^ marsupialis At the smaller Hato Masaguaral site in the Venezuelan llanos, marsupial biomass was twice that of rodents. The smaller Masaguaral site has a poor carnivore fauna, further supporting the predator hypothesis put forward here. Predation also constitutes a viable explanation for the lack of spiny rats in large plots, since efforts in trying to link relative abundances of Proechimys setosus with possible competitive interactions have not been successful. In addition, the discriminant function suggested that the forest plots that lacked spiny rats were apparently suited for the species. Proechimys setosus is otherwise fairly common on other forests not very distant from the Rio Doce Park (see also Davis, 1946). Furthermore, Stal lings (in press) has sampled four other forests in different regeneration stages in the Rio Doce Park, therefore over a wider spectrum of habitats, and has also failed to record P^ setosus Predator pressure, on the other hand, should be more or

PAGE 210

202 Proportion 0.8 0,60,4 0.2 SMALL PLOT MEDIUM PLOT BARRO COLORADO LARGE PLOT D. marsupialis I i Proechimvs spp. Other Species Figure 4-4. Proportion of relative biomasses of Ds. marsupialis M. incana M. nudicaudatus 0^ trinitatis and remaining species together, as function of size category, for both primary and secondary forests lumped. The data for Barro Colorado Island come from Glanz (1982), and relative proportions were calculated for only the small mammal portion of the BCI mammalian community.

PAGE 211

203 less constant throughout the different habitats of the Rio Doce Park, and therefore exert a similar effect among the distinct communities. Spiny rats have, on an anecdotal account, been described as "the base source of protein for lowland predators of the Brazilian Subregion." (Hershkovitz, 1969, page 45). Recently, Emmons (1987) provided quantitative evidence demonstrating that in the Peruvian tropical forest of Cocha Cashu, ocelots take 39 kg/km^/year of Proechimys spp. out of a standing crop biomass of 61 kg/km^/year. It should be noticed that ocelots are not the only predators of Proechimys and as well as of Didelphis marsupialis and the magnitude of predation is thus certainly higher. However, even under high predator pressure, one could argue that Proechimys setosus should appear in the sample in a long-term study. That need not be the case. Although a rare event, in this study five species were represented by only one capture, while two of these were only recorded in the last two months of field work. Furthermore, only three individuals of P. setosus were trapped in 17 months of sampling in one of the forests. If in very low population density, a species may not be registered in a even large sample. Predator pressure thus remains the only viable f — — explanation to the absence (or most probably, undetectable density) of spiny rats from the larger forest plots. Because of small sample sizes it is difficult to generalize the conclusions of the effects of predation to all members of the small mammal community. Nonetheless, there are suggestions that Didelphis

PAGE 212

204 marsup1a11s by locally depressing the relative abundances of other species, may affect the species richness and species diversity of these communities. Chapter 3 showed that larger forest plots, especially secondary forest sites, tend both to have a higher number of species and also to be more diverse. Here I determined that the number of D^ marsupialis (N) on a transect trapline significantly affects the total number of species present on a trapline (S=8.88 1.38 ln(N), R2z0.38, p<0.007, n=18). If we separate the analysis by forest type, Didelphis abundances account for a larger amount of variation in the number of species of secondary forests (R2=0.43)' but do not affect species richness of primary forests. Predation on the rodent species Proechimys setosus may also have a positive impact on other species in the community. Although no competitive interactions were detected between spiny rats and other species, I noted that eight of fourteen species in the largest secondary plot, where no Proechimys were recorded, were also rodents (57 %), in sharp contrast with the other forests, which had from two to five species. Of these, the three Oryzomys species are trophically similar to Proechimys (Eisenberg and Thorington, 1973), while the diets of two others ( Oxymycterus sp. and Akodon cursor) have the insectivore and, to a lesser extent, the frugivore portions of their diets overlapping with spiny rats. Therefore, the existence of competition for resources is not an unlikely possibility.

PAGE 213

205 An alternative to the predation hypothesis presented here relates to density compensation mechanisms (Cody, 1975). One could contend that populations of species do go extinct in small areas due to demographic/stochastic and genetic considerations (MacArthur and Wilson, 1967; Soule, 1980), and thus independent of predator pressure. The higher abundances of Didelphis marsupial is could be explained as the result of the species broadening its niche space where it became less constrained through the local extinction of other species (MacArthur et al., 1972). Two observations refute this hypothesis. First, the sizes of the small forest plots in this study are probably not enough to readily drive local small mammal populations to extinction. In addition, recolonization potential among tropical small mammals may be very high in not truly insular circumstances, and this is probably the situation here (Malcolm, 1987; see also the problem of "compensatory effects," as suggested by Lomolino, 1986). Both aspects would prevent, or at least retard, demographic collapses in the small forests. A further indication arises from the observation that both medium size and one small size forests had the same number of species (nine), which is even more than one of the large forest plots (six), suggesting that local extinction is an infrequent occurrence. Secondly, the fine-grained pattern of the interactions observed here between D^ marsupialis and other species suggests that indeed active competition and possibly predation by opossums is taking place.

PAGE 214

206 We can not totally discard, however, possible compensatory responses in smaller plots, specially in the form of "excess density compensation" (Case et al., 1979; Faeth, 1984). This particular form of response should occur when most species missing, or at much lower densities at that community, are represented by interference competitors. In this way, resources that were released following the loss (or reduction in population sizes) of other community members, are much more effectively used by remaining species. In fact, interference competition would constitute the general mode operating at these marsupial and rodent communities, as species have broad overlap in resource use, and do not actively defend resources. Therefore, it may not be coincidental that the communities with the highest biomasses were those in smaller and medium size forests. I have shown that small mammal species are able to respond to a decrease in opossum densities in secondary forests (Figure 4-2), but the effect is less conspicuous in primary forests (Figure 4-3). This is most probably related to forest structure. Small mammal diversity has been found to be dependent, on variable levels of magnitude, on habitat structural parameters, and especially on habitat heterogeneity (Rozenzweig and Winakur, 1969; M'Closkey, 1976; Dueser and Shugart, 1978; August, 1983). As it was hypothesized before (Chapter 3), the relative homogeneity of primary forests in relation to secondary stages, both horizontally and vertically, prevents additional species to be successfully packed in the available habitat, even if area is

