Citation
Fruit-frugivore interactions in Euterpe palm forests of the Amazon River floodplain

Material Information

Title:
Fruit-frugivore interactions in Euterpe palm forests of the Amazon River floodplain
Creator:
Moegenburg, Susan Marie
Publication Date:
Language:
English
Physical Description:
ix, 163 leaves : ill. ; 29 cm.

Subjects

Subjects / Keywords:
Birds ( jstor )
Ecology ( jstor )
Environmental conservation ( jstor )
Floodplains ( jstor )
Forests ( jstor )
Fruits ( jstor )
Mammals ( jstor )
Nutrients ( jstor )
Species ( jstor )
Trees ( jstor )
Dissertations, Academic -- Zoology -- UF ( lcsh )
Euterpe -- Harvesting -- Amazon River Region ( lcsh )
Floodplain forest ecology -- Amazon River Region ( lcsh )
Frugivores -- Amazon River Region ( lcsh )
Zoology thesis, Ph.D ( lcsh )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis (Ph.D.)--University of Florida, 2000.
Bibliography:
Includes bibliographical references (leaves 144-162).
General Note:
Printout.
General Note:
Vita.
Statement of Responsibility:
by Susan Marie Moegenburg.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
The University of Florida George A. Smathers Libraries respect the intellectual property rights of others and do not claim any copyright interest in this item. This item may be protected by copyright but is made available here under a claim of fair use (17 U.S.C. §107) for non-profit research and educational purposes. Users of this work have responsibility for determining copyright status prior to reusing, publishing or reproducing this item for purposes other than what is allowed by fair use or other copyright exemptions. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder. The Smathers Libraries would like to learn more about this item and invite individuals or organizations to contact the RDS coordinator (ufdissertations@uflib.ufl.edu) with any additional information they can provide.
Resource Identifier:
025860661 ( ALEPH )
47116857 ( OCLC )

Downloads

This item has the following downloads:


Full Text









FRUIT-FRUGIVORE INTERACTIONS IN EUTERPE PALM FORESTS OF THE
AMAZON RIVER FLOODPLAIN
















By

SUSAN MARIE MOEGENBURG


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


2000




























For Ben and Peter














ACKNOWLEDGMENTS

This study was possible only with the generous support, hard work, and financial assistance of many people and institutions. I am especially grateful to my advisor, Doug Levey, for his scientific advice and enduring support through the years. I have learned much about tropical ecology and conservation from other committee members Jack Putz, Colin Chapman, and Richard Bodmer, been challenged to consider multiple scales by Buzz Holling, and benefited from "brainy" discussions with Katie Sieving. I also thank Lauren Chapman for advice and for serving at my defense.

I am fortunate to have interacted with so many friends and colleagues at the

University of Florida. For lively discussions I especially thank Sophia Balcomb, Kevin Baldwin, Maria Ines Barreto, Paula Cushing, Scot Duncan, John Paul, Roberto Porro, Ltciano Verdade, Dan Wenny, and Amy Zanne. Special thanks also go to Juan Posada for abiding support and love. For help of various sorts I thank Richard Fethiere, Bree Darby, and Mark Stowe.

My life has been enriched tremendously through my experiences in Brazil. I am grateful to Jose Fragoso and Kirsten Silvius for infecting me with Amazon fever. Data collection would not have been possible, or as enjoyable, without the help from Batista Ferreira, Rick Newman, AnaLicia Castelo Branco Pina, the staff of the Estagdo Cientifica Ferreira Penna, especially Pao and Madruga, and the fruit harvesters Cl6sio, Jodo Domingo, and Zeca. I thank Mario Hiraoka for the use of his property along the Rio Maracapacu-Mirim, and Patricia and Mauricio Almeida, Michael and Colleen Collins,








and Beto Verissimo for accommodations in Bel6m. Friends at Imazon, especially Mark Cochrane and Chris Uhl, helped by offering good advice and shaded hammocks at crucial times.

Many people at the Museu Paraense Emilio Goeldi also facilitated this project. I especially thank Mario Jardim and Pedro Lisboa for logistical support at key moments. I am indebted to David Oren for the use of mist nets. My identification skills of birds improved greatly from the expertise of Jos6 Maria Cardoso da Silva and Mario Cohnhaft. I also thank Bill Overal for the use of laboratory space and good music during my aquatic invertebrate identifications, and Mauricio Camargo for help with the fish identifications.

I appreciate the institutional support from the Department of Zoology, University of Florida, and from the Brazilian organizations IBAMA (Brazilian Institute for the Environment), and CNPq (National Institute for Research).

I was able to carry out this research only with generous financial support from the Dickinson family, Tinker Foundation, Sigma Xi Scientific Society, Lincoln Park Zoo, American Bird Conservancy, National Geographic Society, and the U.S. Environmental Protection Agency.

Finally, I express my gratitude to my family and friends, whose support and love I carried with me to the rainforest.















TABLE OF CONTENTS



ACKNOWLEDGMENTS ............................ ................. iii

ABSTRACT .......................................................................... viii

CHAPTERS

1 GENERAL INTROD U CTION .................................................. ............ ............. 1

2 RESPONSES OF VEGETATION, BIRDS, AND MAMMALS TO
MANAGEMENT FOR EUTERPE OLERACEA............................... ............6

Introduction .......................................................... ........................................ . . . . . 6
M ethods ............... ............................................................................ 9
S tu d y S y stem ............... ...................... ............................................................ ..... 9
V egetation and M icroclim ate Sam pling .............................................................. 15
Bird Censusing ................................................. 16
Avian Perch Use and Availability ....................................... 17
M am m al Trapping ....................................................................................... 18
D ata A n aly sis ............. . .......... .. ............................ ........... ................... 19
V egetation and m icroclim ate sam pling .............. ........................ ............. 19
B ird cen su sin g ................. ................................... ............. 19
Avian perch use and capture times.......... ....... .. .............. ..... 22
M am m al trap p ing ............................................................................................ 2 2
R e su lts ............... .... .. .................................................................... ....................... 2 2
Vegetation and Microclimate .................. ...................22
B ird C en su sin g .......... ......................................................................................... 2 7
Avian Perch Use and Capture Times .................................................. ............... 34
M am m al T rap p ing .............................................................................................. 3 7
D iscussion ............................................................................................ ............. 39
C hanges in F lora and F auna ................................................................................. 39
Mechanisms of Bird Community Change ................. . .................. 42
Comparison to Alternative Forest Management Systems ...................................... 44
Recommendations and Conclusions ..................................................................45







V









3 LINKING FRUIT AND FRUGIVORE ABUNDANCE: EXPERIMENTAL
EVIDENCE FROM AMAZONIAN BRAZIL ............ .......................... 47

In tro d u c tio n . .............................................................................................................. 4 7
Methods ................................... ................... 51
Study Species .............................................................................. ............... 51
Study Sites and Plots ............................... ........ ................ 52
F ru it A v ailab ility ................... ... .. ... .. ....................................................... 54
Parrot Use of Palm Forest Versus Non-palm Forest............................... .............56
Frugivore Responses to Experimental Fruit Harvest ................ ................ 57
Frugivore surveys ............... ............. .............. 57
Fruit rem oval .................................................................................................... 59
Statistical Analyses............................................60
R esults .............................................................. .. .. .................... ........................ 6 1
Parrot Use of Palm Forest Versus Non-Palm Forest .............................. .............. 61
Frugivore Responses to Experimental Fruit Harvest ............................................. 63
Fruit availability ................................................................................ . . . . .. 63
Birds: community-level responses ................................... 63
Birds: species-level responses........................................................... .............. 69
M ammals ................................. ....... ............. ..... 72
Bruchid Beetles ....................................................... 74
D isc u ssio n ........ ..................... ..... . .......... .................................................... 7 5
Frugivore Responses to Fruit Harvest................................. 75
Birds - com m unity level ................................................................... ..............75
Birds- species level.............................................78
M a m m a ls ............................................................................................................. 8 1
B ru chid B eetles ............... ...... .. . ............... ... ... . ............ .............. 82
Implications of Fruit Harvest for Frugivores and Fruiting Plants ............................ 83


4 FALLING FRUITS AND FEEDING FISHES: EUTERPE OLERACEA FRUITS IN
AMAZONIAN FLOODPLAIN FOREST NUTRIENT AND FOOD CYCLES ....... 88

In tro d u ctio n ............................................................................................................... 8 8
M e th o d s .................................................................... ......................................... ...... 9 0
Study Sites and Plots .............................................. 90
Fruit Production .................................................................................... ........... 93
N utrient L oss from Fruits ............... .... ..... .. ...... ..... ............. 95
Responses of Aquatic Animals to Fruit Harvest by People ..................................... 96
Statistical A n aly ses ........... ... .. ...... ............................ ...................... ............. 97
R e su lts ...................................................................................................................... 9 8
E. oleracea Fruit Production.......................................... 98
Nutrient Loss from Fruits ............................................ 98
Responses of Aquatic Animals to Fruit Harvest by People 1............................... 100
Discussion .............. ....... 103
D isc ssi n . ... ... ... ... ... ... ... ................................................................................. 1 0 3
Fruit Production, Nutrient content, and Nutrient loss ........................................ 105
Aquatic Animal Responses to Fruit Harvest by People ..................................... 107










5 SPATIAL AND TEMPORAL VARIATION IN HYDROCHORY IN
AM AZONIAN FLOODPLAIN FOREST ............................................................ 111

In tro d u ctio n ..... .............................. ................................................................. 1 1 1
M e th o d s ......................................................... .................................................. . . 1 14
R esults ....................................................................... ....................... . ............. 116
D iscussion .............................................................................................. 121

6 GENERAL CONCLUSIONS.......... . .............. ....................... 131

E cological C onsiderations ................................................................................... 13 1
C onservation C onsiderations......................... ...................................................... 133

APPENDICES

A DIET CLASSIFICATION OF SPECIES OBSERVED IN E. OLERACEA
FOREST .................................................................................... .. 136

B RESULTS OF REPEATED-MEASURES ANOVAS ....................... ........... 139

LIST OF REFERENCES . ........................... 144

B IO G R A PH IC A L SK ET CH ....................................................................................... 163














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

FRUIT-FRUGIVORE INTERACTIONS IN EUTERPE PALM FORESTS OF THE AMAZON RIVER FLOODPLAIN

By

Susan Marie Moegenburg

December 2000


Chair: Douglas J. Levey
Major Department: Zoology

While fruit production in tropical forests varies naturally over space and time, its abundance and distribution in some forests is altered by management to increase fruit production and by harvest. In many tropical areas, forest fruits constitute a substantial portion of human diets and are important market commodities. The effects on forests of fruit management and harvest, however, remain little known. I studied enrichment and harvest of fruit from the palm Euterpe oleracea in palm-dominated (> 300 trees ha-') floodplain forests, estimated to cover over 10,000 km of the Amazon River delta. In this area people harvest up to 9,000 kg of E. oleracea fruit ha- yr', reducing fruit availability to arboreal, terrestrial, and aquatic frugivores, and removing nutrients from the system.

Forests managed for E. oleracea were of lower, more open stature, dominated by E. oleracea, and nearly devoid of understory vines, lianas, and small trees, as compared with control forests. Also, the managed forest bird community was dominated by fruitand seed-eaters, whereas the control forest community contained more insect-eaters. I








also evaluated responses of birds and mammals, seed-feeding beetles, and aquatic invertebrates and fish, to fruit harvest. Four 1.8 ha circular plots were divided into three sections, which were assigned one of three ripe fruit harvest treatments: high (-90% removal), low (-50% removal), and control (0% removal). The community of frugivorous birds and mammals showed significant responses to the high fruit harvest treatment. There were 25% fewer bird species, 58% fewer mammal species, 29% fewer bird individuals, and avian frugivores spent 68% less time in high removal plots compared to control plots. The community of frugivorous birds showed no significant responses, however, to low removal. Also, the abundance of fruit-eating fish, but not invertebrates, was reduced in high-fruit removal treatments. Fruit decomposition experiments demonstrated rapid loss of both nitrogen and phosphorous from submerged fruits. High levels of E. oleracea fruit harvest thus appear to have substantial ecological effects on various trophic levels and ecosystem processes, whereas low levels have few measurable impacts. These results provide an initial ecological basis for determining sustainable levels of fruit harvest.













CHAPTER 1
GENERAL INTRODUCTION

Here and there, shooting above the more dome-like and sobre trees, were the smooth columnar stems of palms, bearing aloft their magnificent crowns of finely-cut fronds. Amongst the latter the slim assai-palm was especially noticable, growing in groups of four and five; its smooth, gently curving stem, twenty to thirty feet high, terminating in a head of feathery
foliage, inexpressibly light and elegant in outline. (Bates 1864, p.3)



Among the most conspicuous features of neotropical rainforests are the high

proportion (70 - 93%) of tree species that produce fleshy fruits, the seeds of which are dispersed by vertebrates (Gentry 1982, Jordano 1992), and the high proportion of vertebrates that include fruit as a major component of their diet (e.g., 35 - 39% of bird species) (Karr et al. 1990). This high diversity of interactions among fruits and frugivores has inspired hypotheses that the abundance and distribution of fruit determine the abundance and distribution of frugivores, and, conversely, that the behavior of fruiteating animals influences the abundance and distribution of fruit producing plants (McKey 1975). Despite a legion of studies addressing these hypotheses, however, the importance of fruits and frugivores to the ecology and evolution of one another, relative to other biotic and aiotic factors, is far from understood (Levey and Benkman 1999).

Increasingly in tropical forests, natural patterns of fruit abundance and

distribution are modified by human activities, including fruit extraction for sale and consumption. In Iquitos, Peru, for example, hundreds of species of fruits and seeds are harvested from wild-occurring trees and regularly sold in markets (Vasquez and Gentry








1989). Not only is fruit extraction from forests a widespread, and sometimes intensive, practice today, but anthropological (Bale6 1988) and archeological (Roosevelt et al. 1996) evidence indicate that, in the Amazon basin, fruit harvesting and forest management have occurred for millenia.

The objective of this work is to evaluate how contemporary fruit management and harvest by humans affect fruit-frugivore interactions in Euterpe oleracea - dominated rainforests of the eastern Brazilian Amazon. This multi-stemmed palm is the source of two of the most economically and socially important non-timber products of the region: heart-of-palm and fruit, locally known as agai (Richards 1993, Anderson 1988). Owing to its economic value, natural abundance, and ease of management (Anderson 1990a,b), E. oleracea is featured in several extractive reserves - areas designated by the Brazilian government for the long-term, sustainable management of both biodiversity and forest products (Fearnside 1989). To achieve these multiple goals, a balance must be struck between levels of management and extraction that maximize profits and those that minimize impacts on biodiversity. An ancillary goal of this project, therefore, is to provide data on which to base management and extraction guidelines to help extractive reserves meet their goals.

Like many Amazonian palms (Kahn 1991), E. oleracea is hydrophilic and rarely grows in non-inundated sites. I worked in two types of floodplain forests dominated by E. oleracea. Data collection for Chapters 2 and 5 took place in sediment-laden whitewater river floodplain forest along the Rio Maracapaci-Mirim, Para State, Brazil. Study sites lie approximately 80 km from the port of Bel6m. Data collection for Chapters








3 and 4 took place in nutrient-poor blackwater river floodplain forest within the Caxiuani National Forest, Para, Brazil.

The high diversity of tree species characteristic of tropical forests (Gentry 1982) means that each species tends to occur at low densities. Low conspecific tree densities limit the amount of fruit, and therefore profits, that can be extracted from a given area of forest. To increase commercial tree densities and profits, people therefore manage forests to augment the production of fruit-producing trees. In the case of E. oleracea, management includes enriching forests through planting and re-locating E. oleracea seeds and seedlings, removing understory plants perceived to compete with E. oleracea, and removing canopy trees that shade E. oleracea and may limit fruit production (Anderson 1990a, b). In Chapter 2, I examine the effects of these management activities on forest structure and bird communities by comparing managed and non-managed forests. In particular, I measure canopy height, canopy density, stem density, basal area, and stem composition, in five managed and five non-managed forests. In addition, I census understory birds with mist-nets in the same ten forests. The differences in the structure and composition of both the vegetation and the bird communities between the two forest types are compared to vegetation and birds in forests subjected to other types of management, such as selective logging and agroforestry.

Forest management, by definition, alters forest structure and composition. Fruit extraction, on the other hand, is frequently viewed as having no ecological impacts (e.g., Peters 1992). This view is not supported, however, by studies demonstrating that fruit and frugivore abundance are linked across space and time (Levey 1988, Loiselle and Blake 1991). These correlative studies, however, have limited power to inform








predictions of how frugivores may respond to fruit harvest. In Chapter 3, I take advantage of the E. oleracea system - high-density production of easily-removed fruit that is eaten by a diverse community of frugivores - to conduct what is apparently the largest experimental manipulation of fruit abundance, to date. In replicated plots I mimic subsistence- and commercial-level fruit extraction and compare the species richness, the number of individuals, and visit lengths of fruit-eating birds and mammals. I also monitor the oviposition behavior of seed-feeding beetles. The results of this experiment not only confirm that fruit abundance drives fruit-eating bird abundance, but also that fruit extraction by humans affects the abundance, distribution, and behavior of fruiteating animals.

While fallen fruits and seeds of trees growing in non-inundated forests become available to terrestrial fruit- and seed-eaters (DeSteven and Putz 1984, Bodmer 1989, 1990, Levey and Byrne 1993, Chapman and Chapman 1996), fallen fruits and seeds in inundated forests also become available to fruit- and seed-eating fish and aquatic invertebrates (Goulding 1980). Fallen fruits and seeds not consumed by these organisms may decompose and release nutrients back into their cycles. In Chapter 4, I investigate these two aquatic aspects of E. oleracea fruit-frugivore interactions. First, to evaluate fruits as nutrient sources in floodplain forests, I determine the nutrient content of fresh E oleracea fruit, the loss of nutrients over time from submerged fruits, and the total amount of N and P in E oleracea fruits per area of forest. Second, to test the effect of human harvest of E. oleracea fruit on aquatic animals, I compare the species richness and abundance of aquatic invertebrates and fish in the same experimental plots used in Chapter 3.








In many fleshy fruit-producing plants, the dispersal of seeds by fruit-eating

vertebrates is only the first of several dispersal events (Wenny 1999). For species that grow in inundated forests, subsequent dispersal of seeds is often accomplished by water. In Chapter 5, I explore the dispersal of seeds by water in tidally-flooded forests, where water levels fluctuate according to daily and lunar cycles. I measure water depth as a function of lunar phase and distance to streams, and quantify the effect of water depth on dispersal of seeds of different species. Based on the results, I argue that water, like other agents of dispersal, is unpredictable over time and space and that this unpredictability selects for multiple mechanisms of seed dispersal within and among fruits of individuals and species.













CHAPTER 2
RESPONSES OF VEGETATION, BIRDS, AND MAMMALS TO MANAGEMENT FOR THE NON-TIMBER FOREST PRODUCT EUTERPE OLERACEA Introduction

Extraction and marketing of non-timber forest products (NTFP) can promote forest conservation and human well-being in several ways (Fearnside 1989, Peters et al. 1989, Panatoyu and Ashton 1992, Peters 1992, Plotkin and Famolare 1992, Arnold and Perez 1998). First, harvest of NTFP, such as fruits, seeds, resins, barks, and fibers, tends to be less destructive to forests than logging. Second, harvest of NTFP provides income for local, forest-dwelling people, furnishing a profitable alternative to forest clearing. Third, the economic benefits of NTFP harvest are realized primarily in the long-, rather than the short-term, term, thereby motivating long-term forest management and planning. Such benefits of NTFP harvest have been embraced by conservationists and form the foundation of extractive reserves, such as those in the Brazilian Amazon (Fearnside 1989, Alegretti 1990, Nepstad and Schwartzman 1992, Mattoso and Fleischflesser 1994, Alves 1995).

One limitation to profitable harvest of many tropical NTFPs is the low density at which source species often occur. Low densities of source plants increase the time required to search for individuals, to travel among individuals, and to transport goods (Peters 1992, Salafsky et al. 1993). To decrease these collecting costs and to increase the economic value of their property, many landowners increase productivity through so-








called enrichment management - "increasing the density of crop-producing species by planting in their natural habitats" (Schulze et al. 1994, p. 582). Such enrichment has apparently been practiced for millenia (Balee 1989, G6mez-Pompa and Kaus 1992, Pinedo-Vasquez and Padoch 1996) and may have created extant "oligarchic forests" (sensu Peters et al. 1989), in which NTFP species occur at high densities. Today, enrichment management is encouraged in various types of forests, including indigenous lands (McCann 1999), extractive reserves (Kainer et al. 1998), privately owned areas (Anderson and loris 1992, Peters 1996), and others (Schulze et al. 1994, Pefia 1996, Ricker et al. 1999).

The effects on forests of NTFP enrichment are distinct from the effects of NTFP harvest, yet have received much less attention from those interested in the development of sustainable NTFP use (Bawa 1992). Enrichment often occurs in secondary forests and often accompanies other management activities, such as logging, liberation treatments, and planting of timber species (Schulze et al. 1994, Salick et al. 1995, Bawa and Seidler 1998). Enrichment management can significantly augment the density of species used by humans. For example, Salick et al. (1995) studied the responses of both timber and NTFP species to logging and liberation (i.e., canopy opening) treatments in primary rainforest in Nicaragua. NTFP and timber species richness and density were higher both one year and nine years post-logging, suggesting that logged forests allowed to regenerate can be enriched for economically-valuable species. This is also the case in the Amazon River floodplain, long inhabited by humans who removed valuable timber trees and enriched for a variety of marketable species (Anderson 1990b, Anderson et al. 1995). Similar systems are known from diverse forests in Asia, Europe, and Latin America








(Pinedo-Vasquez and Padoch 1996, Shankar et al. 1996, Freese 1997, McShane and McShane-Caluzi 1997, Wollenberg and Ingles 1998). While these studies illustrate that enrichment management increases the density of harvested species, less is known about the capacity of NTFP-enriched forests to conserve biological diversity and ecosystem functions. The viability of NTFP enrichment and harvest as a conservation tool depends not only upon sustaining production of forest products, but also upon sustaining biodiversity in enriched forests.

I evaluate the ecological impacts of enriching Amazonian floodplain forests with Euterpe oleracea, known locally as agai. This native palm is the source of two important NTFPs: palm heart and fruit. The E. oleracea system represents a complex interaction among palm trees, humans, and animal communities. Enrichment with E. oleracea likely alters forest structure, similar to management for timber species; however, it also creates high density stands of a fruit resource used by a suite of frugivorous animals. I focus my comparisons ofE. oleracea-enriched and control forests on vegetation structure and composition, and understory bird and mammal communities. Birds and mammals are known to respond to variation in vegetation structure and associated microclimatic fluctuations across natural environmental gradients and human disturbances (MacArthur and MacArthur 1961, Karr and Freemark 1983, Johns 1988, Thiollay 1992, Mason 1996. Greenberg et al. 1997, Bawa and Seidler 1998, Putz et al. 2000).

I chose to study the E. oleracea system for several reasons. In contrast to upland areas in the Amazon basin, where conversion of forest to pasture or agriculture captures the highest short-term profits, management of this floodplain for forest products is considered the most rational and profitable land-use option (Anderson 1990b, Anderson








and loris 1992, Anderson et al. 1995, Hiraoka 1995). Indeed, demand for floodplain forest products, especially palm fruit and palm heart, is high and increasing (Pollack et al. 1995). Approximately 40% of the 25,000 km2 Amazon estuary is enriched primarily for E oleracea production and several areas within Brazil's 22,000 km2 system of extractive reserves feature E. oleracea as an NTFP on which residents' incomes are based (Mattoso and Fleischflesser 1994, Alves 1995).



Methods

Study System

The range of the pinnately leaved, multi-stemmed E. oleracea includes the

Brazilian States of Para, AmapA, Tocantins, and Maranhdo, the Pacific coast of Colombia, and northern Ecuador, Trinidad, Venezuela, and the Guianas (Henderson 1995). Throughout the Amazon River estuary, E. oleracea, which I also call by the colloquial name "agai", grows in monodominant stands in floodplain forests, known as virzea and igapo. Individual trees contain up to 25 stems (Henderson 1995), which become reproductive at heights of 4 - 30 m, depending on growing conditions (Strudwick and Sobel 1988). E oleracea produces 1 cm diameter fleshy-fruited drupes (Roosmalen 1985), which are consumed by many species of frugivores (Chapter 3).

E. oleracea is among the most widely-used forest species in the Amazon River

estuary, where its products are central in the diet, culture, and economy of the inhabitants (Bates 1864, Anderson 1988, 1990b, Strudwick and Sobel 1988, Anderson et al. 1995, Hiraoka 1995, Muiiz-Mirit et al. 1996). From E. oleracea people harvest fruits, which are produced into a popular beverage, and the apical meristem, from which heart-of-palm








is extracted. E oleracea fruits have been harvested for centuries (Bates 1864), but largescale commercial harvest of E oleracea heart-of-palm began relatively recently, and coincided with overharvest of populations of E. edulis in southern Brazil (Hiraoka 1995). During the late 1960s, as populations of E. edulis declined to the point of low returns, palm heart factories moved from the Atlantic forest region to the Amazon estuary. At the same time, the market for sugar cane, grown for decades by the estuary's rural inhabitants, collapsed. Rural workers turned to extracting palmito from wild, highdensity agai stands and managing their own forests for agai production. This process was fueled by the growing number of palmito processing factories in the region and by the expanding market for agai fruit, particularly in Belem. Thus began what Hiraoka (1995) has termed the "agaization" of the estuary, in which management that favored E. oleracea became the dominant land use strategy.

Selling heart-of-palm and fruit from E. oleracea are profitable ventures. On the 15 km'2 Combu Island, near Bel6m, for example, E oleracea palm heart and fruit sales comprise 85% (US$3,400) of the income for the majority of the island's 97 households (Anderson and loris 1992). Families selling agai fruit can earn up to US$235 per hectare per year (Peters 1992). During the months of low agai fruit production (November March), income is supplemented by selling palm hearts. Throughout the region, palm heart sales generate approximately US$300 million per year and employ nearly 30,000 people (Clay 1997). The low suitability of estuarine floodplain for other uses and the relative profitability of agai products make agai management and harvest the preferred land use in the region (Hiraoka 1995).








Agai-enriched forest stands originate in one of two ways. Plots formerly devoted to swidden agriculture or sugar cane can be transformed to agai agroforests, after a fallow period of 5 - 7 years (Hiraoka 1995), or mature mixed forest can be gradually converted to agai-dominated forest. Both processes involve planting of agai seeds and seedlings, protection and thinning of agai plants, and removal of non-harvested plants that might compete for light and nutrients (Anderson and loris 1992). In addition, plants that impede walking through the forest, or in which snakes are thought to hide, are removed. According to Hiraoka (1995), agai production reaches peak levels 6 - 10 years after this "tolerant" (sensu Anderson 1990b) management begins. Acai-enriched forests contain other native and exotic harvested species such as tapereba (Spondias mombin), rubber (Hevea brasiliensis), andiroba (Carapa guianensis), bacuri (Rheedia macrophylla), cacao (Theobroma cacao), coconut (Cocos nucifera), guava (Psidium guajava), and mango (Mangifera indica) (Anderson 1990b, Anderson et al. 1995).

Management for agai and other land uses results in a mosaic of forest types that includes "home gardens", agai-dominated forest, and mature, non agai-enriched forest (hereafter "control") from which other forest products are gathered. Streams often form the boundaries between enriched and control forests, and forests dedicated to agai management are easily distinguished from control forests. For example, in enriched forests agai genets almost always contain remnant stumps of stems harvested for heartof-palm. Another clue is discarded, harvested infructescences from which fruits have been removed. Tree species dominant in control forests include Mauritiaflexousa, Virola surinamensis, Pterocarpus officinalis, and Rhizophora mangle (Anderson et al. 1995).








This research took place during 1997 in the southeastern portion of the Amazon estuary, Para State, Brazil (1045'56" S, 48'57'41" W; Fig. 2-1). The hundreds of islands comprising this region are characterized by extremely low-elevation, low-pH Entisol soils, low plant species diversity, and average temperature and precipitation of approximately 25 C and 3000 mm, respectively (Hiraoka 1995). Rainfall is seasonal, with highs from January to May, and lows from June to November.





N





BRAZIL







0Belem



;',@Abaetetuba ,





7'30 Kilometers

- Study islands


Figure 2-1. Study site within Brazil, and study islands within site.








Study sites were located in floodplain forest near the mouth of the Tocantins River, which reverses its flow twice per day with tidal influx. During the full and new moons, and more regularly during the rainy season, the tidal influx swells the river enough to overflow its banks and flood the forests. The restriction of agricultural possibilities by these floods has led to management for agai production as the primary land use in the area. Study sites lie within 20 km of the nearest city, Abaetetuba, and within 80 km of the capitol city Belem (pop. -169,807,000). Most agai fruit and palm heart extracted from this area are sold in Abaetetuba, with occasional shipments selling in the Belem market. Agai fruit is the primary commodity sold by the rural people, with occasional sales by most producers of palm heart. Production ranges from 8,640 - 13,734 kg fruit/ha/year, which totals approximately US$6,594 - 14,936 per ha (Mufiiz-Miret et al. 1996).

Data were collected in ten forest stands: five enriched primarily for agai tree production through enrichment (hereafter "enriched"), and five not enriched for agai (hereafter "control"). These forests were identified through conversations with local people. Enriched forests had been so managed for > 20 yr and were approximately 4-5 ha in size. The sizes of the control forests ranged from 10 - 100 ha. Although some of the enriched and control stands were adjacent to each other (Fig. 2-2), the locations at which data were collected within the forests were all > 500 m apart. Because agaienriched stands tend to be close to homes, which tend to be close to rivers, most enriched stands were closer to rivers than were control stands. Aside from differing in management activities, the ten forests were assumed similar in soils, topography, and vegetation because of their proximity to one another in the floodplain. Other studies have shown



















CPO




0 '






0










CL








high similarity in forest composition among floodplain forest sites in this region (Anderson et al. 1995). Control forests have long been subject to subtle management, in which people selectively remove trees for timber and remove other products, such as palm leaves for thatch, bark for medicines, and fruits for food. Nevertheless, they differ from agai-enriched forests by not being managed specifically for agai production. Vegetation and Microclimate Sampling

Vegetation structure and composition were evaluated in each of the 10 stands along a 50 x 1 m transect. Finding homogenous patches in the varzea, which is riddled with streams and permanently inundated low-lying areas, and where management activities themselves are patchy, proved difficult. Fifty meters was chosen as the minimum transect length that allowed a characterization of the vegetation while remaining within a homogeneous stand of either enriched or control forest. Within such stands, transect starting points and orientations were located randomly.

Along each transect, data were collected on vegetation structure and composition that could be directly affected by management activities and that might affect vertebrate communities. I collected two types of data at 5 m intervals, for a total of ten points per transect. Canopy cover above 1.5 m was measured using a spherical densiometer, an instrument for measuring forest overstory density (Lemmon 1957). Canopy height was visually estimated after practicing with a clinometer at known distances from trees. Along the entire length of each transect, I counted and measured the dbh (diameter at breast height) of all stems above 1.5 m. Stems were recorded as belonging to one of the following categories: E. oleracea; Mauritiaflexuosa (Palmae); Raphia taedigera (Palmae); Montrichardia linifera (Araceae); non-woody vine; woody liana; or hardwood stem. The








palms and M. linmifera were counted separately from the other types of stems because they lack woody bark and may therefore represent different foraging substrates for vertebrates than do vines, lianas, or hardwood trees. Hardwood stems include trees, treelets, and shrubs.

Differences in vegetation structure between forest types can lead to differences in microclimate. In turn, microclimatic differences may affect forest use by vertebrates. To test for microclimatic differences in my forests, I recorded temperature and relative humidity during the first day of mist netting (see below) in each of the 10 forests at 0600, 1200, and 1800 hrs. In each of the ten forests all three measurements were taken in the same location, which was always shaded understory. Measurements were taken with a thermometer and a hygrometer.

Bird Censusing

I censused the avian communities in the same ten agai-enriched and control stands using mist nets during September - November 1997, which coincided with the agai fruiting season. These data were supplemented with casual observations of birds seen or heard during mist netting or vegetation sampling. Although the use of nets introduces biases associated with differences in net visibility among habitats being sampled, and differences among species in their probability of being captured (Karr 1981, Remsen and Good 1996), I chose to use them for several reasons. Specifically, nets allow the detection of species that are more secretive or nocturnal, do not vocalize consistently (e.g., some hummingbirds and cotingas), or that are not vocalizing during the period of sampling. In addition, mist nets were used to sample birds in the only other comprehensive studies of








birds in floodplain forest in this region (Novaes 1970, Lovejoy 1974), so their use in this study allows a direct comparison to other studies.

In each stand, 12 nets (36 mm mesh, 12 x 2.6 m) were placed in three groups of

four, each group comprising a net "lane" of approximately 50 m. Lanes within each stand were 50 - 100 m apart. An attempt was made to place the lanes parallel to each other, but this was not always possible owing to the interruption of vegetation cover by streams and trails, which were avoided. Nets were opened at 05:45 (dawn) on two consecutive days at each site. They were left open for 12 hours on the first day and 6 hours on the second (216 net-hours per stand). To avoid a seasonal bias, I alternated between sampling enriched stands and control stands.

Nets were checked at least hourly. All birds were removed, placed in a cotton bag, and carried to a central station for identification and measuring before being released at the point of capture. The first time each bird was captured I clipped a feather on the left wing in order to identify re-captures, which were not included in analyses. Avian Perch Use and Availability

Availability of perches can be an important mechanism driving differences among bird communities in areas with differences in vegetation, because foraging birds tend to select perches of a given size (Kendeigh 1945, Morse 1976, Robinson and Holmes 1982). The availability of perches of certain dimensions may be altered by management, which removes some types and sizes of trees, vines, lianas, and other understory vegetation (hereafter "stems"). Selection for perch sizes by birds can be tested by comparing the size distribution of perches used by birds with that expected if birds were selecting perches based on the availability of stems of different sizes in the forest. To do this, I determined








the perch sizes and types used by four antbird (Formicariidae) species and compared them with stem sizes and types available in the five enriched and five control forest stands. The four focal species were chosen because they were commonly seen and forage almost exclusively in the understory, in contrast to other species that utilize both understory and canopy strata and would therefore be difficult to observe. During the same period as mist-netting, I observed Thamnophilus nigrocinereous, T. punctatus, Cercomacra tyrannia, andMyrmotherula axillaris, and recorded the diameters of perches they used while foraging. I observed birds opportunistically as I encountered them, following individuals as long as possible. At least five individuals of each species were followed, but some individuals were followed multiple times. I determined the availability of stems that birds could use as perches along the transects I used to sample vegetation. I measured the diameters of all perch-like stems (including vines, liamas, tree branches, palm leaf petioles) that touches my outstretched arms as I walked along the transect.

Mammal Trapping

As with birds, mammals were censused to evaluate their use of agai-enriched versus control forest. Mammals were trapped with Sherman (6.6 x 7.7 x 19.8 cm) and Tomahawk (52.8 x 15.4 x 15.4 cm) live traps during December 1996 - February 1997. Because this was a period of low E oleracea fruit production, differences in mammal captures between enriched and control forests would represent responses of mammals to differences in factors other than fruit availability (e.g., vegetation structure, microclimate, nest sites). Traps were set at 10 m intervals on a grid system in two, one-hectare plots one in an agai-enriched stand and one in a control stand. The number and position of








traps fluctuated throughout the study, but each forest typically contained 20 small and 4 large traps. Traps were moved periodically from ground level to 4 m, in response to changing flood levels and in order to sample both terrestrial and understory habitats. Understory traps were tied onto tree limbs or, occasionally, placed on platforms that were hung next to trees. Traps were never opened for more than four consecutive nights. Trap effort totaled 521 small and 172 large trap-nights in control forest, and 509 small and 151 large trap-nights in agai-enriched forest.