PAGE 215

207 increased. Therefore, since primary forests are resource poor from the start, even if predation reduces opossum densities in forests of larger sizes, this does not result in more resources being released to other potential members of the community. Conversely, the structural complexity of secondary forests can certainly be used to increase community membership and to maintain populations at higher levels, all which result in increased species richness and diversity. For the above reasons, the debate on whether to use univariate or multivariate methods in describing species habitat utilization (Boecklen, 1986; Willig et al., 1986; Blake and Carr, 1987) is indeed an important one. Most species in this study were found to be non-selective by analyzing overall habitat variable group differences. However, habitat utilization by several small mammals were found, in variable degrees, to be univariatly correlated with some environmental parameters, specially ones which characterize mid-stage secondary forests. Therefore, even though spatial use on these forests may be primarily random, species may utilize it in slightly different manners. This may indicate that resource partitioning has developed in response to competition, which may be important in a guild with such broad habitat and diet overlaps. The degree of resource overlap, be it in form of habitat use, in diet, or in on other critical resource for the species, has been frequently utilized as a measure of competition (as in M'Closkey, 1976; Cody, 1974; Meserve, 1981; Nitikman and Mares, 1987). This method has been criticized, as other hypotheses may also

PAGE 216

208 apply for a same habitat segregation outcome (May, 1975). Yet other studies conducted concomitantly with removal experiments, have both yielded conclusions implicating competition as a structural force in the community (e.g.. Price, 1978), as well as failing to detect it as a major factor (e.g., Murua et al 1987). Similar efforts to avoid the problems recognized by May (1975) are represented by other approaches which infer competition by analyzing both habitat requirements and species interactions in a same regression model (Hallett and Pimm, 1979; Hallett, 1982; Dueser and Porter, 1986; Lacher et al in press). However, as Pielou (1975) pointed out, the potential circularity involved in inferring competition from habitat segregation data is only a problem if ecological similarity has not been established. In the present study it is safe to state that most members of the community do have broad diet overlaps, and thus differential use of habitat would constitute a strong suggestion of the existence of competition avoidance mechanisms. The reason for failing to detect competitive interactions for all but one species may be linked to variable degrees of resource partitioning within the scale of a single trap station. The fact that species richness of the structurally complex secondary forests respond significantly more to increase in size than the relatively homogenous primary forests lends definite support to this possibility. The predation hypothesis is potentially a powerful one, as it can be directly tested with field experiments. Didelphis marsupial is is a

PAGE 217

209 species which can be readily trapped, has a wide distribution and by being extremely common does not pose problems with conducting natural removal experiments. Such studies should be urgently conducted, as the possible conservation implications of confirming these results are far reaching. The deforestation process in neotropical regions, especially in areas like the Brazilian Atlantic forest (Fonseca, 1985a), is approaching a stage in which conservation units of reduced size are the rule rather than exception. The maintenance of most biological diversity as possible may have to involve direct manipulation of species populations, as key elements of the communities such as top predators are unlikely to survive over the long run. Therefore, the artificial removal of portions of the populations of common, dominant species, may prove an useful tool in retarding processes of local extinctions and improving overall biological diversity. In addition, if one is searching for an easily detectable indicator species, the absolute densities of Didelphis marsupialis could prove a highly reliable index of species diversity.

PAGE 218

CHAPTER 5 CONCLUSION AND SYNTHESIS The results obtained in this study confirmed the previously suspected notion that Atlantic forest small mammal communities are diverse and rich (Davis, 1946; Laemert et al., 1946) especially when compared to adjoining biomes like the Cerrado (Alho and Pereira, 1986) and the Caatinga (Mares et al., 1981). Species richness is also comparable and often superior, to Amazonian tropical forests (see Malcolm, 1987; Konecny, in press). One of the most interesting features concerning species composition and relative abundances characterizing these communities is the overwhelming dominance of marsupial over rodent species. Although the aggregate data pooled across all forests plots indicate that there are more rodent than marsupial species, single communities were in general dominated by marsupials. This has seldom, if ever, been observed in neotropical forests. The result may be attributed to the widespread occurrence of second growth habitat, which is better able to provide marsupials with a more steadier and productive resource supply (Charles-Dominique et al., 1981). This problem deserves further investigation, since no single conclusion was definitely drawn. 210

PAGE 219

211 I should also stress that the trapping methods used were not adequate to represent the whole complement of the marsupial and especially the rodent communities. The thick-tailed opossum Lutreolina crassicaudata was not recorded probably because it occurs primarily in wetland habitat (Nowak and Paradiso, 1983), which was poorly represented in the samples of this study. The large terrestrial rodents Agouti and Dasyprocta were seen at all forests plots, but none was caught. The capybara Hydrochaeris is confined to riverine environments, and was commonly observed along rivers close to the study sites. However, there were not estimates of population density or biomass for these species, and comparisons between forest plots are not possible. Other mammals were occasionally trapped: the common long nosed armadillo Dasypus novemcinctus was a frequent occurrence in both smaller plots; although present at all forests, young Nasua were occasionally trapped at the medium sized plots; marmosets of the genus Cal 1 ithrix were present at all forests, but infrequently caught, the same being true for squirrels ( Sciurus aestuans). None of these species was, however, systematically trapped, hindering any comparisons between forests plots of different sizes and structure. Nonetheless, it is felt that the occurrence of all these species was not dependent on the two main variables measured in this study, i.e., forest size and habitat structure. Despite the traditional view that tropical forest diversity depends on the maintenance of climax stages (Denslow, 1985),

PAGE 220

212 disturbance was observed to promote species diversity in the communities studied, as was originally suggested by Connell (1978). Corroborating recent findings in other tropical small mammal communities (Charles-Dominique, 1983; Delany, 1986; Isabirye-Basuta and Kasenene, 1987), second-growth forest plots were found to be more rich and diverse than comparable size climax forests. Moreover, it became clear that there are probably very few or no small mammal species trapped in this study which one could safely regard as being obligatory primary forest dwellers. Lower productivity of primary forests may play a role in reducing species diversity, as small mammals rely mainly of fruits and insects, resources that may be more abundantly and steadily available in pre-climax stages. Species richness and diversity only responded to increase in area size between second-growth forest patches. Therefore, we were able to demonstrate that area-species relationships should only be applied when habitat features are taken into account. The island biogeographical extinction/colonization model (MacArthur and Wilson, 1967) was not incompatible with the results obtained, especially since this study dealt with species possibly with high colonization potential (Malcolm, 1987). But on the other hand, the proximate mechanism by which diversity in areas of increasing size is promoted and maintained is best described by the suggestion put forward by MacArthur (1957) and Williams (1964). Secondary forests are more diverse due to both the increase in alpha-diversity, possible by the presence of well developed