Depending on availability, bait included bananas, papaya, pineapples, and corn, but in any given night the same bait was used in all traps. I used fruit for several reasons. First, most small, non-volant neotropical mammals include fruit in their omnivorous diets (Emmons 1997). Second, fruit is often used as bait in studies of non-volant neotropical mammals (e.g., Adler 1998). Furthermore, fruit was consistently available to me during the trapping period. Bait was placed in the traps in the late afternoon and traps were checked the following morning. Animals were removed from the traps, identified, measured, marked and released at the trap site. Henna hair dye was used to place a unique mark on each animal's flank, which aided in identifying re-captures. Data Analysis

Vegetation and microclimate sampling

Canopy height and canopy density were compared between enriched and control forests with nested ANOVAs on transformed data. Total basal area was compared with a t-test. Because understory vertebrates might be sensitive not only to the total density of stems but also to the types of stems available, I separately compared the densities and dbhs of palms, vines, lianas, hardwoods, and M linifera with t-tests. To assess the








availability of small perches to vertebrates, I compared the distributions of small (0-5 cm dbh) tree stems, vines, and woody lianas between the two forest types with a chi-square test of homogeneity. Finally, temperature and relative humidity were compared with Wilcoxon Rank Sums tests.

Bird censusing

Bird communities in the two forest types were compared with three methods commonly used in community comparisons. First, to quantitatively describe the communities, I calculated the total number of captures, species richness, and species diversity for all birds captured in nets, and species richness for all birds captured and observed. Species diversity was estimated with Simpson's Index:



1
D
$,p i2
i=1


where pi = proportion of species i, and s= the number of species.

The second comparison of the bird communities evaluated species composition using Dice's Index (1945). This index, which estimates the similarity of two communities, is calculated as:

2a
(2a + b + c)'


where a is the number of species common to both habitats, and b and c are the number unique to the two habitats. A value of I indicates complete species overlap. I used this index to answer two questions. First, how similar is the bird community in enriched








forest to that in control forest? I pooled all species caught in agai- enriched forest and pooled all those caught in control forest, then calculated the index. Second, I asked the question: how similar are enriched forests to one another, and how similar are control forests to one another? To answer this, the index was calculated for all pairs of sites within each forest type, which resulted in 10 index values for agai-enriched forest and 10 for non- enriched forest. I used a Wilcoxon Rank Sums test on these values to evaluate which forest type was most similar among sites.

Responses of bird species to forest management may differ among members of different dietary guilds and among those that utilize different strata of the forest 1988, Thiollay 1992, Mason 1996, Greenberg et al. 1997). In particular, understory insectivores tend to decline dramatically in enriched forests. I tested the hypothesis that bird species would differentiate between enriched and control forest based on diet and forest strata preferences. I used correspondence analysis, which is an ordination procedure useful for testing hypotheses about joint relationships between variables with categorical data (James and McCulloch 1990). I first assigned species to guilds (using Hilty and Brown 1986, Levey and Stiles 1992, Ridgely and Tudor 1994, Mason 1996, and Greenberg et al. 1997, and personal observations of food consumption) and strata preferences (using Parker et al. 1996). Some species, generally considered omnivorous, were classified as granivorous or frugivorous if they were observed primarily eating seeds or fruit during this study. Correspondence analysis (SAS 1996) was based on the number of captures of each species in the two forest types.








Avian perch use and capture times

I tested if the four antbirds that I observed showed preferences for perches of

certain sizes, and if the abundance of those perch sizes differed between forest types. To test for perch size preferences, I compared the size distribution of perches used by birds with the distribution that would be expected if birds were simply using stems according to their availability with a chi-square test.

Use of a forest type might also depend upon favorable microclimatic conditions. I predicted that, if temperature and relative humidity influence the activity of birds, then this may be reflected in the time of day at which I captured them. Therefore, for all species that were captured in both forest types (= 15 species), I compared capture times with a one-way ANOVA.

Mammal trapping

Capture rates of mammals were standardized for the number of trap-nights (= number of traps x number of nights open), which was higher in control than in agaienriched forest. To compare trap success between stands, I pooled all individuals of the three species that I captured. Captures were dominated, however, by one species (Marmosa murina), so it alone was used to compare the number of individuals between forest types using a chi-square test.



Results

Vegetation and Microclimate

Vegetation structure differed substantially between agai-enriched and nonenriched stands. Forest canopy averaged 16.2 m in control stands and 10.2 m in enriched









stands (F,90 = 111.50, p < 0.01). Canopy density was 86.6% in control stands, but only 77.4% in enriched stands (F,.0 = 96.70, p < 0.01). Stem density (t = 3.91, df= 8, p <

0.05) and basal area (t = 3.72, df= 8, p < 0.01) were similarly greater in the control stands (density: 4680 + 1197; area: 16.4 + 2.0) than in the enriched stands (density: 1640 + 1267; area: 3.4 + 2.9).

Agai- enriched and control stands also differed in vegetation composition (Fig. 23). Agai- enriched stands contained six times the number of reproductive-sized acai adults (t = 3.86, df = 8, p < 0.01), eleven times the number of juveniles with trunks (t = 2.97, df = 8, p < 0.02), and seven times as many juveniles without trunks (t = 3.96, df = 8, p <

0.01) than did control stands. On the other hand, control stands contained four times more small trees (< 10 cm dbh; t = 3.84, df= 8, p < 0.01), five times more non-woody vines (t = 2.95, df= 8, p < 0.02), and 84 times more woody lianas (t = 4.51, df= 8, p <

0.01). Stems of agai trees were larger in enriched stands, while vines had greater diameters in control stands (Fig. 2-4).

These differences, particularly the virtual lack of vines and lianas in the agaienriched stands, led to significantly different distributions of small-diameter trees, vines, and lianas (together "stems") in the two forest types (Figs. 2-5a & b; X2 = 20.27, df = 9, p < 0.001). While most small stems in the control stands measured less than 1 cm dbh, stems were found in all size classes between 0 - 5 cm (Fig. 2-5a). In contrast, most small stems in agai-enriched stands were either 0 - 1 cm or 4 - 5 cm, with relatively few stems at all, and no vines or lianas, between 2 - 4 cm (Fig. 2-5b).













80 enriched a8000+1 7000- 0 non-enriched
a7000E 6000 5000c. 4000S 30002
a 20001000
0 in











Stem types



Figure 2-3. Numbers of stems of different types (mean + s.e.) in five stands enriched with E. oleracea and five non-enriched stands. Bars show standard error and asterisks indicate significant (p < 0.05) differences between stand types.









* enriched

0 non-enriched


Stem types




Figure 2-4. Diameters (dbh) of different types of vegetation (mean + s.e.) in five forest sites managed for E. oleracea and five non-managed sites. Bars show standard error and asterisks indicate significant (p<0.05) differences between forest types.


11- * 109
8
7
6
5
4
3 2
1
0 O-)
0











a.

I] lianas O vines a trees


0.0 . . . .
C) W) C W) 0 ) 0 W


0.0 0
0 ,0 9) . V9 t 9 '. 9
o i_' Il CI 4 4 V
W W r 1 W
-~~ - l Cl et


Diameter (cm)




Figure 2-5. Distribution of small (0-5 cm) trees, vines, and lianas among size classes in two different stand types.
a) E. oleracea-enriched forest
b) Non-enriched forest








Measurements of temperature and relative humidity showed that microclimate fluctuated more throughout the day in enriched stands than in control stands. Dawn and dusk temperatures were similar in the two forest types, but noon temperature was 33.7 C in the enriched stands, but only 31.2C in the control stands (Table 2-1). Relative humidity did not differ significantly between the two forest types. Table 2-1. Temperature (C) and relative humidity (%) in agai-enriched and control stands. Significant difference at p < 0.05 is indicated by *.

Agai-enriched Control
Dawn temperature 25.6 26.5 Dawn relative humidity 88.5 86.5
Noon temperature 33.7 31.2 Noon relative humidity 69.25 75.5 Dusk temperature 28.9 29.3 Dusk relative humidity 82 81.5 Bird Censusing

I captured 54 bird species in nets and observed an additional 16, for a total of 70 species among all ten sites (Table 2-2). Species accumulation curves indicate that most species were detected after the first 15 of the 18 hrs of netting; few new species accumulated in the final three hours (Fig. 2-6). Although 70 species is a relatively low species richness for an Amazonian site (Terborgh et al. 1990, Thiollay 1994), it is not especially low for floodplain forests in eastern Amazonia, which are known to have lower species diversity than interfluvial and western Amazonian sites (Lovejoy 1974). Two earlier studies (Novaes 1970, sampling for 12 months; Lovejoy 1974, sampling at two sites for 22 months) in similar habitat found between 62 and 87 bird species, indicating that my results fall within the expected species richness range for the region.









Table 2-2. Birds captured and observed in agaf-enriched (1-5) and control (6-10) stands. Numbers correspond to individuals captured, x's represent birds observed but not captured. Of the species captured in nets, those found only in aqaf-enriched stands and not in control stands are in bold. The second and third columns represent each species' dietary guild and preferred forest strata. The final two columns refer to the number of agaf-enriched and non-enriched stands, respectively, in which each species was detected.


ArM-enriched ~t~ind~


Control stands


species diet strata 1 2 3 4 5 6 7 8 9 10 # 1-5 # 6-10 Buteo magnirostris c c x 0 1 Micrastur gilvicollis c u/m x x 1 1 Columbina talpacoti o t 1 1 0 Leptotila rufaxilla g t 1 1 1 1 1 4 1 Amazona amazonicus f c x x 1 1 Brotogeris versicolurus f c x 0 1 Crotophaga major o t/c 1 1 0 Amazilia sp. n 1 1 0 Campylopterus largipennis n u/m 1 0 1 Glaucis hirsuta n u 1 1 0 Threnetes leucurus n u 1 1 0 2 Phaethornis ruber n u x 1 x 0 3 Phaethornis sp. n 1 1 0 Thalurania sp. n 1 0 1 Thaluraniafi4rcata n u/m 2 2 1 1 1 4 1 4 3 Trogon viridis o c x 1 0 Trogon collaris o m/c x 1 0 Chloroceryle americana p u - 2 1 x 2 1









Table 2-2 continued
Chloroceryle inda p u 1 1 1 2 2 2 Chloroceryle aenea p u 2 1 1 0 3 Celeus flavus o c 1 x 1 1 Ramphastos tucanus f c x x 1 1 Xiphorhynchus picus i m 1 2 2 1 1 3 2 Xiphorhynchus guttatus i u/c 0 1 Glyphorhyncus spirurus i u/m 4 11 10 5 3 0 5 Nasica longirostrus i m/c x 0 1 Xenops minutus i u/m 2 1 1 1 1 3 1 5 Cercomacra tyrannia i u x 2 1 1 2 x x 6 4 4 Formicivora grisea i u/m 1 3 2 4 x 4 1 Hypocnemoides melanopogon i u x 1 1 1 1 3 Microhopias quixensis i U1/t 2 - 1 0 Myrmotherula axillaris i u/m 3 4 4 11 5 4 10 3 3 5 Sclateria naevia i 1/t x 1 2 1 1 1 3 x 4 4 Thamnophilus nigrocinereous i u/m 1 4 5 6 4 6 5 2 5 4 5 Thamnophilus punctatus i u/m 1 0 1 Thamnophilus amazonicus i u/m x x 1 1 Pipra aureola f u/m 1 6 2 6 3 6 7 1 5 4 5 Manacus manacus f u 1 1 1 1 2 Querula purpurata f c x 1 0









Table 2-2 continued
Pachyramphus castaneus o c 1 1 0 Pachyramphus polychopterus o c 1 1 0 Eleania chiriquensis o c 2 1 0 Myiopagis gaimardii o c 2 1 0 Mionectes oleaginea f u/c 2 1 4 2 1 Lophotriccus galeatus i m/c 1 0 1 Todyrostrum maculatum i c 2 1 2 0 Tolmomyiasflaviventris o c 1 1 0 Tyrannus melancholicus o c x 1 0 Myiozetes similis o m/c 1 1 0 Phaeomyias murina o c 1 1 0 Myiozetetes cayanensis o c x 2 2 0 Tyrannulus elatus i c x 1 0 Turdusfumigatus o t/m 3 1 2 0 Turdus albicollis o u/rm 3 1 0 Basileuterus rivularis i t/u 2 1 1 1 0 4 Cacicus cela o m/c x 1 0 Vireo olivaceus o c x x x x x x 2 4 Geothlypis aequinoctialis o u 1 1 0 Cyanerpes caeruleus o c x 1 0 Coerebaflaveola o c 1 1 2 0 Thraupis episcopus f c x x x x 4 0









Table 2-2 continued
Thraupis palmarum f c x x 1 1 Ramphocelus carbo f u/c 1 4 32 22 1 1 1 1 5 3 Eucometis penicillata f u/m x 1 3 2 2 2 2 2 3 5 Tachyphonus rufus f u/c 1 1 0 Oryzoborous angolensis g u/m 3 1 0 Sporophila americana g u/m 1 7 2 0 Sporophila nigricollis g u 1 1 0 Saltator maximus m/c 1 1 0 Saltator orenocensis g u/c 1 1 0
diet: c=carnivore; p=piscivore; i=insectivore; o=omnivore; f=frugivore; g=granivore; n=nectarivore strata: c=canopy; m=midstory; u=understory; l=low understory; t=terrestrial











301

* Im
15


to.
- 10


5.


0

1 2 3 4 5 6 7 8 9 101112131415161718

Hours netting




Figure 2-6. Bird species accumulation curves during 216 mistnet-hours in ten stands.
Squares represent E. oleracea-enriched stands, while x's represent non-enriched
(control) stands.






Species richness was nearly twice as high in enriched stands than in control stands

when only netted birds were considered (43 vs 25 species; X2 = 4.76, df = 1, p < 0.05) and when both netted and observed birds were considered (57 vs. 36 species; X2 = 4.74,

df = 1, p < 0.05). I found no differences between enriched and control stands in the

number of individuals captured (194 vs 180; X2 = 0.52, df= 1, p > 0.50) nor in species

diversity (Simpson's Index: 8.2 vs 9.4; Mann-Whitney U test, U = 14, p > 0.10).








The difference in species richness between the two forest types reflects their

nearly completely different sets of species. Of the 70 species observed overall, only 24 were common to both forest types, and Dice's Index of Similarity of the two forest types was low (0.5567; 1.0 indicates complete overlap). Values of Dice's Index were also significantly (Mann-Whitney U test, U = 11.063, p < 0.01) greater among the nonenriched forests (0.6004 + 0.080) than among the enriched forests (0.3966 + 0.123). This higher degree of similarity among control sites suggests a core set of species typical of control forest. Indeed, six species (Glyphorhynchus spirurus, Xenops minutus, Myrmotherula axillaris, Thamnophilus nigrocinereous, Pipra aureola, Eucometis penicillata) were caught in every control forest. In contrast, only one species, the silverbeaked tanager (Ramphocelus carbo), was caught in every agai-enriched forest.

Moreover, some species were observed in only one of the two forest types. For example, I observed nine species in the control sites only, and these tended to be smallbodied, nectarivorous, piscivorous, or insectivorous species that inhabit forest understory [four hummingbirds (Campylopterus largipennis, Phaethornis ruber, Thalurania sp., Threnetes leucurus), pygmy kingfisher (Chloroceryle aenea), black-chinned antbird (Hypocnemoides melanopogon), slaty antshrike (Thamnophilus punctatus), scale-crested pygmy-tyrant (Lophotriccus pileatus), river warbler (Basileuterus rivularis)]. Of the species observed but not caught, the raptor Buteo magnirostris and the parakeet Brotogeris versicolurus were seen only in control forest. On the other hand, 32 species were observed in agai-enriched forest only, and 20 of these were represented in only one of the sites (#4). These species tended to be larger-bodied frugivores or omnivores, many of which occupy the forest canopy (Table 2-2).








Correspondence analysis revealed significant relationships among diet, forest

strata preference, and forest type preference (X2 = 170.24, df = 105, p < 0.001; Fig 2-7). Most notably, insectivores were associated exclusively with the understory of control stands, while frugivores were associated with the understory and midstory of both enriched and control stands. Omnivores and piscivores, on the other hand, were associated almost exclusively with enriched stands; omnivores in the canopy, and piscivores in the midstory and understory. Nectarivores and granivores showed few preferences, being caught in all strata of both stand types.

In addition to comparing the species composition of the bird communities, I also compared capture rates of bird species, as an indication of forest type preferences. Most of the frugivorous, granivorous and omnivorous species were caught with greater frequency in the agai-enriched forest, whereas most of the non-frugivorous species (insectivores, nectarivores, and piscivores) were caught more in the control forest (Fig. 28, X = 6.37, df= 1, p < 0.01).

Avian Perch Use and Capture Times

Numerous mechanisms, including understory perch availability and microclimate, may underlie the observed differences in avian communities. The four antbird species all used perches between 0.5 and 3.0 cm, although they differed in their perch use according to body size (Fig. 2-9). The smallest of the observed species, M. axillaris, used mostly <2.5 cm perches, while the larger C. cinerescens used mostly 0.5 - 4.0 cm perches. The larger T nigrocinereous and T puntatus both used perches of all diameters between 0.5

5.0 cm. Perch use by all four species differed significantly from that expected if birds used perches according to their availability in the environment, indicating perch size











































Figure 2-7. Results of correspondence analysis between dietary guild, forest type, and forest strata preferences of 56 bird species captured in E. oleracea-enriched and nonenriched (control) stands. Circled letters represent guilds: f--frugivore; i=insectivore; n=nectanrivore; g=granivore; o=omnivore; and p=piscivore. Other notation indicates forest type/strata. Forest types are e=enriched; ne=non-enriched. Strata are: c=canopy; m=midstory; u=understory; t--terrestrial. Plot shows that insectivores were associated exclusively with the understory of non-enriched stands, omnivores were associated with the canopy of enriched forests, and frugivores were associated with the understory and midstory of both enriched and non-enriched forests. Piscivores, granivores, and nectarivores showed fewer forest type and strata preferences, being associated with both forest types and all strata.










2000-


more captures in afai-enriched stands


onuuvoras
Iinsectivores

frugivores nectarivores piscivores



-500
more captures in non-enriched stands




-1000species abbreviations


Figure 2-8. Percent difference in bird species abundance, estimated from mist-net captures, in five forest sites enriched for E. oleracea and five non-enriched sites. Change is shown relative to non-enriched forest, so that positive values indicate more captures of that species E. oleracea-enriched forest, and negative bars, more captures in non-enriched sites. Species are grouped according to dietary guilds. Abbreviations are as follows: Ep = Eucometis penicillata, Mm = Manacus manacus, Pa = Pipra aureola Mo = Mionectes oleaginea, Tr = Tachyphonus rufus, Rc = Ramphocelus carbo, Sn = Sporophila nigricollis, Sm = Saltator maximus, So = Saltator orenocensis, Oa = Oryzoborous angolensis, Lr = Leptotila rufaxilla, Sa = Sporophila americana, Pc = Pachyramphus castaneus, Pp = Pachyramphus polychopterus, Tf = Tolmomyias flaviventris, Ms = Myiozetes similis, Pm = Phaeomyias murina,,Ga = Geothlypis aequinoctialis, Ct = Columbina talpacoti, Cm = Crotophaga major,Cf = Celeusflavus, Ec = Eleania chiriquensis, Mg = Myiopagis gaimardii, Mc = Myiozetetes cayanensis, Cof = Coerebaflaveola, Ta = Turdus albicollis, Tuf = Turdusfumigatus, TI = Threnetes leucurus, Cl = Campylopterus largipennis, Pr = Phaethornis ruber, Ts = Thalurania sp., Thf = Thaluraniafurcata, As = Amazilia sp., Gh = Glaucis hirsuta, Ps = Phaethornis sp., Gs = Glyphorhyncus spirurus, Br = Basileuterus rivularis, Hm = Hypocnemoides melanopogon, Tp = Thamnophiluspunctatus, Lg = Lophotriccus galeatus, Xm = Xenops minutus, Ma = Myrmotherula axillaris, Cet = Cercomacra tyrannia, Tn = Thamnophilus nigrocinereous, Scn = Sclateria naevia, Xg = Xiphorhynchus guttatus, Xp = Xiphorhynchuspicus, Mq = Microhopias quixensis, Tm = Todyrostrum maculatum, Fg = Formicivora grisea, Ca = Chloroceryle aenea, Ci = Chloroceryle inda, Cha = Chloroceryle americana.








selection (M. axillaris: X'2 = 78.1, df= 2, p < 0.001; C. tyrannia: X2 = 40.6, df= 2, p <

0.001; T nigrocinereous: X2 = 42.3, df= 2, p < 0.001; T. punctatus: X2 = 42.4, df= 2, p <

0.001). Small-diameter perches, especially small-diameter vines and lianas, were extremely rare in agai-enriched forests (Fig. 5).

For the 15 bird species that were captured in both forest types, I compared the mean capture times as a way to assess activity levels over the course of the day. Mean capture time was significantly later (11:01 hrs) in the control forest than in the agaienriched forest (10:06 hrs; F,.,48 = 7.04, p < 0.008), implying that activity is concentrated earlier in enriched forest, perhaps because greater mid-day temperatures there deter midday activity of birds.

Mammal Trapping

The same three species of marsupials were caught in both agai-enriched and control stands (Table 2-3). There were no differences in capture rates between stand types ofMarmosa murina, the smallest species captured (X2 = 2.37, df= 1, p > 0.05). Too few individuals of the other species (Didelphis marsupialis, Caluromys philander) were captured to statistically compare. Capture success in small traps was nearly double in enriched stands (6.3 vs. 3.8%), but capture success in large traps was greater in control stands (2.3 vs. 1.3%).

Table 3. Summary of small mammal ca tures in the two forest types.
species # agaf- # control mean mass mean mass enriched + s.e. male + s.e female Marmosa murina 31 20 57+11.01 48.62+5.42 Didelphis marsupialis 2 3 510 775 Caluromys philander 1 1 149 200












a.

U M. axillaris
0 C. cinerescens


CV
.o 0.3

0.2


0.1 0.0







0.6- b.

0.5
O T. nigrocinereous
0.4' T. punctatus JJ







0.3




0 1 I"- F'l
0.0 r C. C V" " .
0.6















Stem diameter (cm)





Figure 2-9. Small tree branches, vines, and lianas used by the antbird species Myrmotherula axillaris, Cercomacra cinerescens (a), Thamnophilus nigrocinereous, and T. punctatus (b). Perches between 1 - 5 cm diameter are largely absent from E. oleracea-enriched stands.








Discussion

My purpose in this study was to compare the understory vegetation, birds, and mammals in agai-enriched and non-enriched Amazonian floodplain forests. Enrichment management of tropical forests is a recommended strategy to increase profits derived from forests by increasing the density of economically-valuable species (Schulze et al. 1994, Kainer et al. 1998, Ricker et al. 1999). Because enrichment with species that yield nontimber products can result in long-term increases in profits, this type of management is promoted in tropical forests in many regions (Schulze et al. 1994, Kainer et al 1998, Ricker 1999). This study revealed large differences between enriched and control forests in the composition and structure of both vegetation and bird communities. The mammal community, on the other hand, showed equally low diversity in both enriched and control forests. The apparent effects of enrichment management on forests have not before been appreciated, but should be considered wherever this strategy is employed. Changes in Flora and Fauna

Most of the changes I found in vegetation structure and composition are not

surprising given that the goal of management is to increase again production. As expected, enriched forests have higher densities of agai (Anderson and Jardim 1989, Hiraoka 1995, Pollack et al. 1995). Lower overall stem density and abundance of non-palm stems reduce perceived competition for nutrients and light; removal of vines and lianas also reportedly make enriched forests safer places to work. Thinning and canopy opening are recommended practices for increasing palm growth rates and fruit production (Anderson and Jardim 1989, Jardim and Rombold 1994) and are common throughout the Amazon








estuary (Anderson et al. 1995, Mufiiz-Miret et al. 1996). By all measures, aqai management succeeds in augmenting palm productivity (Anderson and Jardim 1989)

This study reveals, however, that management is also associated with a

substantially altered bird community. Agai-enriched stands contained a number of species not usually found in floodplain forests, and lacked others that are typical of floodplains of the region (Novaes 1970, Lovejoy 1974). Furthermore, the community associated with enriched stands was biased towards fruit- and seed-eating frugivores, granivores, and omnivores, whereas understory insectivores were underrepresented.

As suggested for forests enriched for other non-timber products (Greenberg et al. 1997), agai-enriched forest may serve as "secondary habitat" for some bird species. For example, frugivores may forage in agai patches but may nest and conduct other nonforaging activities in control habitat. I found no nests in enriched forest, but found four nests of T. nigrocinereous and one ofM. axillaris in control forest. Moreover, frugivore use of agai-enriched forest likely increases during aqai fruit production (when mist netting occurred). More detailed observational studies are required to determine the relative importance and seasonal use of different forest types for birds, many of which likely use more than one type.

In any study using mist nets, some caution is warranted in drawing inferences about bird communities because of potential biases (Karr 1981, Van Remsen and Good 1996). For example, differences in vegetation structure between enriched and control forest could differentially influence the capture of birds. The lower canopy in enriched sites, for instance, could result in a higher probability of capture of a canopy-dwelling bird. There was no evidence for this, however, as canopy-dwellers found in enriched








forest (e.g., Myiozetetes cayanensis, Pachyramphus castaneus, P. polychopterus, Eleania chiriquensis) were neither caught nor observed in control forest. Furthermore, several of the canopy species, such as Formicivora grisea and Coerbaflaveola, were not caught in control forest in this study nor by Novaes (1970) or Lovejoy (1974), and are considered edge or disturbance species (Hilty and Brown 1986).

Surprisingly, only three species of mammals were captured, and I saw no evidence, such as partially eaten fruits, feces, or tracks, of other species (except domesticated pigs). If the mammal community consists only of the three marsupials that were captured, then it is very depauperate compared with other Amazonian sites. People currently hunt D. marsupialis (S. Moegenburg pers. obs.), and it is possible that past hunting, over the long period during which humans have occupied the region (Hiraoka 1995, Roosevelt et a. 1996), reduced the species diversity of mammals.

The three mammal species that were captured showed no differences in abundance between forest types. This may be due to several factors. First, these three species are habitat generalists and omnivores and may not discriminate between the two forest types sampled in this study. As has been shown for marsupials in other habitats (Malcomb 1988), they tend to be less sensitive to habitat alteration than are other types of mammals, such as rodents. A second explanation may be the timing of trapping, which took place during the winter, when agai fruit is scarce. During the summer, when agai fruit is abundant, marsupials may utilize agai-enriched forest to a greater extent. Marsupials are known to respond to phenology of fruit production in other neotropical forests (Charles-Dominique et al. 1981). The mammalian community was undoubtedly








more diverse in the past, but has probably suffered from hunting and the high human population density in the region.

Mechanisms of Bird Community Change

The differences in the bird communities between enriched and control forests likely result from the differences in vegetation structure in these two forest types. In particular, the abundance of fruit-producing agai could explain the dominance by fruit- and seed eating species in enriched forests. Six of the 28 fruit-eating species caught in enriched forests showed evidence of agai fruit consumption (defecation of agai seeds or purple-stained bills). While the bounty of agai may be the lure for some frugivores, others may respond to an overall increase in fruit production resulting from the more open canopy in enriched forests. A similar result has been found in selectively logged forests (Johns 1988, Thiollay 1992, Mason 1996) and in coffee agroforests (Greenberg et al. 1997). In general, frugivores respond to spatial and temporal variation in fruit resources (Levey 1988, Loiselle and Blake 1991, Rey 1995). Nectar and fruit tend to be ephemeral resources, so may only function to attract animals to a habitat during the season of production (Levey and Stiles 1992).

A second indirect effect of agai management on avian communities could be the change in relative abundance of palm and non-palm trunks and leaves. The high abundance of agai trunks and leaves, and the low abundance of bark-producing plants, in the enriched understory could render the habitat less suitable for bark foraging (e.g., Glyphorhyncus spirurus, Xenops minutus) and foliage gleaning (e.g., Basileuterus rivularis, Hypocnemoides melanopogon) insectivorous species. Palm leaves are structurally well defended against herbivores, and may support fewer or different








arthropods than non-palm leaves. Likewise, barkless agai trunks provide poor substrates for bark-dwelling invertebrates and may therefore discourage use of the habitat by insect feeders.

Another mechanism for the differences in the bird community could be the lower abundance of small diameter perches in enriched forest. The four species of antbirds displayed preferences for small perches 0.5-3.0 cm in diameter, which were relatively and absolutely more common in control forest, despite the abundance of again leaves in the former. Some understory insectivores, such as Cercomacra cinerescens and Xenops minutus, specialize their foraging on small vines and lianas (Ridgeley and Tudor 1994). In addition, understory frugivores such as Manacus manacus and Pipra aureola use small understory perches for displaying (Snow 1962a,b). Indeed, the lack of small vines, lianas, stems, and branches in the understory could explain why, despite the higher abundance of fruit in enriched forest, four species of frugivore (M. manacus, P. aureola, Eucometis penicillata, and Mionectes oleaginea) were more abundant in the control forests. Owing to the dependence of many birds on small perches, the management practice of methodically removing vines and lianas may have one of the most detrimental effects on the understory bird community.

Finally, the higher temperature and lower relative humidity at noon, and the overall greater fluctuation in temperature and humidity in the enriched forests, could affect bird activity and community composition. For the 15 species that were caught in both forest types, mean capture time was significantly later (by one hour) in the control forest, suggesting that bird activity extends later into the day there, perhaps due to more favorable microclimatic conditions (Karr and Freemark 1983). Some species that were








captured primarily in enriched forest, such as Formicivora grisea, are considered secondary forest or scrub dwelling birds (Ridgeley and Tudor 1994). To the contrary, the river warbler, Basileuteris rivularis, is typical of forest interior species that seem sensitive to microclimatic changes associated with opening of the forest canopy and understory (Thiollay 1992, Mason 1996), further indicating that the abiotic differences that I recorded may indeed influence bird distribution and behavior. Such impacts of microclimate on birds have been implicated in other enriched forests (Karr and Freemark 1983, Johns 1988, Stouffer and Bierregaard 1995) and coffee agroforests (Bawa and Seidler 1998).

Comparison to Alternative Forest Management Systems

The differences in bird community structure between agai-enriched and control forests are comparable to those between agroforests, plantations, and logged and control forests. In particular, canopy dwelling and fruit-eating species tend to remain or become more abundant in modified forest, while forest interior, insect-eating species decline (Bawa and Seidler 1998). These patterns have emerged from agroforests in Sumatra (Thiollay 1995), shade coffee plantations in Mexico (Greenberg et al. 1997), cacao plantations in Brazil (Alves 1990), and logged forests in Venezuela (Mason 1996), French Guiana (Thiollay 1992), Malaysia (Johns 1988), Borneo (Lambert 1992), and Indonesia (Marsden 1998). As in this study, likely mechanisms for these bird community responses include higher flower and fruit production (Greenberg et al. 1997), greater canopy openness (Thiollay 1992, 1995, Mason 1996), lower understory stem densities (Mason 1996) warmer understory temperatures (Thiollay 1992), and fewer understory microhabitats such as vine tangles (Thiollay 1995) in modified forest. Forest management








has substantial ecological impacts, whether it is for timber, coffee, or non-timber forest products like agai.

Recommendations and Conclusions

My results indicate that the role of forests enriched for NTFPs in conservation should be re-evaluated. In particular, enriched forests such as those in this study can be valuable complements to, not replacements for, forests dedicated to conservation. Based on this study, I offer several recommendations to reduce the effects on bird communities of forest enrichment with E. oleracea. Further research is required on other non-timber forest product yielding species to assess the generality of both the results of this study and these recommendations.

Wherever enrichment takes place, steps should be taken to maintain

characteristics, such as canopy height and density, of control forests. Moreover, as little modification as possible should be done to the understory vegetation of enriched stands. Removal of small trees, vines, and lianas decreases perching and foraging substrates for understory animals. Furthermore, maintenance of both understory vegetation complexity and a dense canopy can help prevent large fluctuations in understory microclimate, to which understory animals can be sensitive. Research is needed to identify ways to successfully enrich forests with economic species without substantially modifying overall forest structure and composition.

My results suggest that control forests may play an important role in conserving regional biodiversity by supporting certain species, especially understory insectivores, not found in enriched forests. In my study area, the area of forest left control is shrinking, however, because market demand for agai fruit is growing, which encourages








landowners to increase agai enrichment in forests (Hiraoka 1995, Mufiiz-Miret et al. 1996).

Encouraging landowners to maintain a portion of control forest in the present state could be accomplished in several ways. One method would be to promote markets for other non-timber forest products, such as andiroba, which are harvested from control forests. A second method might be a community-wide certification program (Kiker and Putz 1997), in which people or families that maintain a certain percentage of their forest as control become certified producers of agai and earn more per volume of fruit than do non-certified producers. Higher prices captured for both agai fruit and palm heart or improved market access could balance any profits lost by leaving some forest control. Such incentive programs could be viable, because the landscape itself promotes the maintenance of the existing system, in which not all forest is intensively enriched. The estuary region consists of islands surrounded by rivers; most households and enriched forest lie along the river margins. Because the islands are roughly circular in shape, the individual land holdings take the shape of pie slices, joining at their points near island centers. Island centers are difficult to access and therefore usually remain non-enriched. The aggregation of the control areas of various families' forests means that if a certification program is implemented, then all the parcels of control forest would remain grouped. This may be the most effective manner in which to promote both forest utilization and conservation in the agai management system.












CHAPTER 3
LINKING FRUIT AND FRUGIVORE ABUNDANCE: EXPERIMENTAL EVIDENCE FROM AMAZONIAN BRAZIL Collecting the fruits produced by oligarchic forests is one of the most benign forms of resource exploitation practiced in Amazonia. If harvests are properly controlled and conducted in a non-destructive fashion, fruit collection has minimal impact on forest structure and function. Canopy cover and floristic composition are maintained, the constituent fauna is preserved, and the cycling of water and nutrients remain essentially unaltered. The fact that many oligarchic forests occur in habitats subjected to seasonal flooding and sediment deposition suggests that nutrient losses resulting from fruit removal are likely to be quite small.
Peters, 1992, pg. 17.


Introduction

The ubiquity of fruit-eating animals and fleshy fruit-producing plants in

neotropical forests (Gentry 1982, Karr et al. 1990, Robinson and Redford 1986, Levey and Stiles 1992) suggests that fruit-frugivore interactions influence the ecology and evolutionary history of the organisms involved (McKey 1975, Snow and Snow 1980, Bodmer 1989, Fleming 1991, Levey and Stiles 1992). One prediction stemming from this idea is that the abundance and distribution of fruits and frugivores should be linked (McKey 1975). Indeed, several studies demonstrate correlations between fruit and frugivore abundance both within and among habitats (Snow 1962a, b, Crome 1975, Worthington 1982, Wheelwright 1983, Levey 1988, Sargent 1990, Loiselle and Blake 1993), and others show that frugivores apparently track fruit resources over space and time (Loiselle and Blake 1991, Powell and Bjork 1995, Rey 1995, Kinnaird et al. 1996, Bjork 2000). Moreover, fruit availability seems to regulate populations of some highly-








frugivorous tropical animals (Foster 1982a, Terborgh 1986b, Snyder et al. 1987, Abramson et al. 1995, Adler 1998, Wright et al. 1999).

Another series of studies, however, calls into question the significance of fruitfrugivore relationships. Very few vertebrates, for example, are completely dependent on fruit; indeed many of them switch to other foods when fruits become scarce (Loiselle and Blake 1993, Pina 1999). In addition, fruit-frugivore interactions vary greatly over space (Bronstein and Hoffmann 1987, Fleming 1991, Chapman in prep.) and time (Herrera 1998), so their role in selecting for traits such as fruit morphology or frugivore behavior might in fact be limited (Herrera 1992). These results have led some (Levey and Benkman 1999) to question the ecological and evolutionary significance of fruitfrugivore relationships. That these relationships are more complex than was previously thought indicates the need for more refined research questions, such as: which frugivores are sensitive to fruit abundance and which are not? At what scale do frugivores detect variation in fruit abundance? At what hierarchical level (e.g., populations, communities) do frugivores respond? To date, however, answering these types of questions has been hindered by, among other things, the logistical difficulties of experimentally manipulating fruit abundance in forests.