PAGE 221

213 mid-stories, and to the increase in beta-diversity, resulting from the existence of a larger number of patches of different structure. Species diversity of primary forests, in contrast, is less capable of responding to increase in size because they are characteristically resource poor habitats for small mammals. Species interactions can also have a direct effect on community structural parameters. As observed in other studies, patterns of relative abundances of small mammals can be influenced by microhabitat requirements, spatial segregation and competitive interactions (Rosenzweig and Winakur, 1969; M'Closkey, 1976; Dueser and Shugart, 1978; August, 1984; Nitikman and Mares, 1987). In this study, however, rodent and marsupial species, with rare exceptions, did not select microhabitats from available habitat. Therefore, we can exclude this variable as a possible source of variation in the patterns observed here. The species displayed a high degree of spatial and diet overlap, suggesting that there is a good possibility that competitive interactions may play a role in the structure of these communities. Indeed, the common opossum, Didelphis marsupialis was found to depress densities of several species of small mammals in this study. Opossums have undoubtedly the higher level of resource overlap with other species, and their relative abundances affect both the number of species in a forest plot as well as local abundances of individuals.

PAGE 222

214 The previous result leads naturally to the question of what influences Didelphis marsupial is densities between forest plots. Predation by other vertebrates, especially mammalian carnivores and owls, is the most likely explanation. Opossums are one of the most frequent and reliable prey items for mammalian carnivores in neotropical forest communities (Ludlow and Sunquist, 1987; Emmons, 1987; Konecny, in press). Since long-term viability of top predator populations depends on the size of the area, we should expect predation pressure to be lower (or nonexistent) in very small forest plots. Larger forest patches, such as those found in the Rio Doce State Park, possess a larger and more diverse fauna of predators, with consequent effect on opossum populations. With Didelphis numbers reduced by predation, other potential tenant species of these communities have access to resources previously exploited by opossums. But once more, only secondary forests should display such changes in community parameters in response to lower opossum densities, because primary forests are, in themselves, unable to maintain a higher number of species. These findings have significant consequences for fauna! conservation schemes for this region. It became evident that the vast majority, if not all of these small mammal species, do not depend on primary forests. On the other hand, the reduction in the area size of these forest patches certainly reduces species richness and diversity. It is possible that the widespread alteration of climax forests across

PAGE 223

215 the Atlantic coastal region of Brazil did not yet lead to massive extinctions of these small mammalian species, but the small area of the forests may in the long run lead to the extinction of some species. Given that only about 5 % of the region is presently forested (Fonseca, 1985), and that abandoned plantations and pastures exist (Fonseca, 1983), there is a chance that efforts to increase the size of protected areas will clearly yield beneficial results. Some younger forests and other abandoned fields that are not overexploited or eroded, have been observed to possess a high regeneration potential. These have to be considered a very important conservation resource and their preservation and protection should be actively stimulated.

PAGE 224

LITERATURE CITED Alho, C. J. R. 1981. Small mammal populations of Brazilian Cerrado: the dependence of abundance and diversity on habitat complexity. Rev. Brasil. Biol. 41 (1):223-230. 1982. Brazilian rodents: their habitats and habits. In: M. A. Mares and H. H. Genoways (eds.), Mammalian Biology in South America. Special Publication Series Volume 6, Pymatuning Laboratory of Ecology, University of Pittsburgh Press, Pittsburgh. L. A. Pereira. 1985. Population ecology of a Cerrado rodent community in Central Brazil. Rev. Bras. Biol. 45(4):597-607. and A. C. Paula. 1986. Patterns of habitat utilization by small mammal populations in cerrado biome of central Brazil. Mammalia 50(4): 447-460. Almeida, F. A. and M. Z. P. Rocha. 1977. Estabelecimento de areas minimas de preservacao dos di versos ecosistemas terrestres do Brasil. In: Encontro Nacional sobre Conservacao de Fauna e Recursos Faunisticos, organized by IBDF and Academia Brasileira de Ciencias, Brasilia, D. F. Alonso, M. T. A. 1977. Vegetacao. Regiao Sudeste. In: Geografia do Brasil. Institute Brasileiro de Geografia e Estatistica, IBGE, Rio de Janeiro, R. J. Atramentowicz, M. 1982. Influence du milieu sur I'activite locomotrice et la reproduction de Caluromys philander (L.). Revue d'Ecologie 36:376-395. August, P. V. 1983. The role of habitat complexity and heterogeneity in structuring tropical mammal communities. Ecology 64:1495-1513. 1984. Population ecology of small mammals in the Llanos of Venezuela. In: R. E. Martin and B R. Chapman (eds.), Contributions in Mammalogy in Honor of Robert L. Packard. Texas Tech. University, Lubbock, Texas. 216