Responses of frugivores to fruit abundance may depend upon many factors, including the scale at which frugivores operate in the environment, their degree of frugivory, their body size and coloration, and their social structure (Howe 1979, Pratt and Stiles 1983, Levey 1988, Chapman et al. 1989, Wheelwright 1991, Loiselle and Blake 1993, Westcott and Graham 2000). Large-bodied tropical animals with large ranges may not respond to local variations in fruit availability (Bourne 1974, Jardim and Oliveira








1997, Pina 1999), whereas smaller fruit-feeding mammals, birds, and insects might (Janzen 1970, 1971, Wright 1983, 1991, Levey 1988, Fleming 1992, Loiselle and Blake 1993, Adler 1998). In addition, highly frugivorous animals may display more sensitivity to fruit abundance than partially frugivorous ones (McKey 1975, Pratt and Stiles 1983, Levey 1988, Loiselle and Blake 1991, Chapman and Fedigan 1994, Allen 1997, Peck et al. 1999). Birds that are conspicuous to predators, either through coloration (Howe 1979) or vocalization (Bourne 1974, Snyder et al. 1987) may respond less to fruit abundance than birds that can conceal themselves in treetops. Finally, flocking species, such as parrots, may respond not only to food abundance but also to the presence of conspecifics (Snyder et al. 1987, Chapman et al. 1989).

If animals do respond to decreased fruit abundance, it may be on several levels: communities may contain fewer species (Martin and Karr 1986, Develey and Peres 2000); species may be represented by fewer individuals (Karr 1976, Levey 1988, Chapman et al. 1989, Loiselle and Blake 1993, Galetti and Aleixo 1998); and individuals may alter their behavior (Davidar and Morton 1986, Sargent 1990, Sherman and Eason 1998, Shepherd and Boates 1999). Such changes may, in turn, alter processes such as fruit removal and seed dispersal (Pratt and Stiles 1983, DeSteven and Putz 1984, Wheelwright 1991, Strahl and Grahal 1991, Chapman and Chapman 1995, Allen 1997, Hamann and Curio 1999, Loiselle and Blake in prep.).

Among the tropical fruits consumed by frugivores, palms are often considered

"keystone" resources that maintain populations during periods of fruit scarcity (Terborgh 1986b, Peres 2000). Palms are eaten by a high diversity of birds (e.g., Galetti et al. 1999) and mammals (Terborgh 1986a), often comprising an important component of the diet








(Foster 1982b, Terborgh 1986a, Snyder et al. 1987, Bodmer 1990, Allen 1997, Adler 1998). In southern Brazil, for example, Galetti et al (1999) documented one mammal and 14 bird species feeding on Euterpe edulis palm fruits, which comprise between 8 - 30% of the diets of these birds. Because of the apparent importance of palms to frugivores, then, palm-frugivore interactions might be the ideal system in which to test responses of frugivores to changes in fruit abundance.

In some tropical forests, natural variation in palm fruit availability (Foster 1982a, Terborgh 1986a, reviewed by Fleming 1991, Chapman et al. 1994) is magnified by extraction of wild fruits by people (Peters et al. 1989, Vasquez and Gentry 1989, Allen 1997, see also Galetti and Aleixo 1998). In the Amazon River estuary in eastern Brazil, for example, the palm Euterpe oleracea, which occurs in high-density stands (>300 trees ha~') across ca. 10,000 km2 (Calzavara 1972, Peters et al. 1989, Kahn 1991), produces highly-valued fruit that people harvest by climbing stems and removing infructescences (Anderson 1988, 1990a, Strudwick and Sobel 1988). Many people harvest only a small amount of fruit for household use, but some households practice intensive extraction of up to 9,000 kg fruit ha' yr'- (Muiiz-Miret et al. 1996). While it is recognized that many frugivores also consume E. oleracea fruit (Strudwick and Sobel 1988, Sick 1993), impacts of the harvest on frugivores have not been examined.

I took advantage of the E. oleracea system to test the effects of different levels of fruit abundance on frugivore habitat selection and behavior. I first monitored the abundance of parrots, parakeets, macaws (the most frequent visitors to E. oleracea trees), and E. oleracea fruit during one season of fruit production in four E. oleracea-dominated forests and four forests in which other species of palms occurred but were not dominant.








As this demonstrated a correlation between E oleracea fruit availability and frugivore abundance, I then experimentally removed E. oleracea fruit at two levels of intensity and monitored the community-level responses of mammals, birds, and seed-feeding bruchid beetles. These data were used to evaluate species-level responses of birds to determine which frugivores are most sensitive to fruit harvest and at which level of fruit reduction they cease to visit E. oleracea forest sites. With this information I formulate harvest recommendations that optimize both fruit utilization by people and maintenance of frugivore communities.

This research has special relevance to Amazonian Brazil, the forests of which are under increasing pressure from logging, mining, and ranching (Anderson 1990a). Brazil has implemented a strategy to conserve forests through a system of extractive reserves: areas designated for the long-term sustainable harvest of primarily non-timber forest products, including fruit (Fearnside 1989, Alegretti 1990, Mattoso and Fleischflesser 1994). E oleracea and related species are extracted in several of these reserves, so understanding the consequences of its fruit extraction is crucial to developing harvest guidelines that can maintain forest biodiversity (Vasquez and Gentry 1989, Peters 1990).


Methods

Study Species

The multi-stemmed E. oleracea occurs throughout the Brazilian States of Para, Amapi, Tocantins, and Maranhdo, along the Pacific coast of Colombia and northern Ecuador, and in Trinidad, Venezuela, and the Guianas (Henderson 1995). Across approximately 10,000 km2 of floodplain forests in the Amazon River estuary, E. oleracea forms monodominant stands, some of which are the result of historical or contemporary








management (Calzavara 1972, Peters et al. 1989, Kahn 1991). Individual genets (hereafter "trees" for simplicity) contain up to 25 slender stems that reach heights of 30 m (Henderson 1995). Reproductive stems produce infructescences bearing several thousand purple-black globose drupes ca. 1 cm in diameter.

The ubiquity of E. oleracea across inhabited estuarine regions suggests its

importance the diet, culture, and economy of the people in Amazonia (Anderson 1988, 1990b, Strudwick and Sobel 1988, Hiraoka 1995, Mufiiz-Mirit et al. 1996). E. oleracea yields two of the region's most profitable non-timber forest products: heart-of-palm and fruit. People harvest E. oleracea fruit by climbing stems and removing infructescences with machetes or knives. Fruits are processed into a drink consumed daily by many thousands of people. Where people have access to markets, they often increase fruit production by enriching their forests with E oleracea; they then sell the fruit not consumed in the household. On Combu Island near Bel6m, Brazil, several dozen families earn more than US$3,400 per year from the sale of E. oleracea fruit (Anderson and loris 1992).

Study Sites and Plots

This study was carried out at the 33,000 ha Estagio Cientifica Ferreira Penna (Ferreira Penna Scientific Station; 1o42'30" S, 510 31'45" W), operated by the Museu Paraense Emilio Goeldi of Bel6m and located within CaxiuanA National Forest in the municipality of Melgago, Para State, Brazil (Fig. 3-1). Average rainfall is 2,500 - 3,000 mm, mean annual temperature is 26C, and mean annual relative humidity is 85%. The vegetation is evergreen humid rainforest, and the majority of the station is non-flooded, terra firme forest (Lisboa 1997). Approximately 3,300 ha, mostly along the blackwater Bay of Caxiuani, is extremely low-lying forest that inundates when the river levels rise








during the rainy seasons (December - May) and high tides. Water depths reach their maxima (ca. 1 m) in May. Much of this floodplain forest is dominated by E oleracea, with Virola surinamensis (Myristicaceae) and Pterocarpus santalinoides (Fabaceae) also common (Ferreira et al. 1997). Although there are no homes and no current management of this forest, several lines of evidence suggest that the area of floodplain forest dominated by E. oleracea (hereafter "palm forest") resulted from past human management. First, areas of highest E. oleracea density occur closet to the river margin, where human dwellings tend to sit. Second, local people commented that the names of some of the sites (e.g., Moreira, Fazenda) reflect previous uses or names of inhabitants. E. oleracea fruit begins to appear in these sites in May and persists until September. In most years people occasionally extract E oleracea fruit from these areas; however, in 1997 and 1998 local inhabitants respected my request to refrain from extraction.

Data collection occurred during the fruiting seasons of 1997 and 1998. Within palm forest I located four sites, called Plaquinha, Fazenda, Miriti, and Moreira, located approximately 1 km apart along the bay, in which I worked in both years. Within the non-floodplain forest I located an additional four sites, known as Estaqgo Sur, Estaio Norte, Heliporto, and Invent6rio, in which I collected data during 1997 only. In 1997, I randomly located 1 ha square plots in each of the eight sites, in which I surveyed parrots (all sites) and monitored E oleracea fruit abundance (palm forest sites only). These plots were demarcated with colored flagging and were bisected by a 100 m-long transect.




























XField Station o Study Sites

Figure 3-1. Study site within Brazil, and E oleracea-dominated study plots and field station within site.



Fruit and frugivore surveys in 1998 were conducted in one circular plot per site, with diameter 150 m (area = 1.8 ha; Fig. 3-2), located roughly in the center of highest density of E. oleracea at Plaquinha, Fazenda, Miriti, and Moreira. Each plot was divided into three equal subplots, which were further divided in half by a transect along which all censusing took place. Along each transect I constructed a "footbridge" of tree trunks to facilitate walking during flooded periods. Subplot edges, plot perimeters, and transects were demarcated with colored flagging.

Fruit Availability

In both years I censused fruit availability by examining E. oleracea trees along transects and recording the number of ripe infructescences by size (small, medium, or








large). During 1997 I estimated fruit abundance during July and August in the palm forest plots by walking the 100 m-long transect in the middle of each and recording infructescences on 50 trees adjacent to the transect. In 1998, fruit surveys were carried out twice monthly during the entire fruiting season (May - September) on 20 trees adjacent to transects in each of the three subplots, for a total of 60 trees per plot. The individuals observed were not necessarily the same in all censuses, but rather were chosen randomly from those occurring along the transects.


plot area = 1.8ha
- subplot boundaries
- - -. transects

Figure 3-2. Schematic map of study plots, divided into three subplots, each with a 75 m transect down the middle. One subplot was randomly designated high-removal, another low-removal, and the third, control. During frugivore censusing, observer 1 began near the edge of the control subplot, while observer 2 started at the plot center. Bold and italicized numbers next to arrows indicate paths and directions walked by observers 1 and 2, respectively.








To convert the number and sizes of infructescences to total available fruit mass, I first weighed the fruits from five harvested small, medium, and large infructescences. These average weights were then multiplied by the total number of small, medium, and large infructescences recorded in each plot or subplot. This total fruit mass was then divided by the number of trees censused (50 in 1997, 20 in 1998) for a per tree estimate of fruit availability. To determine the total fruit available per area, the per tree fruit availability was multiplied by the number of trees in the hectare plots (1997) or subplots (1998). This was done on a per-subplot basis in 1998 due to differences among subplots in E. oleracea tree densities.

Parrot Use of Palm Forest Versus Non-palm Forest

Before evaluating the community-level responses of frugivores to fruit harvest, I wanted to determine unmanipulated frugivore densities and seasonal fluctuations in palm forest and non-palm forest. I therefore monitored the visitations of the most abundant palm forest frugivores (parrots, parakeets, and macaws) to both forest types during the E. oleracea fruiting season of 1997. I conducted four surveys in the four palm forest sites (Plaquinha, Fazenda, Miriti, and Moreira) and three in the four upland sites (Estagio Sur, Estagdo Norte, Heliporto, and Invent6rio). Surveys involved slowly walking along the 100 m-long transect that bisected each 1 ha plot from 0700- 1000 hrs and recording all pssitacids within the plot, including species using the area, size of groups, and time spent in palm trees and eating fruit (in palm forest).

I estimated the number of minutes that pssitacids spent in the censused areas

conservatively as the longest interval between observations of a group. At the end of the season, an average group size was calculated based on all observations in which group size could be ascertained. Over the census period I detected no changes in parrot group








sizes (Chapman et al. 1989), which were: Pyrrhuraperlata (12), Pionites leucogaster (4), Deroptyus accipitrinus (3), Amazona amazonicus (2), and Ara spp. (2). For each group observed, I multiplied the number of minutes it spent in the plot by the average group size for that species. The sum of this value for all groups observed during a census was termed the parrot visitation in that plot.

Frugivore Responses to Experimental Fruit Harvest

The truest experimental design is a pre-test - post-test of controls and treatments (James and McCulloch 1995). To maximize the robustness of my experiment in 1998, I therefore surveyed both fruit abundance and frugivores during a pre-harvest phase (May and June) and a post-harvest phase (July and August) in three treatments. Survey methods were identical in the two phases.

Frugivore surveys

While frugivore surveys in 1997 focussed on pssitacids only and were meant to establish a correlation between fruit and frugivore abundance, surveys in 1998 were expanded to include a variety of types of animals that likely perceive fruit abundance on different scales. I therefore censused fruit-eating mammals, fruit-eating birds, and seedfeeding bruchid beetles. Moreover, although their abundance was not expected to change in response to fruit harvest, I censused non-frugivorous birds, as a general gauge of seasonal fluctuations in bird abundance. Animals were classified as non-frugivores, frugivores, partial frugivores, or granivore/frugivores based on personal observations of fruit-eating and the literature (Hilty and Brown 1986, Levey 1988, Ridgeway and Tudor 1994, Emmons 1997, Levey and Stiles 1992, 1994). Only two species of nonfrugivorous mammals were observed in the plots during censuses: southern tamandua








(Tamandua tetradactyla) and grisons (Galactis vittata). One species of bruchid beetle (Pachymerus sveni) is known to feed on E. oleracea seeds (Johnson et al. 1995).

Surveys of birds, mammals, and beetles were conducted on the same days as the fruit surveys during 1998. From 0700 - 1130, two observers simultaneously walked transects in a pre-determined pattern, one starting at the plot edge and the other at the plot center (Fig. 3-2). They then slowly walked toward the center or edge, respectively, allowing thirty minutes to walk one-way on a transect. In this way the observers were always in different subplots and walked each transect twice during the 4.5 hr survey. Because surveys were conducted at ca. two week intervals, with few intervening visits to plots, it is unlikely that animals became habituated to our presence and modified their behavior. Each observer recorded on a schematic map, resembling Fig. 3-2, all birds and frugivorous mammals seen and heard, paying particular attention to the subplot in which animals were located. Multiple observations were recorded to estimate visitation time in the plot. The simultaneous observations by two observers allowed for more accurate mapping of animals. Later, final maps of animal locations and visitation times during each survey were compiled. I developed rules to aid in estimating the number of individuals and visitation times of birds. First, the number of individuals of flocking species, if not readily countable, was assumed to be the average number in flocks, as in 1997 (see above). Second, we frequently detected a bird first in one subplot, and later in another subplot. In these cases, the individual was assumed to have spent equal amounts of time in the two subplots. Final summaries of census data therefore included number of species, number of individuals, and visitation times (mins) for frugivorous birds, nonfrugivorous birds, and mammals.








Bruchid beetles were not surveyed directly, but rather by the number of eggs they oviposited. Under five E. oleracea trees in each subplot (N=15 per site) I placed a mesh bag (mesh size 1 cm2, allowed easy passage by beetles) containing 10 E. oleracea seeds. At each bi-monthly sample, seeds were removed, placed in plastic bags, and replaced with fresh ones. After ca. 6 months, seeds were checked for bruchid beetle emergence. These emergence rates were reduced by a factor that controlled for beetles that were in the seeds prior to their use in the experiment (i.e., beetles that had entered seeds while they were still on trees). For each batch of seeds used, therefore, a sample (>100 seeds) was stored in dry, plastic bottles fitted with mesh "windows" and later (after 4-9 months) checked for beetle emergence. An average base infestation rate was thereby obtained, and this was subtracted from the actual rates observed in the experimental seeds. Fruit removal

Mid-way through the 1998 fruiting season (early July) I initiated the E. oleracea fruit harvest experiment. In each plot one subplot was randomly selected as highremoval (-100% ripe fruit removed), and one as low-removal (-50% ripe fruit removed). These harvest intensities mimic extraction that is done for both consumption and marketing of fruit, and for household consumption only, respectively. In each plot one subplot was left as a control (0% fruit removed) to mimic the condition of non-harvested forests. All harvests were carried out one day prior to frugivore surveys in that plot.

Four local teenagers were contracted as harvesters. They were told from which areas to harvest, but were not told which infructescences to harvest. This allowed the extraction to be carried out in as realistic a manner as possible; that is, harvested infructescences represented those that would actually be selected by local people. To remove fruit, harvesters climbed the palms and cut off the infructescences with machetes.








This involved climbing up to 120 palms per plot. No trees were climbed in control plots. On the ground, fruit was removed from infructescences, put into baskets, and used for consumption. Harvest from some palms was impossible, as they were too thin and fragile to allow climbers to safely reach the infructescence. After the initiation of the harvests, bird, mammal, and bruchid beetle surveys continued as before. In each plot four harvests and four surveys (on the day subsequent to harvest) were conducted during July and August.

Statistical Analyses

I tested for a correlation between ripe E. oleracea fruit and the number of minutes that parrots spent in plots in 1997 using SAS JMPIN (1996).

The fruit removal experiment was a repeated measures factorial design with two between (site and treatment) and two within (pre-vs post-harvest and census) factors. Each of the eleven response variables (Appendix 3-2; Tables 1-9) was therefore used in a repeated measures ANOVA, conducted with SuperANOVA (Abacus Concepts 1993). If necessary, data were first transformed to achieve normality and equal variances. The experimental design, with only four replicates, had low power to detect true differences between treatments (Zolman 1993). To increase power, and reduce Type II error, I therefore chose 0.1 as the initial significance value rather than the traditional value of

0.05. This value was adjusted, however, using a Bonferroni sequential technique (Rice 1989), because multiple (three) comparisons were conducted on each data set.

In addition to testing, on the community level, the existence of a relationship between fruit and frugivore abundance, I sought to provide data on which to base sustainable fruit harvest guidelines. In other words, I wanted to determine which frugivore species were most sensitive to fruit harvest, and the threshold harvest intensity








beyond which they ceased to visit the plots. To do this, I first used the data from the 1998 fruit and frugivorous bird surveys in logistic regression analyses using SAS JMPIN Software (1996). In logistic regression, a continuous independent variable (fruit abundance) is used to predict the state of a discreet, dependent variable (presence vs. absence of each frugivore species). I first performed the analyses to evaluate which species' presence or absence was significantly determined by the abundance of fruit ("fruit-sensitive species"). These analyses only included those visitations of > 5 min, to exclude birds that landed in plots but did not stay. I then used the logistic linear model to calculate, for each fruit-sensitive species, how much fruit would be required for a 25, 50, 75, and 99% probability of that species visiting a plot. Second, I tested for relationships between the amount of fruit available and the number of minutes that each species spent in subplots by regressing the number of minutes against kg fruit available, after both variables were log-transformed. In practical terms, managers could use these data to establish the minimum amount of fruit that should remain in the forest to ensure, within a chosen probability, the persistence of a certain species.




Results

Parrot Use of Palm Forest Versus Non-Palm Forest

Pssitacids apparently use E. oleracea-dominated forest to a greater extent than

non-palm forest, and this use is associated with ripe palm fruit abundance (Fig. 3-3a). In 1997, fruit abundance was greatest at the initial survey and declined steadily through late August, as did the length of parrot visits. I found a positive correlation between ripe fruit abundance and the number of minutes parrots spent in palm forest plots (r=-0.42,









F1,12=8.4, p<0.01). The number of minutes that parrots spent in non-palm forest showed no such temporal trend and was consistently lower than in the E. oleracea forest (Fig. 33b).


a.








fruit abundance


-80

-70

-60

-50

-40

-30


early July late July early August late August


- III
late July early August late August

Timing

Figure 3-3. Censuses in 1997. Shown are mean + s.e.
a) Fruit abundance (kg) and parrot visits (min) were positively correlated in
I ha plots at four E. oleracea -dominated forest sites.
b) Parrot visits (min) in I ha plots at four non-E. oleracea forest sites.


350300250

200150100500-








Frugivore Responses to Experimental Fruit Harvest Fruit availability

As in 1997, fruit production in 1998 peaked in July and declined through August (Fig. 3-4a). There were no significant differences in ripe fruit availability among treatments in the pre-harvest phase (F2,.s=0.95, p=0.44; Fig. 3-4, Table Al). Fruit harvest significantly reduced fruit abundance (Figs. 3-4a, b & c). Low-removal treatments had on average 41% less fruit than control subplots (F,18s=4.15, p=0.09), whereas highremoval treatments had on average 75% less fruit (F.18s=10.52, p=0.02). Low-removal and high-removal treatments did not differ significantly in their fruit availability (F1.s=1.46, p=0.27).

Birds: community-level responses

Forty-one species of frugivorous birds visited plots during 1998 (Appendix 3-1). Their responses to fruit harvest are indicated by significant "harvest x treatment" interaction terms in the repeated-measures ANOVA (Tables A1-A3; Appendix 3-2).

Prior to fruit harvest, no differences existed (Fig. 3-5) among treatments in the number of frugivorous species (F2,18=0.34, p=0.72), individuals (F2.is=0.13, p=0.88), or their visit lengths (F2z8=0.04, p=0.96). Once fruit harvest began, however, the number of species declined by 25% (F,18s=13.08, p=0.01), the number of individuals declined by 29% (F1,18=34.96, p=0.001), and frugivores spent 68% less time (F,18=22.50, p=0.003), in the high-removal subplots, compared with the controls (Fig. 3-5). Moreover, highremoval treatments had significantly fewer species (F1.8=11.60, p=0.01), individuals (F,.,8=70.61, p=0.0002), and visitation times (Flsl=22.24, p=0.003) than the low-removal treatments (Fig. 3-5). In contrast, low-removal subplots showed no significant









pre-harvest


post-harvest


100

80 a. control 60

40 20

0


b. low
removal





i ii


Census


< <


Figure 3-4. Average E. oleracea fruit abundance in four 0.6 ha sites, over time. Censuses in May and June are pre-harvest; those in July and August are post-harvest. Shown are mean + s.e.
a) Control.
b) Low-removal: 4-56% less fruit than controls.
c) High removal: 43-75% less fruit than controls.









post-harvest


+-


108
6
4
2

0

25
201510
5
0


900


600


300-


C.






rF -17


control


- a


b
,-T.,


high control


Treatment



Figure 3-5. Effects of E. oleracea fruit harvest on frugivorous birds. Shown are mean + s.e. Columns labeled with different letters differed significantly (p < 0.05). No differences existed pre-harvest.
a) Number of species: post-harvest, high-removal treatments had 25% fewer
species than controls.
b) Number of individuals: post-harvest, high-removal treatments had 29%
fewer individuals than controls. Low removal treatments had slightly more
individuals.
c) Visit lengths: post-harvest, visits to high-removal treatments were 69%
shorter than to controls.


b.


b


high


low


I I I m R R


I


pre-harvest


FI_








differences from controls in the number of species (F1.1s=0.04, p=0.84), or visit durations (FI,18=0.01, p=0.98), but had more individuals (F . =6.20, p=0.047).

I recorded 40 species of non-frugivorous birds in the plots (Appendix 1). Nonfrugivores showed no differences in number of species, number of individuals, or visitation times among treatments (Fig. 3-6) either before (all p's>0.51) or after (all p's>0.73) fruit harvest (Tables A5-7).

The repeated-measures ANOVA's revealed several significant main and

interaction effects in addition to the "harvest x treatment" interaction. As these effects are not central to the study questions, they will be reviewed here, but not discussed in detail.

Numbers of both frugivorous and non-frugivorous birds varied spatially and

temporally, as indicated by the significant effects of site, harvest, and time (Tables AlA6). The significant "site" effect indicates variation among my four replicate sites in the abundance and behavior of birds. Fazenda and Miriti consistently had more individuals of frugivorous birds that spent more time than did Moreira and Plaquinha. The reverse pattern was true for non-frugivorous birds, which regularly were represented by more species at Moreira and Plaquinha than the other two sites.

In several of the repeated-measures models "harvest" was a significant effect

(Tables Al-A6). This factor compares the responses of birds in May and June (the preharvest phase) with those in July and August (the post-harvest phase), independent of treatment. This effect of harvest on the number of frugivorous bird species and individuals depended, however, on site (significant harvest x site interaction). At Fazenda and Miriti, both frugivorous and non-frugivorous birds were more abundant and








pre-harvest


post-harvest
_ T


c 1 h c I h


Treatment


Figure 3-6. Effects of E. oleracea fruit harvest on non-frugivorous birds. Shown are mean + s.e. No significant differences existed among treatments either pre-or postharvest.
a) Number of species.
b) Number of individuals.
c) Visit lengths.








paid longer visits to sites in July and August than in May and June, whereas at Plaquinha and Moreira, the abundance of birds did not vary much between the pre- and post-harvest phases.

The factor "time" was significant in all of the models. This factor explains

variation among the four censuses within each harvest phase (i.e., early May, late May, early June, etc.). In five of the six ANOVA's, the effect of time depended upon "harvest" (significant harvest x time interaction). In these cases, the effect of time was important in the pre-harvest phase, but not important in the post-harvest phase - this is likely due to several factors. First, as fruit production began and progressed in May and June, the number of frugivores increased. In the post-harvest phase however, fruit abundance was kept at a more constant level by the experimental manipulations, and the number of frugivores overall in plots also stayed more constant. Non-frugivorous bird abundance increased also over time in the pre-harvest phase but did not fluctuate in the post harvest phase. Non-frugivores likely responded to other factors, such as water levels or their breeding cycles. In two of these five ANOVA's, (frugivorous bird species and frugivorous bird individuals) the interaction between harvest and time also depended on site (significant harvest x time x site interaction). This can be explained by the fact that the number of frugivores increased dramatically over time within the pre-harvest phase at Fazenda and Miriti, but not at Plaquinha and Moreira. The increase in bird abundance may be due, in part, to the arrival of species that immigrate from other areas in Amazonia.








Birds: species-level responses

In addition to evaluating community-level responses to fruit harvest, I used three methods to identify "fruit-sensitive species" - those whose presence and/or visitation times were affected by fruit abundance. First, I simply counted the species that no longer visited high-removal and low-removal subplots, once fruit harvest began. Of the regular visitors, six species ceased visits to the high fruit -removal sites. These included a tanager (Thraupispalmarum), a trogon (Trogon collaris), a parrot (Deropteus accipitrinus), two parakeets (Pionites leucogaster and Brotogeris versicolorous), and a macaw (Ara macao). Only one species, Trogon viridis, stopped visiting low-removal sites.

Second, logistic regression revealed that the presence of six of the 20 species analyzed could be predicted by the abundance of fruit (Table 3-1). The probability of these six highly frugivorous species occurring in plots increased as fruit abundance increased, and these are the species most responsible for the lower species richness in high-removal treatments. Indeed, two of them (Ara and Pionites) are among those that ceased visits altogether. The most sensitive to fruit availability was Rhitipterna simplex, requiring 62 kg and 869 kg fruit for a 25 and 99% chance of occurring, respectively. The least sensitive was Vireo olivaceus, requiring 0 kg and 350 kg fruit for a 25 and 99% chance of occurring, respectively.

Finally, I used linear regression to evaluate the relationship between the length of frugivore visits and the abundance of fruit. The length of frugivore visits was less related to fruit abundance than was frugivore presence or absence (Table 3-2). Only three species (Pionites leucogaster, Ramphastos tucanus, Vireo olivaceus) demonstrated a









Table 1. Results of logistic regression between E. oleracea fruit abundance (independent variable) and the probability of the presence of six frugivorous bird species (dependent variable). All df= 1.

Percent probability of occurring
25 50 75 99
species X p < kg fruit needed per ha Ara macao 2.87 0.090 73 197 238 718 Pionites leucocephala 5.02 0.020 47 140 233 528 Pyrrhuraperlata 8.82 0.003 13 87 160 395 Attila cinnamomeus 3.66 0.050 100 203 307 635 Rhitipterna simplex 2.90 0.080 103 363 622 1448 Vireo olivaceus 3.16 0.080 32 42 172 583














Table 2. Number of minutes that frugivorous bird species spent in treatment subplots (mean + s.e.), and its relationship with kg fruit per subplot.


number of minutes


regression on fruit abundance


species control low removal high removal r' df/ F p Columba speciosa 0.25+0.25 1.03+0.94 0.78+0.75 ns Ara macao 3.83+1.74 12.50+8.76 2.81+2.07 ns Ara severa 2.19+1.58 2.36+2.01 0.75+0.46 ns Amazona amazonicus 7.39+3.29 14.00+5.34 2.83+1.31 ns Pyrrhura perlata 65.64+31.83 142.72+84.79 34.86+14.39 ns Pionites leucogaster 20.94+11.67 31.92+14.21 11.50+6.86 0.13 1,25/3.48 <0.07 Brotogeris versicolorus 6.67+6.10 19.47+9.27 0.11+0.11 ns Ramphastos vittelinus 17.65+5.72 19.99+13.82 9.85+3.76 ns Ramphastos tucanus 28.52+8.34 26.90+14.50 13.42+4.02 0.08 1,56/4.62 <0.04 Celeus undatus 7.14+6.66 2.86+1.85 2.47+0.95 ns Trogon viridis 30.08+15.37 6.93+3.12 6.36+2.68 ns Eleania flavogaster 15.07+3.72 16.31+9.14 11.63+3.02 ns Tyrannulus elatus 6.84+3.76 3.06+1.39 7.26+2.84 ns Rhitipterna simplex 3.94+2.15 7.50+4.85 0.35+0.27 ns Attila cinnemomeus 13.56+7.99 1.53+0.91 0.81+0.65 ns Turdus albicollis 1.72+0.92 5.07+2.66 7.79+4.36 ns Cacicus cela 0.94+0.42 2.53+0.89 1.31+0.56 _ns Vireo olivaceus 14.14+6.23 23.86+8.02 8.50+3.14 0.16 1,52/9.68 <0.003 Ramphocelus carbo 0.13+0.08 3.58+2.26 0.15+0.09 ns








significant response to fruit abundance. Because these three were among the most frequent visitors, the community as a whole showed a strong response to fruit harvest (Fig. 3-5c).

Mammals

Five species of mammals known to eat E. oleracea fruit were observed in the

plots: Guianan squirrel (Sciurus aestuans), South American coati (Nasua nasua), brown capuchin monkey (Cebus apella), red-handed howler monkey (Alouatta belzebul), and golden-handed tamarin (Saquinus midas). As with birds, a significant response of mammals to fruit harvest is indicated by a significant "harvest x treatment" interaction.

Similar to the observed responses of birds, mammals showed no differences in the number of species, individuals, or visitation times among treatments in the pre-harvest phase (Fig. 3-7, Tables A7-A9). Post-harvest, however, the number of mammal species was 58% lower in the both high-removal (F,18=9.59, p=0.02) and the low-removal (F1 as=9.59, p=0.02) treatments, compared with the controls; the low- and high-removal treatments did not differ (F,18=0.01, p=0.99). The lower species richness in high-removal subplots reflects the lack of visits there by both howler monkeys and tamarins once fruit harvest began. Howler monkeys, but not tamarins, also stopped visits to low-removal treatments after fruit harvest; squirrels and capuchin monkeys made less frequent visits post-harvest to low-removal treatments, which resulted in the lower species richness. Too few observations of coatis were made to detect a pattern for that species. Unlike species richness, the number of individuals and their visit times of mammals did not differ among the treatments after fruit harvest (Fig. 3-7; all p's>0.09).

As with birds, mammal species (F1,18=18.78, p=0.005) and individuals

(F1,18=29.26, p=0.043) showed an overall significant response to harvest, independent of









pre-harvest a.






rfi FLFf


post-harvest


4- b. 3
2


1~r�~rF+]


c 1 h c 1 h


Treatment


Figure 3-7. Effects of E. oleracea fruit harvest on fruit-eating mammals. Shown are mean + s.e. Columns labeled with different letters differ significantly (p < 0.05). No differences existed pre-harvest.
a) Number of species: post-harvest, both high- and low-removal treatments had
58% fewer species than controls.
b) Number of individuals.
c) Visit lengths.


1 0.8
0.6
0.4
0.2-


II I








treatment (Tables A7 & A8). I found both more species and individuals in all three treatments post-harvest (Fig. 3-7), when overall fruit availability in plots was greater. This effect varied, however, among the sites, being stronger at Fazenda and Miriti than at Plaquinha and Moreira.

Bruchid beetles

Beetle emergence showed no differences among treatments, post harvest (Fig. 38; F2.,15 = 0.346, p=0.71). The number of eggs oviposited was most affected by site (F3.150 = 2.90, p=0.036), with the greatest emergence rates at Plaquinha and the lowest rates at Fazenda.


0.16

0.12

0.08

0.04

0-


control


low
Treatment


high


Figure 3-8. Effect of E. oleracea fruit removal on the number of eggs oviposited by bruchid beetles on E. oleracea seeds. No significant differences were found among low-removal, high-removal and control sites. Shown are mean + s.e.








Discussion

Frugivores responded strongly to both natural and experimental declines in palm fruit abundance. When intensive fruit extraction by people was mimicked (75% ripe fruit; high-removal treatment), frugivorous bird species richness, abundance, and visits, and frugivorous mammal species richness, were significantly reduced. Removing 41% of the ripe fruit (low-removal treatment) had no such effects. That non-frugivorous birds showed no differences among treatments is additional evidence that frugivores responded to fruit abundance rather than to other factors. Although these effects were tested on a small scale relative to the home ranges of most frugivores, they may occur at larger scales since human fruit harvest is both extensive and intensive throughout the region. Frugivore Responses to Fruit Harvest

Birds - community level

Spatial and temporal fluctuations in tropical frugivore species richness,

abundance, and visit lengths have long been noted (Davis 1945, Fogden 1972, Karr 1976, Terborgh 1977). These have been linked to fruit abundance at various scales (Snow 1962a,b, Levey 1988, Loiselle and Blake 1993, Wright et al. 1999), and several studies have demonstrated that frugivores apparently track fruit abundance across space and time (Loiselle and Blake 1991, Powell and Bjork 1995, Rey 1995, Kinnaird et al. 1996, but see Herrera 1998). Nevertheless, many other factors (e.g., weather, breeding cycles) also influence the abundance and distribution of birds in forests (Karr 1976, Terborgh 1977, Karr et al. 1990), and these factors are often not considered in studies of fruits and frugivores. While fruit abundance is frequently implicated as the factor driving frugivore changes in abundance and movements, this idea has not yet been tested experimentally in a natural system.








This study is apparently the largest controlled manipulation of arboreal fruit abundance, to date. Along with manipulations of fruit abundance on the forest floor (Adler 1998, Sherman and Eason 1998), it provides strong evidence that not only are abundances of fruits and frugivores linked, but that fruit is the mechanism behind frugivore responses. If the strong influence of fruit abundance found in these studies represents a general pattern, then fruit abundance may help explain frugivore species diversity, abundance, and behavior across many scales: in foraging flocks (Chapman et al. 1989, Develey and Peres 2000) and individual fruiting trees (Howe and Vande Kerckhove 1980), within (Loiselle and Blake 1993) and among (Levey 1988) understory habitats, across regions (Rey 1995, Levey and Stiles 1992), and through time (Martin and Karr 1986, Bronstein and Hoffmann 1987).

Frugivore communities not only had lower species richness and number of

individuals in areas of high fruit removal; the species comprising the communities also spent 68% less time there. The length of individual visitations among all species ranged between 0.13 - 65 min. These visit lengths are similar to those found for frugivores in individual trees in both Papua New Guinea (Pratt and Stiles 1983) and Costa Rica (Wheelwright 1991). The amount of time that frugivores spend in fruiting patches has rarely been examined directly at levels above individual trees, although Sargent (1990) found that as the amount of fruit on both fruiting plants and in fruiting neighborhoods declined, so did visitations by both flocking and non-flocking bird species.