PAGE 225

217 Austad, S. N. and M. E. Sunquist. 1986. Sex-ratio manipulation in the common opossum. Nature 324:58-60. Avila-Pires, F. D. and E. Gouvea. 1977. Mamiferos do Parque Nacional do Itatiaia. Bol. Mus. Nac. 291:1-29. Bigger, M. 1976. Oscillations of tropical insect populations. Nature 259:207-209. Bisbal, F. J. 1986. Food habits of some neotropical carnivores in Venezuela (Mammalia, Carnivora). Mammalia 50(3): 329-339. and J. Ojasti. 1980. Nicho trofico del zorro Cerdocyon thous (Mammalia, Carnivora). Act. Biol. Venez. 10(4):469-496. Bishop, I. R. 1974. An annotated list of Caviomorph rodents collected in North-Eastern Mato Grosso Brazil. Mammalia 38(3):489-502. Blake, J. G. and J. R. Karr. 1987. Breeding birds of isolated woodlots: area and habitat relationships. Ecology 68:1724-1734. Boecklen, W. J. 1986. Effects of habitat heterogeneity on the species-area relationships of forest birds. J. Biogeogr. 13:59-68. Borchert, M. and R. L. Hansen. 1983. Effects of flooding and wildfire on valley side wet campo rodents in Central Brazil Rev. Bras. Biol. 43(3):229-240. Botelho, J. R. and P. M. Linardi. 1980. Alguns ectoparasitos de roedores silvestres do municipio de Caratinga, Minas Gerais, Brasil. I. Relacoes pulga/hospedeiro. Rev. Bras. Ent. 24(2):127-130. Bourliere, F. 1983. Species richness in tropical forest vertebrates. In: G. Maury-Lechon, M. Hadley and T. Younes (eds.), The Significance of Species Diversity in Tropical Forest Ecosystems. Report of the meeting of the lUBS Working Group on Species Diversity/Decade of the Tropics Programme, Paris, France.

PAGE 226

218 Brokaw, N. V. L. 1985. Treefalls, Regrowth, and Community structure in tropical forests. In: S. T. A. Pickett and P. S. White (eds.), The Ecology of Natural Disturbance and Patch Dynamics. Academic Press, New York. Brown, J. H. 1978. The theory of insular biogeography and the distribution of boreal birds and mammals. Great Basin Nat. Mem. 2:209-227. Buckley, R. 1982. The habitat-unit model of island biogeography. J. Biogeogr. 9:339-344. 1985. Distinguishing the effects of area and habitat type on island plant species richness by separating floristic elements and substrate types and controlling for island isolation. J. Biogeogr. 12:527-535. Cabrera, A. 1957. Catalogo de los mamiferos de America del Sur. I. Rev. Mus. Cienc. Nat. "Bernardino Rivadavia" 4:1-309. 1961. Catalogo de los mamiferos de America del Sur. II. Rev. Mus. Cien. Nat. "Bernardino Rivadavia" 4:310-732. Carvalho, C. T. 1965. Bionomia de pequenos mamiferos da Boraceia. Rev. Biol. Trop. 13(2):239-257. Case, J. T., M. E. Gilpin, and J. M. Diamond. 1979. Overexploitation, interference competition, and excess density compensation in insular faunas. Amer. Nat. 113:843-854. CETEC, 1981. Programa de pesquisas ecologicas no Parque Estadual do Rio Doce. Sistema Operacional de Ciencia e Tecnologia — SOCT — Fundacao Centre Tecnologico de Minas Gerais — CETEC, Belo Horizonte, M. G., Brazil. Relatorio Final, Segundo Volume. Charles-Dominique, P. 1983. Ecology and social adaptations in didelphid marsupials: comparison with eutherians of similar ecology. In: J. F. Eisenberg and D. G. Kleiman, Advances in the Study of Mammalian Behavior. Special Publication No. 7, American Society of Mammalogists.

PAGE 227

219 M. Atramentowicz, M. Charles-Dominique, H. Gerard, A. Hladik, C. M. Hladik, and M. Prevost. 1981. Les mammi feres frugivores arboricoles nocturnes d'une foret guyanaise: inter-relations plant-animaux. Rev. D'Ecologie, 35:341-435. Cody, M. L. 1974. Competition and the Structure of Bird Communities. Princeton University Press, Princeton, New Jersey. 1975. Towards a theory of continental species diversities: bird distributions over mediterranean habitat gradients. In: M. L. Cody and J. M. Diamond, Ecology and Evolution of Communities. Harvard University Press, Cambridge, Massachussets. Coimbra-Filho, A. F. and R. A. Mittermeier. 1977. Conservation of the Brazilian lion tamarins ( Leontopithecus rosalia ) In: Prince Rainier of Monaco and G. H. Bourne (eds.). Primate Conservation. Academic Press, New York. Conant, F., P. Rogers, M. Baumgardner, C. McKell, R. Dasmann and P. Reining. 1983. Resource Inventory and Baseline Study for Developing Countries. American Association for the Advancement of Science, Washington, D. C. Connell, J. H. 1961. The effects of competition, predation by Thais lapi llus and other factors on natural populations of the barnacle, Chthamalus stellatus Ecol. Monogr. 31:61-104. 1975. Some mechanisms producing structure in natural communities. In: M. L. Cody and J. M. Diamond (eds.). Ecology and Evolution of Communities. Harvard University Press, Cambridge, Massachussets. 1978. Diversity in tropical rain forests and coral reefs. Science 199:1302-1310. and R. 0. Slatyer. 1977. Mechanisms of succession in natural communities and their role in community stability and organization. Am. Nat. 111:1119-1144. Connor, E. and E. McCoy. 1979. The statistics and biology of the species-area relation. Am. Nat. 113:791-833.

PAGE 228

220 Cordero, G. A. and R. A. Nicolas B. 1987. Feeding habits of the opossum ( Didelphis marsupial is ) in Northern Venezuela, Fieldiana, Zoology 39:125-131. Crespo, J. A. 1982. Ecologia de la comunidad de mamiferos del parque nacional Iguazu, Missiones. Rev. Mus. Arg. "Bernardino Rivadavia" 3: (2):45-162. Dalby, P. L. 1975. Biology of Pampa Rodents. Michigan State University Series, Vol. 5, Number 3, East Lansing, Michigan. Darlington, P. J. 1957. Zoogeography: the Geographical Distribution of Animals. Wiley and Sons, London. Davis, D. E. 1946. The annual cycle of plants, mosquitoes, birds and mammals in two Brazilian forests. Ecological Monographs 15:244-295. 1947. Notes on the life histories of some Brazilian mammals. Bol Mus. Nac. 76:1-8. Delany, M. J. 1986. Ecology of small rodents in Africa. Mammal Rev. 16(1): 1-41. Denslow, J. S. 1985. Disturbance-mediated coexistence of species. In: S. T. A. Pickett and P. S. White (eds.). The Ecology of Natural Disturbance and Patch Dynamics. Academic Press, New York. Diamond, J. M. 1975. The island dilemma: lessons of modern biogeographical studies for the design of natural reserves. Biol. Conserv. 7:129-146. Dietz, J. M. 1983. Notes on the natural history of some small mammals in Central Brazil. J. Mamm. 64(3):521-523. Dueser, R. D. and W. C. Brown. 1980. Ecological correlates of insular rodent diversity. Ecology 61:50-56. and J. G. Hallett. 1980. Competition and habitat selection in a forest-floor small mammal fauna. Oikos 35:293-297. and J. H. Porter. 1986. Habitat use by insular small mammals: relative effects of competition and habitat structure. Ecology 67(1) : 195-201