My results differ from those of Galetti and Aleixo (1998) who studied the impacts of reduced Euterpe edulis palm fruit abundance resulting from harvest of trees for heartof-palm. These authors found no difference in the abundance of frugivorous birds in








harvested vs. unharvested forest, and argue that because E. edulis produces fruit concurrent with many other plant species, the 96% reduction in palm density in harvested forest represented a small overall loss of fruit. Clearly, such a large reduction in fruit abundance would induce large responses from frugivores in my Amazonian site. One explanation for the differences in outcome between these two studies could be differences in the spatial and temporal scales of frugivore censusing:Galetti and Aleixo censused birds in a much larger area from which palms had been harvested 5 - 10 yrs previously In addition, fruit fall rather than fruit production was measured; fruit fall does not always represent the amount of fruit available to frugivores (Chapman et al. 1994). The lag time between harvests and censuses, and the lack of data on fruit production, may have made finding a relationship between fruit and frugivores inherently unlikely. Alternatively, the larger scale at which they censused may indicate that the small-scale differences I measured do not hold at larger scales (but see Rey 1995).

Although I consider it unlikely, frugivores may have responded to the human

activity associated with my fruit harvests rather than to the reduction of fruit. In reality, there were two differences between the experimental treatments and the controls: tree climbing and fruit removal. Many of the frugivores observed in this study certainly are sensitive to human activity and would not remain in an area while fruit harvest was occurring. Others, such as Amazona parrots, are relatively insensitive to human activity at my sites. Because fruit harvest occurred a full day prior to the frugivore surveys, however, I believe that any disturbance, aside from the odor of humans (which could have affected mammals but not birds) caused by the activity would no longer have been present.








Birds- species level

The lower species richness in high-removal treatments has at least three

components. First, ten "fruit-sensitive" species, defined as those that either discontinued visits to high-removal sites (six species) or whose visits depended, according to logistical regression, on fruit abundance (4 additional species), account for the 25% reduction in species richness in high-removal sites. A second factor, however, is the lower consistency with which other frugivores visited those subplots. Some species, such as Turdus albicollis and Pipra rubrocapilla, continued to visit high-removal sites, but did so more sporadically once fruit harvest began. Lowered visitation rates could result if reduced fruit availability caused by harvest forced frugivores to search a wider area for food (Chapman et al. 1989, Fleming 1991). A third component of lower species richness in high-removal treatments is the fact that fewer individuals were visiting those sites. With such a diverse frugivore community (41 species), the observed 29% reduction in frugivorous individuals would almost certainly lead to the loss of some species, just by chance alone.

What determines which birds visit particular areas and how long they spend

there? The most obvious factor is degree of frugivory; both Levey (1988) and Loiselle and Blake (1993) showed that the abundance of manakins, the most heavily frugivorous birds in their study sites, was correlated with small-scale fruit abundance. In Papua New Guinea, Pratt and Stiles (1983) found that highly frugivorous fruit pigeons spent more time in fruiting trees than did partially frugivorous bowerbirds, which spent more time than did birds of paradise, which eat fruit only occasionally. I found that the presence of six species, and the visit lengths of one additional species (Ramphastos tucanus), were predicted by the abundance of fruit in 0.6 ha subplots. All of these species include








substantial amounts of fruit in their diet, but probably not more so than the frugivorous species that did not respond to fruit abundance. In other words, the responses of individual species do not appear to be determined solely by their degree of frugivory, a result also found by Wheelwright (1991). As Levey (1988) suggested, fruit abundance may not be equivalent to fruit availability, because competetive interactions may limit access to fruit for some species. Toucans, for example, are known to supplant other bird species feeding on the same trees (Bourne 1974).

In addition to the abundance of their primary food source, the abundance,

distribution, and behavior of birds may be influenced by alternative food abundance, crypsis, body size, and social systems (Howe 1979, Pratt and Stiles 1983, Chapman et al. 1989, Wheelwright 1991). Of the birds recorded in this study, only five qualify as highly frugivorous (Pipra rubrocapilla, Gymnoderusfoetidua, Cotinga cayana, Cotinga cotinga, Querulapurpurata); the others commonly consume alternative foods, especially insects. I did not monitor insect abundance, but in other tropical forests it tends to be low in the dry season (Davis 1945, Karr 1976, Develey and Peres 2000), which is when my study took place. If insect abundance varied over the course of my study and among my treatments, inducing the differences among treatments in frugivores, then similar or stronger responses should have been detected for non-frugivorous bird species. Because neither the species richness, number of individuals, nor visitation times of non-frugivores differed among treatments, it seems unlikely that insect abundance played a role in frugivore abundance and distribution.

Howe (1979) suggested that crypsis determines, in part, the amount of time

frugivores spend in fruit patches, because colorful birds are more obvious and vulnerable








to predation. In support Howe's hypothesis, Pratt and Stiles (1983) found that less conspicuously colored fruit pigeons spent more time in fruit trees than did more colorful bowerbirds and birds of paradise. On the other hand, Wheelwright (1991) found no such pattern in Costa Rica; time spent in fruiting trees by quetzals, bellbirds, toucanets, and robins was unrelated to plumage coloration (Wheelwright 1991). My results agree with those of Wheelwright; fruit-sensitive species included the dull-colored red-eyed vireo (Vireo olivaceus) and grayish mourner (Rhitipterna simplex) as well as the conspicuous white-fronted parakeet (Pionites leucogaster) and scarlet macaw (Ara macao).

Body size may also affect time spent in fruiting trees, for two reasons. First, large birds, like colorful birds, may be more vulnerable to predation (Howe 1979, Pratt and Stiles 1983). Second, large birds generally require greater amounts of food, which are unlikely to be provided in a small area of forest (Fleming 1991). Larger birds might therefore be expected to move more and spend less time per tree or fruiting area. My results do not, however, indicate that body size greatly affected which species were sensitive to fruit harvest; fruit-sensitive species ranged from the 16 g red-eyed vireo Vireo olivaceus to the 1250 g red-and-green macaw Ara chloroptera. The reasons for this result may lie in the experimental design; despite being a relatively large-scale manipulation of fruit abundance, the area of my experimental treatments (0.6 ha) likely encompassed only a small portion of the home ranges of many or most frugivores in the community (Fleming 1991). Recent work (Westcott and Graham 2000) indicates that even small Mionectes (11 g) flycatchers can have home ranges >28 ha, and larger birds such as parrots and toucans are known to range over many kilometers per day (Bourne 1974, Snyder et al. 1987, Bjork 2000).








The reasons underlying the 68% decrease in visitation times in high-removal sites may relate most to the behavioral attributes of the most numerous group of birds in the plots - parrots, parakeets, and macaws. Eleven of the 41 frugivore species are members of the Pssitacidae, and their visits frequently accounted for >50% of the total communitylevel visit minutes. Throughout the neotropics, parrots rely heavily on palm fruit (Snyder et al. 1987, Abramson et al. 1995, Galetti et al. 1999), so their high abundance in E. oleracea forests is not surprising. Also, parrots move and respond to fruit abundance in groups (Chapman et al. 1989); a group of eight parakeets visiting a plot translated into many more visitation minutes than did a visit by a single individual of another species. Furthermore, pssitacids frequently vocalize upon arrival in a feeding area (Snyder et al. 1987). This behavior may serve to relay information about food resources to other individuals, attracting them to the sites of high fruit abundance and leading to higher visit times there.

Mammals

Both high and low intensities of fruit removal reduced the species richness of fruit-eating mammals in plots. The number of individuals and the time they spent, however, were unaffected by fruit harvest. Howler monkeys, tamarins, and squirrels showed the most significant responses to fruit harvest, apparently avoiding the areas from which fruit had been harvested. Of the three primate species observed in plots, howler monkeys are the most frugivorous at my study site, composing up to 71% of their diet with fruit (Jardim and Oliveira 1997, Pina 1999), whereas tamarins and capuchin monkeys have more mixed diets of insects and fruit (Chapman and Fedigan 1990, Peck et al. 1999). That these large-bodied mammals responded to fruit harvest treatments on a scale of 0.6 ha indicates their sensitivity to small-scale differences in food availability








within their much larger home ranges. A similar result was found by Allen (1997), who found that agouti and tayra responded to differences in availability ofMauritiaflexuosa (Palmae) fruit by removing a significantly greater proportion of fruit from areas with low fruit availability. Fruit abundance not only affects fruit-eating mammal behavior; it can also regulate fruit-eating mammal populations (Foster 1982b, Terborgh 1986b, Bodmer 1989, 1990, Adler 1998, Wright et al. 1999). Intensive harvest of E. oleracea fruit over a large area could therefore have population-level impacts on frugivorous mammals. Bruchid beetles

I presumed that bruchid beetles would have relatively small home ranges and thus would be sensitive to my experimental reduction in their sole developmental food source. Beetles could have responded to reduction in fruit abundance with either an increase or a decrease in the number of eggs oviposited. Increased oviposition would have been expected if competition for seeds existed, so that beetles had fewer seeds on which to oviposit in harvested areas (Siemens and Johnson 1996). Decreased oviposition would have been expected if fruit harvest reduced beetle populations, so that fewer females were available to oviposit (Wright 1990).

Contrary to these predictions, beetle oviposition did not vary among treatments. This lack of response may have several explanations. First, bruchids may actually have larger ranges than initially suspected. Wright (1983) showed that bruchid oviposition rates on fallen fruits were similar at distances within 16 m of the trunk of fruiting trees; only at 100 m did rates decline. These results imply that bruchids search over wide areas for seeds - beetles may thus have traversed the treatment subplots within my plots. A second potential reason that oviposition did not differ among treatments is the timing of the experiment. If fruit harvest limited the number of ovipositing females, this effect








would only begin to take place after 2 months of harvest, because beetles take >60 days to emerge from seeds (S. Moegenburg unpubl. data). As I only monitored the number of eggs on experimental seeds for 2 months, I may have missed a decline in the number of eggs oviposited due to a decline in the number of females, which my have been detectable later. A final reason for the lack of bruchid response may be the possibility that to them, E. oleracea fruit abundance did not decrease. The fruit harvest clearly reduced fruit availability to arboreal frugivores, but not necessarily to terrestrial granivores. Despite the fruit harvest, many seeds were dropped by frugivores who removed the pulp but left the seed intact. These seeds remained available to bruchid beetles, and may have been sufficiently available to prevent seed limitation for ovipositing females.


Implications of Fruit Harvest for Frugivores and Fruiting Plants

Few studies have evaluated the ecological impacts of non-timber resource harvest. Extraction is not limited to tropical fruits; the harvest (and associated ecological impacts) of non-timber resources occurs worldwide. One study in northeastern North America, for example, found that harvest of worms for fishing bait significantly reduced the foraging efficiency and prey base of semipalmated sandpipers (Shepherd and Boates 1999).

Even fewer studies have attempted to determine harvest levels of non-timber

resources that minimize ecological impacts. Using matrix models and life table analysis, Peters (1990) estimated that 80% of the fruit produced by Griasperuviana (Lethycidaceae) trees in Peru could be harvested without affecting regeneration of the species. Like E. oleracea, G. peruviana occurs in high-density stands and produces fruits








eaten by a variety of frugivores. In Peter's study, however, the effects of an 80% fruit harvest on frugivores were not considered.

As shown in the present study and by Shepherd and Boates (1999), resource extraction may impact animal communities in a variety of situations. In the case of tropical fruit, what are the implications of these impacts for frugivores, fruiting plants, and conservation strategies based on fruit extraction?

Of the birds most sensitive to E oleracea fruit harvest, at least one (scarlet

macaw, Ara macao) is considered vulnerable to extinction from other causes (Parker et al. 1996). Several other species recorded in this study, or known to eat E. oleracea fruit in other regions, such as blue and yellow macaws (Ara araruana), green and red macaws (Ara chloroptera), and golden parakeets (Guarouba guarouba) may also be affected by fruit harvest (Sick 1993). This last species which, according to Sick (1993), favors the fruits from E oleracea, is one of the most threatened pssitacids in the Brazilian Amazon (Oren and Novaes 1986). Parrots, parakeets, and macaws throughout the neotropics are threatened by habitat destruction and capture for the pet trade (Snyder et al. 1991); harvest of their foods by people should be added to the list of threats.

Pssitacids were not the only group of frugivores to respond to fruit harvest,

however. Contrary to conventional wisdom (Anderson 1990b, Peters 1992, Anderson et al. 1995), fruit harvest can have impressive impacts on biodiversity, reducing it, in the case of birds, by 25%, and in the case of mammals by 58%. Moreover, harvest substantially affected bird behavior; the birds that did visit high-removal treatments spent 68% less time there. These substantial effects may reflect the importance of E. oleracea








in their diets if, as has been proposed for other palm fruits, it constitutes a "keystone" resource for them (Terborgh 1986b, Peres 2000).

Reduction of frugivore diversity and visitations may cascade into indirect effects on other organisms. In particular, reduction in frugivore activity may affect the plants in this system, such as E. oleracea and Virola surinamensis, whose seeds are dispersed by frugivorous birds (Bourne 1974, DeSteven and Putz 1984, Strahl and Grahal 1991, Hamann and Curio 1999, Loiselle and Blake in prep.). On the one hand, less time spent in fruiting patches may actually increase dispersal of the seeds that do get ingested, because frugivores are forced to move farther in search of food (Pratt and Stiles 1983). On the other hand, less frequent and shorter visits can mean that fewer total seeds are ingested (Davidar and Morton 1986, Sargent 1990), and that many may fall directly below parent trees (Wheelwright 1991). In the case of E. oleracea, frugivores perform a critical function by removing pulp, because seeds cannot germinate without at least some pulp removal (S. Moegenburg unpubl. data). Likewise, dispersal of seeds by frugivores is critical for E. oleracea because lack of dispersal results in high mortality from insect seed consumers under the parent plant (S. Moegenburg unpubl. data).

Most of the frugivores in this study range over many hectares, so fruit harvest at the level of this experiment likely had no, or few, impacts on them other than the visitation patterns observed. Harvest of E. oleracea fruit occurs, however, over a very large area (up to 10,000 km2 of the Amazon estuary, Calzavara 1972) by many thousands of people. Furthermore, extraction is increasing in response to higher market prices for fruit and the emergence of a forest conservation strategy known as extractive reserves. These reserves, encompassing over 22,000 km2 mostly in the Brazilian Amazon, are








collectively managed by local, forest-dwelling residents for the long-term sustainable use of forest resources, including fruits and other non-timber products (Fearnside 1989, Allegretti 1990). What might be the effects on frugivores across this larger scale at which E. oleracea fruit is harvested?

While my study shows that intensive fruit harvest at a small scale (0.6 ha)

negatively affects frugivores, it has limitations in terms of predicting the impacts of E. oleracea fruit harvest at the larger scales at which people currently practice extraction. Fruit harvest across the home ranges of frugivores could have several types of effects. Specifically, in response to decreased fruit availability throughout their ranges, frugivore numbers may decrease through a functional response, if frugivores vacate the area, or through a numerical response, if frugivores suffer lower reproductive success. Alternatively, they may show neither a functional nor numerical response but rather simply switch diets. Evidence suggests that all three responses may occur, each by different members of the frugivore community. In other systems, for example, frugivorous birds apparently respond functionally to fruit availability by tracking ripe fruit over space and time. Such tracking behavior has been shown across altitudinal gradients (e.g., Loiselle and Blake 1991), latitudinal gradients (Martin and Karr 1986), and agricultural landscapes (Rey 1995). In systems in which frugivores cannot track ripe fruit availability (e.g., non-volant mammals on Barro Colorado Island, Panama), response to decreased fruit abundance is numerical, with frugivores suffering either mortality (Foster 1982a, Wright et al. 1999) or lowered reproductive success (Adler 1998). Finally, some partial frugivores and omnivores (usually the majority of fruit-eaters in the








community) may switch to other diets when fruit becomes scarce (Loiselle and Blake 1993, Pina 1999).

How might the responses of frugivores to large-scale fruit harvest be ascertained? Clearly, a controlled experiment on an adequate scale would be difficult. A potential route might be in an extractive reserve in which fruit is harvested and in which the level of human community organization would allow the orchestration of a large-scale quasiexperiment. In the Cajari River Extractive Reserve in Amapi State, Brazil, for example, several dozen families extract E. oleracea fruit from several thousand hectares. In such a setting it may be possible to request families to harvest from designated areas and not from others, which would establish areas of high and low fruit availability, in which frugivores could be monitored. As in this study, frugivores should be surveyed both preand post-harvest to increase the power of such a quasi-experiment.

Such research needs to be done, on E. oleracea and other species of harvested

fruit. Until it is, however, the data available from my study and others indicate that fruit harvest does affect frugivores, so it should be limited if biodiversity and ecological processes are to be conserved. Forests devoted to fruit extraction are apparently not equivalent, from the point of view of frugivorous birds and mammals, to forests from which fruit is not extracted. Fruit extraction can potentially play an important role in tropical forest conservation, as a long-term renewable source of income. However, forests dedicated to fruit extraction should not be seen as substitutes for non-harvested forests; rather, they should be part of conservation plans that also include non-harvested forests.













CHAPTER 4
FALLING FRUITS AND FEEDING FISHES: EUTERPE OLERACEA FRUITS IN
AMAZONIAN FLOODPLAIN FOREST NUTRIENT AND FOOD CYCLES


Introduction

Approximately 2 - 7% of the Brazilian Amazon exists as floodplain forest, which inundates via two hydrological processes of whitewater, clearwater, and blackwater rivers (Ducke and Black 1953, Sioli 1966, Pires 1974, Prance 1978, Goulding 1980). The first process occurs throughout the basin during the wet season, when rainwater swells rivers, causing them to overflow their banks as much as 13 meters. This high water phase in socalled "igap6"' forests can persist from 3 - 11 months per year (Pires 1974). The second, shorter-phase process is caused by tides from the Atlantic Ocean, which push river water upstream, causing it to spill into forests to depths of 2 - 3m. These flood pulses in "varzea" forests last for about six hours and occur approximately twice per day.