PAGE 229

o 21 and H. H. Shugart. 1978. Microhabitats in a forest-floor small mammal fauna. Ecology 59(1):89-97. Eisenberg, J. F. 1980. The density and biomass of tropical mammals. In: M. Soule and B. Wilcox (eds.), Conservation Biology: An Evolutionary-Ecological Perspective. Sinauer Press, Sunderland, Massachusetts. Eisenberg, J. F. 1981. The Mammalian Radiations. An Analysis of Trends in Evolution, Adaptation, and Behavior. University of Chicago Press, Chicago. M. A. O'Connell, and P. V. August. 1979. Density, productivity, and distribution of mammals in two Venezuelan habitats. In: J. F. Eisenberg (ed.), Vertebrate Ecology in the Northern Neotropics. Smithsonian Institution Press, Washington, D. C. and R, W. Thorington, Jr. 1973. A preliminary analysis of a Neotropical mammal fauna. Biotropica 5(3):150161. Eiten, G. 1974. An outline of the vegetation of South America. Symp. Congr. Int. Primatol. Soc. 5*h, 529-545. Elton, C. 1975. Conservation and the low productivity of invertebrates inside neotropical rainforest. Biol. Cons. 7:3-15. Emlen, J. M. 1973. Ecology: An Evolutionary Approach. Add 1 son-Wesley Publishing Company, Reading, Massachusetts. Emmons, L. H. 1982. Ecology of Proechimys (Rodentia, Echymyidae) in South-Eastern Peru, Trop. Ecol. 23(2): 280-290. 1984. Geographical variation in densities and diversities of non-flying mammals in Amazonia. Biotropica 16(3):210-222. 1987. Comparative feeding ecology of felids in a neotropical rainforest. Behav. Ecol. Sociobiol. 20:271-283. A. Gautier-Hion and G. Dubost. 1983. Community structure of the f rugivorous-fol ivorous forest mammals of Gabon. J. Zool., Lond. 199:209-222.

PAGE 230

222 Enders, R. K. 1935. Mammalian life histories from Barro Colorado Island, Panama. Bull. Mus. Comp. Zool. 78:1-502. Faeth, S. H. 1984. Density compensation in Vertebrates and Invertebrates: A review and an Experiment. In: D. R. Strong, Jr., D. Simberloff, L. G. Abele and A. B. Thistle (eds.), Ecological Communities: Conceptual Issues and the Evidence. Princeton University Press, Princeton, New Jersey. and T. C. Kane. 1978. Urban biogeography. City Parks as islands for Diptera and Coleoptera. Oecologia 32:127-133. Fleming, T. H. 1971. Population ecology of three species of neotropical rodents. Misc. Publ. Mus. Zool. Univ. Michigan 143:1-77. 1972. Aspects of the population dynamics of three species of opossums in the Panama Canal zone. J. Mamm. 53:619-623. 1973. The reproductive cycles of three species of opossums and other mammals in the Panama Canal Zone. J. Mamm. 54:439-455. 1975a. The population ecology of two species of Costa Rican heteromyid rodents. Ecology 55:493-510. 1975b. The role of small mammals in tropical ecosystems. In: F. B. Golley, K. Petrusewicz and L. Ryszkowski (eds.), Small Mammals: Their productivity and Population Dynamics. Cambridge University Press, Cambridge. Fonseca, G. A. B. 1983. The role of deforestation and private reserves in the conservation of the woolly spider monkey ( Brachyteles arachnoides ). Master's Thesis, University of Florida, Gainesville, Florida. 1985a. The vanishing Brazilian Atlantic forest. Biol. Conserv. 34(1): 17-34. 1985b. Observations on the ecology of the muriqui ( Brachyteles arachnoides E. Geoff royi 1806): Implications for its conservation. Prim. Cons. 5:48-52.

PAGE 231

223 L. F. B. Mello and G. Herrmann. 1987. Inventariamento de mamiferos e estudo dos padroes de diversidade de especies de pequenos mamiferos na EPDA de Peti — Minas Gerais. Report to CEMIG, Belo Horizonte, M. G. and K, R. Redford. 1984. The mammals of IBGE's ecological reserve and an analysis of the role of gallery forests in increasing diversity. Rev. Bras. Biol. 44:517-523. Foster, R. B. 1982. The seasonal rhythm of fruitfall on Barro Colorado Island. In: E. G. Leigh, Jr., A. S. Rand, D. M. Windsor (eds.). The Ecology of a Tropical Forest — Seasonal Rhythms and Long-Term Changes. Smithsonian Institution Press, Washington, D. C. Glanz, W, E. 1982. The terrestrial mammal fauna of Barro Colorado Island: Censuses and long-term changes. In: E. G. Leigh, Jr., A. S. Rand, D. M. Windsor (eds.), The Ecology of a Tropical Forest — Seasonal Rhythms and Long-Term Changes. Smithsonian Institution Press, Washington, D. C. Glitzenstein, J. S. P H. Harcombe and D. R. Streng. 1986. Disturbance, succession, and maintenance of species diversity in an east Texas forest. Ecol. Monogr. 56(3):243-258. Hair, J. D. 1982. Measurement of ecological diversity. In: S. D. Schemnitz (ed.). Wildlife Management Techiques Manual. The Wildlife Society, Washington, D. C. Hallett, J. G. 1982. Habitat selection and the community matrix of a desert small-mammal fauna. Ecology 63:1400-1410. M. A. O'Connell and R. L. Honeycutt. 1983. Competition and habitat selection: test of a theory using small mammals. Oikos 40:175-181. and S. L. Pimm. 1979. Direct estimation of competition. Amer. Nat. 113:593-600. Handley, C. 0. Jr. 1976. Mammals of the Smithsonian Venezuela project. Brigham Young Univ. Sci. Bull., 20:1-91, Harper, J. L. 1969. The role of predation in vegetational diversity. In: G. M. Woodwell and H. H. Smith (eds.), Diversity and Stability in Ecological Systems. Brookhaven Symposium in Biology, Brookhaven, Connecticut.

PAGE 232

224 Harris, L. D. 1984. The Fragmented Forest. Island Biogeography Theory and the Preservation of Biotic Diversity. University of Chicago Press, Chicago. Heideman, P. D. L. R. Heaney, R. L. Thomas, and K. R. Erickson. 1987. Patterns of faunal diversity and species abundances of non-volant small mammals on negros island, Philippines. J. Mamm. 68(4):884-888. Hershkovitz, P. 1969. The evolution of mammals on southern continents. VI. The recent mammals of the neotropical region: A zoogeographical and ecological review. Quart. Rev. Biol. 44:1-70. Horn, H. S. 1974. The ecology of secondary succession. Ann. Rev. Ecol. Syst. 1974:25-37. Hueck, K. 1972. As Florestas da America do Sul. Editora Poligono, Sao Paulo, Brazil. Hunsaker, D. 1977. Ecology of New World marsupials. In: D. Hunsaker (ed.), The Biology of Marsupials. Academic Press, New York. Huntly, N. and R. S. Inouye. 1987. Small mammal populations of an old-field chronosequence: successional patterns and associations with vegetation. J. Mamm. 68(4): 739-745. Isabirye-Basuta, G. and J. M. Kasenene. 1987. Small rodent populations in selectively felled and mature tracts of Kibale forest, Uganda. Biotropica 19(3):260-266. James, F. C. and N. 0. Wamer. 1982. Relationships between temperate forest bird communities and vegetation structure. Ecology 63:159-171. Janzen, D. H, 1973. Sweep samples of tropical foliage insects: Effects of seasons, vegetation types, elevation, time of day and insularity. Ecology 54:687-708. and T. W. Schoener. 1968. Differences in insect abundance and diversity between wetter and drier sites during a tropical dry season. Ecology 49:96-110. Johns, A. D. and J. P. Skorupa. 1987. Responses of rain-forest primates to habitat disturbance: a review. Int. J. Primatol. 8:157-191.

PAGE 233

225 Johnson, N. K. 1975. Control of number of bird species on montane islands in the Great Basin. Evolution 29:545-567, Kachigan, S. K. 1982. Multivariate Statistical Analysis. Radius Press, New York. Karr, J. R. 1982. Avian extinction on Barro Colorado Island: A Reassessment. Amer. Nat. 119(2):220-239. and K. E. Freemark. 1985. Disturbance and vertebrates: an integrative perspective. In: S, T. A. Pickett and P. S. White (eds.). The Ecology of Natural Disturbance and Patch Dynamics. Academic Press, New York. Konecny, M. J., in press. Movement patterns and food habits of four sympatric carnivore species in Belize, Central America. In: K. R. Redford (ed.). Mammals of the Americas: Essays in Honor of Ralph M. Wetzel. University of Chicago Press, Chicago. Lacher, T. E. and C. J. R. Alho. in press. Small mammal communities in the Brazilian Pantanal : Population densities, microhabitat affinities and species interactions. J. Mammalogy. M. A. Mares and C. J. R. Alho. in press. The structure of a small mammal community in a central Brazilian savanna. In: K. R. Redford (ed.). Mammals of the Americas: Essays in Honor of Ralph M. Wetzel. University of Chicago Press, Chicago. Laemmert, H. W. L. C. Ferreira and R. M. Taylor. 1946. Investigation of Vertebrate Hosts and Arthropod Vectors. Am. J. Trop. Med. 26(suppl ):23-69. Lee, A. K. and A. Cockburn. 1985. Evolutionary Ecology of Marsupials. Cambridge University Press, London. Leigh, E. G. 1975. Population fluctuations, community stability, and environmental variability. In: M. L. Cody and J. M. Diamond (eds.). Ecology and Evolution of Communities. Harvard University Press, Harvard.

PAGE 234

226 Lidicker, Jr., W. Z. 1975. The role of dispersal in the demography of small mammals. In: F. B. Golley, K. Petrusewicz and L. Ryszkowski (eds.), Small Mammals: Their Productivity and Population Dynamics. Cambridge University Press, Cambridge. Lloyd, M. 1967. "Mean Crowding." J. Anim. Ecol 36:1-30. Loiselle, B. A. and W. G. Hoppes. 1985. Nest predation in insular and mainland lowland rainforest in Panama. Condor 85:93-95. Lomolino, M. V. 1984. Mammalian island biogreography: effects of area, isolation and vagility. Oecologia 61:376-382. 1986. Mammalian community structure on islands: the importance of immigration, extinction and interactive effects. In: L. R. Heaney and B. D. Peterson (eds.). Island Biogeography of Mammals. Academic Press, New York. Lovejoy, T. E., J. M. Rankin, R. 0. Bierregaard, K. S. Brown, L. H. Emmons and M. E. Van der Voort. 1984. Ecosystem decay of Amazon forest fragments. In: M. H. Nitecki (ed.), Extinctions. University of Chicago Press, Chicago. R. 0. Bierregaard, A. B. Rylands, J. R. Malcom, C. E. Quintela, L. H. Harper, K. S. Brown, A. H. Powell, G. V. N. Powell, H. 0. R. Schubart and M. B. Hays. 1986. Edge and other effects of isolation on Amazon Forest fragments. In: M. E. Soule (ed.). Conservation Biology. Sinauer Press, Sunderland, Massachusetts. Ludlow, M. E. and M. E. Sunquist. 1987. Ecology and Behavior of ocelots in Venezuela. Nat. Geog. Res. 3(4):447-461 Lynch, J. F. and D. F. Whigham. 1984. Effects of forest fragmentation on breeding bird communities in Maryland, USA. Biol. Conserv. 28:287-324. MacArthur, R. H. 1957. On the relative abundance of bird species. Proc. Nat. Acad. Sci. 43:293-295. 1972. Geographical Ecology. Princeton University Press, Princeton.

PAGE 235

227 ., J. M. Diamond and J. R. Karr. 1972. Density compensation in island faunas. Ecology 53:330-342. and E. 0. Wilson. 1967. The Theory of Island Biogeography. Princeton University Press, Princeton. Malcolm, J. 1987. Small mammal abundances in isolated and non-isolated primary forest reserves near Manaus, Brasil. In: Resumos do Congresso Brasileiro de Zoologia, Juiz de Fora, Minas Gerais, Brasil. Mares, M. A., M. R. Willig, K. E. Streilein, and T. E. Lacher, Jr. 1981. The mammals of northeastern Brazil: A preliminary assessment. Ann. Carn. Mus. 50:80-137. and Genoways, H. H. 1982. Introduction. In: M. A. Mares and H. H. Genoways (eds.), Mammalian Biology in South America. Spec. Publ. Pymatuning Laboratory of Ecology, University of Pittsburgh. Margules, C. R. and M. B. Usher. 1981. Criteria used in assessing wildlife conservation potential: a review. Biol. Conserv. 21:79-109. May, R. M. 1975. Patterns of species abundance and diversity. In: M. L. Cody and J. M. Diamond (eds.). Ecology and Evolution of Communities. Harvard Univ. Press, Harvard, Massachusetts. M'Closkey, R. T. 1976. Community structure in sympatric rodents. Ecology 57:728-739. Mello-Leitao, C. 1946. As zonas de fauna da America Tropical. Rev. Bras. Geogr. 8:71-118. Melo, D. A. 1977. Observacoes preliminares sobre a ecologia de algumas especies de roedores do Cerrado, municipio de Formosa, Goias, Brasil. Rev. Bras. Pesq. Med. Biol. 10(1): 39-44. Meserve, P. L. 1981. Resource partitioningin a Chilean semi-arid small mammal community. J. Anim. Ecol. 50:745-757. Miles, M. A., A. A. de Souza and M. M. Povoa. 1981. Mammal tracking and nest location in Brazilian forest with an improved spool-and-line device. J. Zoo! Lond. 195:331-347.

PAGE 236

228 Mittermeier, R. A., A. F. Coimbra-Fi Iho, I. D. Constable, A. B. Rylands and C. Valle. 1982. Conservation of primates in the Atlantic forests of Eastern Brazil. Int. Zoo. Yearbook, 1982. Mondolfi, E. 1986. Notes on the biology and status of the small wild cats in Venezuela. In: International Cat Symposium, Caracas, Venezuela. Moojen, J. 1952. Os Roedores do Brasil. Instituto Nacional do Livro, Rio de Janeiro, Brazil. Muller, P. 1973. The Dispersal Centers of Terrestrial Vertebrates in the Neotropical Realm. W. Junk Publishers, The Hague. Murua, R. P. L. Meserve, L. A. Gonzalez and C. Jofre. 1987. The small mammal community of a Chilean temperate rain forest: lack of evidence of competition between dominant species. J. Mammal. 68:729-738. Nitikman, L. Z. and M. A. Mares. 1987. Ecology of small mammals in a gallery forest of Central Brazil. Ann. Carn. Mus. 56:75-95. Nowak, R. M. and J. L. Paradiso. 1983. Walker's Mammals of the World. Johns Hopkins University Press, Baltimore, Maryland. O'Connell, M. A. 1979. Ecology of didelphid marsupials from Northern Venezuela. In: J. F. Eisenberg (ed.), Vertebrate Ecology in the Northern Neotropics. Smithsonian Institution Press, Washington, D. C. Paine, R. T. 1966. Food web complexity and species diversity. Amer. Nat. 100:65-76. Patil, G. P. and C. Taillie. 1983. Diversity as a concent and its measurement. Trans. North Amer. Wildl. Nat. Resour. Conf. 39:334-353. Pearson, 0. P. 1985. Predation. In: R, H. Tamarin (ed.), Biology of New World Microtus American Society of Mammalogists, Special Publication No. 8, Washington, D. C. Peet, R. K. 1974. The measurement of species diversity. Ann. Rev. Ecol. Syst. 1974:285-307.

PAGE 237

229 Phillips, C. J. and J. K. Jones. 1965. Notes on reproduction and development in the four-eyed opossum, Phi lander opossum in Nicaragua. J. Mamm. 50(2):345-348. Pianka, E. R. 1983. Evolutionary Ecology. Harper and Row, Cambridge, Massachusetts. Pielou, E. C. 1974. Population and Community Ecology. Gordon and Breach Publishers, New York. 1975. Ecological Diversity. John Wiley & Sons, New York. Pine, R. W. 1973. Mammals (exclusive of bats) of Belem, Para, Brasil. Acta Amaz. 3:42-79. Poole, R. H. 1974. An Introduction to Quantitative Ecology. McGraw-Hill, New York. Prance, G. T. 1982. Biological Diversification in the Tropics. Columbia University Press, New York. Price, M. V. 1978. The role of microhabitat in structuring desert rodent communities. Ecology 59:910-921. Quinn, J. F. and A. Hastings. 1987. Extinction in subdivided habitats. Conserv. Biol. 1:198-208. Rabinowitz, A. R. and B. G. Nottingham, Jr. 1986. Ecology and behaviour of the jaguar ( Panthera onca) in Belize, Central America. J. Zool. Lond. 210:149-159. Ralls, K. 1976. Mammals in which females are larger than males. Quart. Rev. Biol. 51:245-276. Redford, K. H. 1984. Mammalian myrmecophagy: Foraging, feeding and food preference. Ph. D. Thesis, Harvard University. 1987. Ants and termites as food: Patterns of mammalian myrmecophagy. In: H. H. Genoways (ed.), Current Mammalogy, Volume 1. Plenum Press, New York. Rosenzweig, M. L. and J. Winakur. 1969. Population ecology of desert rodent communities: habitats and environmental complexity. Ecology 50:558-572.

PAGE 238

230 Seifert, R. P. 1975. Clumps of Hel Iconia inflorescences as ecological islands. Ecology 57:1416-1422. Shaffer, M. L. 1981. Minimum population sizes for species conservation. Bioscience 31:131-134. Simberloff, D. S. 1976. Experimental zoogeography of islands: effects of island size. Ecology 57:629-578. 1982. Big advantages of small refuges. Nat. Hist. 91:6-13. and L. G. Abele. 1976. Island biogeograpy theory and conservation practice. Science 191:285-286. 1982. Refuge design and island biogeographic theory: effects of fragmentation. Amer. Nat. 120:41-50. Smythe, N. 1982. The seasonal abundance of night-flying insects in a neotropical forest. In: E. G. Leigh, A. S. Rand, and D. M. Windsor (eds.). The Ecology of a Tropical Forest: Seasonal Rhythms and Long-Term Changes. Smithsonian Institution Press, Washington, D.C. E. Glanz and E. G. Leigh, Jr. 1982. Population regulation in some terrestrial frugivores. In: E. G. Leigh, A. S. Rand, and D. M. Windsor (eds.). The Ecology of a Tropical Forest: Seasonal Rhythms and Long-Term Changes. Smithsonian Institution Press, Washington, D. C. Sokal, R. R. and Rohlf, F. J. 1981. Biometry. W. H. Freeman Co. New York. Soule, M. E. B. 1980. Thresholds for survival: maintaining fitness and evolutionary potential. In: M. E. Soule and B. A. Wilcox (eds.), Conservation Biology. Sinauer Press, Sunderland, Massachusetts. 1983. What do we really know about extinction? In: C. M. Schonewald-Wi Icox, S. M. Chambers, Macbryde and L. Thomas (eds.). Genetics and Conservation. Benjamin Cummings Co., Menlo Park, California. Souza, W. P. 1984. The role of disturbance in natural communities. Ann. Rev. Ecol Syst. 15:353-391.

PAGE 239

231 Streilein, K. R. 1982. Behavior, ecology and distribution of the South American marsupials. In: M. A. Mares and H. H. Genoways (eds.), Mammal in Biology in South America. Special Publication Series, Pymatuning Laboratory of Ecology, University of Pittsburgh Press, Pittsburgh. Sunquist, M. E. S. N. Austad, and F. Sunquist. 1987. Movement patterns and home range in the common opossum ( Didelphis marsupialis ). J. Mamm. 68(1): 173-176. Terborgh, J. W. J. W. Fitzpatrick, and L. Emmons. 1984. Annotated checklist of bird and mammal species of Cocha Cashu biological station, Manu National Park. Fieldiana 21:1-29. Tyndale-Biscoe, C, H. and R. B. Mackenzie. 1976. Reproduction in Didelphis marsupial is and D^ albiventris in Colombia. J. Mamm. 57(2) :249-265. Usher, M. B. 1986. Wildlife Conservation Evaluation. Chapman and Hall New York. Veiga-Borgeaud, T. 1982. Donnees ecologiques sur Oryzomys niqripes (Desmarest, 1819) (Rongeurs ; Cricetides) dans le foyer nature! de peste de Barracao dos Mendes (Etat de Rio de Janeiro, Bresil). Mammalia 46(3): 335-359. Vieira, C. 0. C. 1955. Lista remissiva dos mamiferos do Brasil. Arq. Zool. S. P. 8:341-474. Vuilleumeier, F. 1970. Insular biogeography in continental regions: I. The northern Andes of South America. Amer. Nat. 104:373-388. Walter, H. 1971. Ecology of Tropical and Subtropical Vegetation. Oliver and Boyd Press, Edinburgh. Weaver, M. and M. Kellman. 1981. The effects of forest fragmentation on woodlot tree biotas in Southern Ontario. J. Biogeogr. 8:199-210. Westman, W. E. 1983. Island biogeography: studies on the xeric shrublands of the inner Channel Islands, California. J. Biogeogr. 10:96-118. Williams, C. B. 1964. Patterns in the Balance of Nature. Academic Press, New York.

PAGE 240

232 Willig, M. R. R. D. Owen, and R. L. Colbert. 1986. Assessment of morphometric variation in natural populations: the inadequacy of the univariate approach. Syst. Zool. 35(2):195-203. Willis, E. 0. 1979. The composition of avian communities in remanescent woodlots in southern Brazil. Papeis Avulsos Zool. 33(1):1-25. Wolda, H. 1978. Seasonal fluctuations in rainfall, food and abundance of tropical insects. J. Anim. Ecol. 47:369-381. Wright, P. C. 1985. The costs and benefits of nocturnal ity for Actus trivirqatus (The night monkey). Ph. D. Dissertation, City University of New York, New York. Zimmerman, B. L. and R. 0. Bierregaard. 1986. Relevance of the equilibrium theory of island biogeography and species-area relations to conservation with a case from Amazonia. J. Biogeogr. 13:133-143.

PAGE 241

BIOGRAPHICAL SKETCH Gustavo A. B. da Fonseca was born in Belo Horizonte, Minas Gerais, Brazil, on October 25, 1956. He graduated in ecology at the University of Brasilia, Brazil, in 1978. After graduation, he took the position of mammal specialist at the Brazilian Institute for Geography and Statistics, where he worked until 1981. He arrived at the University of Florida in January, 1982, and received the degree of Master of Arts in Latin American studies in December of 1983. Immediately after he enrolled in the Ph.D. program of the Department of Wildlife and Range Sciences of the University of Florida, receiving his degree in the summer of 1988. He is currently an assistant professor at the Department of Zoology of the Federal University of Minas Gerais, Brazil. 233

PAGE 242

I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, ij%jgcope and quality, as a dissertation for the degree of Doctc Robinson, Chairman /Assoc fate Professor of Forest ces and Conservation I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. ^ J^^ E:. JoKn F. Eisenberg Kathar^ine Ordway Professor of j Ecosystem Conservation I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. n I // l
PAGE 243

I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. •M,'v^^' Larry D. Harris Professor of Forest Resources and Conservation I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Nigel J. H. Smith Professor of Geography This dissertation was submitted to the Graduate Faculty of the School of Forest Resources and Conservation in the College of Agriculture and to the Graduate School and was accepted as partial fulfillment of the requirements for the degree of Doctor of Philosophy. August, 1986 O^U^^<^r ^ < X^%^^^"— ; -' Director, Forest Resources ^nd Conservation Dean, Graduate School

PAGE 244

UNIVERSITY OF FLORIDA 3 1262 08556 7716