When rivers flow into forests, fishes follow, finding detritus, invertebrates, and fruit, on which to feed. Fruits serve as food for at least 200 of the 2,500 - 3,000 species of Amazonian fish (Marlier 1967, Goulding 1980, Waldhoff et al. 1996), some of which consume little else during the rivers' high water phase. Some species, such as Colossoma macropomum, have evolved special dentition that allows them to crush even very hard pericarps (Goulding 1980). Top-swimming fishes gulp floating fruits and seeds, such as SAlternative definitions for varzea and igap6 exist. According to Pires (1974), for example, varzea forests are those flooded by turbid, or sediment-laden, water, while igap6 is forest flooded by blackwater and clearwater rivers. Neither set of definitions is more correct, either ecologically or biogeographically (Gouldming 1980). My preference for the one used above reflects the definition used by Furch (1997) and many local people in the Amazon.








those from the rubber tree, Hevea brasiliensis, while bottom dwellers, such as the catfish Lithodoras dorsalis, eat fruits and seeds that sink (Kubitzki and Ziburski 1994, Goulding 1980). Small fish that cannot ingest whole fruits bite off pieces of pulp from submerged fruits (S. Moegenburg pers. obs.).

Just as in unflooded forests (Hladik and Hladik 1969), not all fruit that falls in flooded forests is immediately consumed. Rather, uneaten fruits remain floating or submerged (Kubitzki and Ziburski 1994) and begin to decompose. During this phase, fruits may play several roles in the floodplain ecosystem. First, they may continue to serve as food for fruit-eating fishes (Araujo-Lima et al. 1986, Henderson and Crampton 1997). Second, decomposing pulp and seeds may furnish nutrition for scavenging and detritous-feeding fish and invertebrates (Irmler 1975, Henderson and Walker 1986, Junk and Robertson 1997). Third, decaying fruits and seeds may release nutrients into the water to be taken up by phytoplankton or plants. Blackwater rivers and the forests they flood are extremely nutrient poor (Sioli 1968, Setaro and Melack 1984, Furch 1997), presumably recycling most nutrients through the 5,000 kg ha-' of leaf litter that falls per year and decomposes (Franken et al. 1979, Furch and Junk 1997b). Nutrient cycling through the 469 kg ha-' yr' of fruits that fall in igap6 forests (Waldhoff et al. 1996) has not been well studied.

In igap6 forests of the eastern Amazon, fruit production at many sites is

dominated by the palm, Euterpe oleracea, which produces one-seeded, spherical drupes approximately 1 cm in diameter (Roosmalen 1985). E. oleracea is hydrophilic, occurring primarily along river margins and in both whitewater and blackwater varzea and igap6. Unlike many other floodplain diaspores, however, neither the fruits nor seeds of E.








oleracea float (Roosmalen 1985). Many animals, including primates (Roosemalen 1985), birds (Moegenburg and Levey in prep.), and fish (Goulding 1980, Waldhoff et al. 1996) consume E. oleracea fruits. In addition, people harvest the fruits to make a nutritious drink (Strudwick and Sobel 1988). In intensively managed and harvested areas, fruit removal can approach 13,000 kg ha' yr-' (Mufiiz-Miret et al. 1996). Fruit extraction may thus remove a significant source of food for fish and invertebrates, and a source of nutrients for floodplain forests.

I studied two aspects of E. oleracea fruits in floodplain systems. First, to evaluate fruits as nutrient sources in floodplain forests, I determined the nutrient content of fresh E. oleracea fruit and the total amount of N and P in E. oleracea fruits per area of forest. Second, to test the effect of human harvest of E. oleracea fruit on aquatic animals, I compared the species richness and abundance of aquatic invertebrates and fish in areas subjected to high-intensity fruit harvest, low-intensity fruit harvest, and no fruit harvest in an experimental manipulation.

Methods

Study Species and Sites

The cespitose E. oleracea ranges throughout Amazonian South America, along the Pacific coast of Colombia and Ecuador, and in Trinidad (Henderson 1995). Individual genets (hereafter "trees" for simplicity) contain up to 25 slender stems that reach heights of 30 m (Henderson 1995). Reproductive stems produce infructescences bearing several thousand purple-black fruits. Fruits contain approximately 37% water (Waldoff et al. 1996). E. oleracea pericarp contains 8.1 and 15.3 % crude protein and fat, respectively. E. oleracea "seed" (presumably seed + endocarp) contains 7.2 2.5, and

8.1% crude protein, crude fat, and crude fiber, respectively (Waldhoff et al. 1996).








This study was carried out from June - September, 1998 at the 33,000 ha Estaqdo Cientifica Ferreira Penna (Ferreira Penna Scientific Station; 142'30" S, 510 31 '45" W), operated by the Museu Paraense Emilio Goeldi of Belem and located within Caxiuan, National Forest in the municipality of Melgaqo, Para State, Brazil (Fig. 4-1). Average rainfall is 2,500 - 3,000 mm, mean annual temperature is 26C, and mean annual relative humidity is 85%. The vegetation is evergreen humid rainforest, and the majority of the station is non-flooded, terra firme forest (Lisboa et al. 1997). Approximately 3,300 ha, mostly along the blackwater Bay of Caxiuani, is extremely low-lying forest that inundates when the river levels rise during the rainy seasons (December - May) and high tides. Water depths reach their maxima (ca. 1 m) in May.

Much of this floodplain forest is dominated by E. oleracea, with Virola

surinamensis (Myristicaceae) and Pterocarpus santalinoides (Fabaceae) also common (Ferreira et al. 1997). Although there are no homes and no current management of this forest, several lines of evidence suggest that the area of floodplain forest dominated by E. oleracea (hereafter "palm forest") resulted from past human management. First, areas of highest E. oleracea density occur closet to the river margin, where human dwellings tend to sit. Second, local people commented that the names of some of the sites (e.g., Moreira, Fazenda) reflect previous uses or names of inhabitants. E. oleracea fruit begins to appear in these sites in May and persists until September. In most years people occasionally extract E. oleracea fruit from these areas; however, in 1998 local inhabitants respected my request to refrain from extraction.

I worked in four blackwater river flooded forest sites (hereafter "palm forest") located approximately 1 km apart along the bay. The water level in all of the sites




Full Text

PAGE 1

FRUIT-FRUGIVORE INTERACTIONS IN EUTERPE PALM FORESTS OF THE AMAZON RIVER FLOODPLAIN By SUSAN MARIE MOEGENBURG 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 2000

PAGE 2

For Ben and Peter

PAGE 3

ACKNOWLEDGMENTS This study was possible only with the generous support, hard work, and financial assistance of many people and institutions. I am especially grateful to my advisor, Doug Levey, for his scientific advice and enduring support through the years. I have learned much about tropical ecology and conservation from other committee members Jack Putz, Colin Chapman, and Richard Bodmer, been challenged to consider multiple scales by Buzz HoUing, and benefited from "brainy" discussions with Katie Sieving. I also thank Lauren Chapman for advice and for serving at my defense. I am fortunate to have interacted with so many friends and colleagues at the University of Florida. For lively discussions I especially thank Sophia Balcomb, Kevin Baldwin, Maria Ines Barreto, Paula Cushing, Scot Duncan, John Paul, Roberto Porro, Luciano Verdade, Dan Wenny, and Amy Zanne. Special thanks also go to Juan Posada for abiding support and love. For help of various sorts I thank Richard Fethiere, Bree Darby, and Mark Stowe. My life has been enriched tremendously through my experiences in Brazil. I am grateful to Jose Fragoso and Kirsten Silvius for infecting me with Amazon fever. Data collection would not have been possible, or as enjoyable, without the help from Batista Ferreira, Rick Newman, AnaLucia Castelo Branco Pina, the staff of the Esta9ao Cientifica Ferreira Penna, especially Pao and Madruga, and the fruit harvesters Clesio, Joao Domingo, and Zeca. I thank Mario Hiraoka for the use of his property along the Rio Maracapacu-Mirim, and Patricia and Mauricio Almeida, Michael and Colleen Collins, iii

PAGE 4

and Beto Verissimo for accommodations in Belem, Friends at Imazon, especially Mark Cochrane and Chris Uhl, helped by offering good advice and shaded hammocks at crucial times. Many people at the Museu Paraense Emilio Goeldi also facilitated this project. I especially thank Mario Jardim and Pedro Lisboa for logistical support at key moments. I am indebted to David Oren for the use of mist nets. My identification skills of birds improved greatly from the expertise of Jose Maria Cardoso da Silva and Mario Cohnhaft. I also thank Bill Overal for the use of laboratory space and good music during my aquatic invertebrate identifications, and Mauricio Camargo for help with the fish identifications. I appreciate the institutional support from the Department of Zoology, University of Florida, and from the Brazilian organizations IBAMA (Brazilian Institute for the Environment), and CNPq (National Institute for Research). I was able to carry out this research only with generous financial support from the Dickinson family, Tinker Foundation, Sigma Xi Scientific Society, Lincoln Park Zoo, American Bird Conservancy, National Geographic Society, and the U.S. Environmental Protection Agency. Finally, I express my gratitude to my family and friends, whose support and love I carried with me to the rainforest. iv

PAGE 5

TABLE OF CONTENTS page ACKNOWLEDGMENTS iii ABSTRACT viii CHAPTERS 1 GENERAL INTRODUCTION 1 2 RESPONSES OF VEGETATION, BIRDS, AND MAMMALS TO MANAGEMENT FOR EUTERPE OLERACEA 6 Introduction 6 Methods 9 Study System 9 Vegetation and Microclimate Sampling 15 Bird Censusing 16 Avian Perch Use and Availability 17 Mammal Trapping 18 Data Analysis 19 Vegetation and microclimate sampling 19 Bird censusing 19 Avian perch use and capture times 22 Mammal trapping 22 Results 22 Vegetation and Microclimate 22 Bird Censusing 27 Avian Perch Use and Capture Times 34 Mammal Trapping 37 Discussion 39 Changes in Flora and Fauna 39 Mechanisms of Bird Community Change 42 Comparison to Alternative Forest Management Systems 44 Recommendations and Conclusions 45 V

PAGE 6

3 LINKING FRUIT AND FRUGI VORE ABUNDANCE : EXPERIMENTAL EVIDENCE FROM AMAZONIAN BRAZIL 47 Introduction 47 Methods 51 Study Species 51 Study Sites and Plots 52 Fruit Availability 54 Parrot Use of Palm Forest Versus Non-palm Forest 56 Frugivore Responses to Experimental Fruit Harvest 57 Frugivore surveys 57 Fruit removal 59 Statistical Analyses 60 Results 61 Parrot Use of Palm Forest Versus Non-Palm Forest 61 Frugivore Responses to Experimental Fruit Harvest 63 Fruit availability 63 Birds: community-level responses 63 Birds: species-level responses 69 Mammals 72 Bruchid Beetles 74 Discussion 75 Frugivore Responses to Fruit Harvest 75 Birds community level 75 Birdsspecies level 78 Mammals 81 Bruchid Beetles 82 Imphcations of Fruit Harvest for Frugivores and Fruiting Plants 83 4 FALLING FRUITS AND FEEDING FISHES : EUTERPE OLERACEA FRUITS IN AMAZONIAN FLOODPLAIN FOREST NUTRIENT AND FOOD CYCLES 88 Introduction 88 Methods 90 Study Sites and Plots 90 Fruit Production 93 Nutrient Loss from Fruits 95 Responses of Aquatic Animals to Fruit Harvest by People 96 Statistical Analyses 97 Results 9g E. oleracea Fruit Production 98 Nutrient Loss from Fruits 98 Responses of Aquatic Animals to Fruit Harvest by People 100 Discussion 203 Fruit Production, Nutrient content, and Nutrient loss 105 Aquatic Animal Responses to Fruit Harvest by People 107 vi

PAGE 7

5 SPATIAL AND TEMPORAL VARIATION IN HYDROCHORY IN AMAZONIAN FLOODPLAIN FOREST 1 1 1 Introduction Ill Methods 114 Results 116 Discussion 121 6 GENERAL CONCLUSIONS 131 Ecological Considerations 131 Conservation Considerations 133 APPENDICES A DIET CLASSIFICATION OF SPECIES OBSERVED IN E. OLERACEA FOREST 136 B RESULTS OF REPEATED-MEASURES ANOVAS 139 LIST OF REFERENCES 144 BIOGRAPHICAL SKETCH 163 vii

PAGE 8

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 FRUIT-FRUGIVORE INTERACTIONS IN EUTERPE PALM FORESTS OF THE AMAZON RIVER FLOODPLAIN By Susan Marie Moegenburg December 2000 Chair: Douglas J. Levey Major Department: Zoology While fruit production in tropical forests varies naturally over space and time, its abundance and distribution in some forests is altered by management to increase fruit production and by harvest. In many tropical areas, forest fruits constitute a substantial portion of human diets and are important market commodities. The effects on forests of fruit management and harvest, however, remain little known. I studied enrichment and harvest of fruit from the palm Euterpe oleracea in palm-dominated (> 300 trees ha"') floodplain forests, estimated to cover over 10,000 kmof the Amazon River delta. In this area people harvest up to 9,000 kg of E. oleracea fruit ha"' yr ', reducing fruit availability to arboreal, terrestrial, and aquatic frugivores, and removing nutrients from the system. Forests managed for£. oleracea were of lower, more open stature, dominated by E. oleracea, and nearly devoid of understory vines, lianas, and small trees, as compared with control forests. Also, the managed forest bird community was dominated by fruitand seed-eaters, whereas the control forest community contained more insect-eaters. I viii

PAGE 9

also evaluated responses of birds and mammals, seed-feeding beetles, and aquatic invertebrates and fish, to fruit harvest. Four 1.8 ha circular plots were divided into three sections, which were assigned one of three ripe fruit harvest treatments: high (-90% removal), low (-50% removal), and control (0% removal). The community of frugivorous birds and mammals showed significant responses to the high fruit harvest treatment. There were 25% fewer bird species, 58% fewer mammal species, 29% fewer bird individuals, and avian frugivores spent 68% less time in high removal plots compared to control plots. The community of frugivorous birds showed no significant responses, however, to low removal. Also, the abundance of fruit-eating fish, but not invertebrates, was reduced in high-fruit removal treatments. Fruit decomposition experiments demonstrated rapid loss of both nitrogen and phosphorous from submerged fruits. High levels of E. oleracea fruit harvest thus appear to have substantial ecological effects on various trophic levels and ecosystem processes, whereas low levels have few measurable impacts. These results provide an initial ecological basis for determining sustainable levels of fruit harvest. ix

PAGE 10

CHAPTER 1 GENERAL INTRODUCTION Here and there, shooting above the more dome-like and sobre trees, were the smooth columnar stems of palms, bearing aloft their magnificent crowns of finely-cut fronds. Amongst the latter the slim assai-palm was especially noticable, growing in groups of four and five; its smooth, gently curving stem, twenty to thirty feet high, terminating in a head of feathery foliage, inexpressibly light and elegant in outline. (Bates 1864, p. 3) Among the most conspicuous features of neotropical rainforests are the high proportion (70 93%) of tree species that produce fleshy fruits, the seeds of which are dispersed by vertebrates (Gentry 1982, Jordano 1992), and the high proportion of vertebrates that include fruit as a major component of their diet (e.g., 35 39% of bird species) (Karr et al. 1990). This high diversity of interactions among fruits and frugivores has inspired hypotheses that the abundance and distribution of fruit determine the abundance and distribution of frugivores, and, conversely, that the behavior of fruiteating animals influences the abundance and distribution of fruit producing plants (McKey 1975). Despite a legion of studies addressing these hypotheses, however, the importance of fruits and frugivores to the ecology and evolution of one another, relative to other biotic and aiotic factors, is far from understood (Levey and Benkman 1999). Increasingly in tropical forests, natural patterns of fruit abundance and distribution are modified by human activities, including fruit extraction for sale and consumption. In Iquitos, Peru, for example, hundreds of species of fruits and seeds are harvested from wild-occurring trees and regulariy sold in markets (Vasquez and Gentry 1

PAGE 11

1989). Not only is fruit extraction from forests a widespread, and sometimes intensive, practice today, but anthropological (Balee 1988) and archeological (Roosevelt et al. 1996) evidence indicate that, in the Amazon basin, fruit harvesting and forest management have occurred for millenia. The objective of this work is to evaluate how contemporary fruit management and harvest by humans affect fruit-frugivore interactions in Euterpe oleracea dominated rainforests of the eastern Brazilian Amazon. This multi-stemmed palm is the source of two of the most economically and socially important non-timber products of the region: heart-of-palm and fruit, locally known as a^ai (Richards 1993, Anderson 1988). Owing to its economic value, natural abundance, and ease of management (Anderson 1990a,b), E. oleracea is featured in several extractive reserves areas designated by the Brazilian government for the long-term, sustainable management of both biodiversity and forest products (Feamside 1989). To achieve these multiple goals, a balance must be struck between levels of management and extraction that maximize profits and those that minimize impacts on biodiversity. An ancillary goal of this project, therefore, is to provide data on which to base management and extraction guidelines to help extractive reserves meet their goals. Like many Amazonian palms (Kahn 1991), E. oleracea is hydrophilic and rarely grows in non-inundated sites. I worked in two types of floodplain forests dominated by E. oleracea. Data collection for Chapters 2 and 5 took place in sediment-laden Whitewater river floodplain forest along the Rio Maracapacu-Mirim, Para State, Brazil. Study sites lie approximately 80 km from the port of Belem. Data collection for Chapters

PAGE 12

3 3 and 4 took place in nutrient-poor blackwater river floodplain forest within the Caxiuana National Forest, Para, Brazil. The high diversity of tree species characteristic of tropical forests (Gentry 1982) means that each species tends to occur at low densities. Low conspecific tree densities limit the amount of fruit, and therefore profits, that can be extracted from a given area of forest. To increase commercial tree densities and profits, people therefore manage forests to augment the production of fruit-producing trees. In the case of E. oleracea, management includes enriching forests through planting and re-locating E. oleracea seeds and seedlings, removing understory plants perceived to compete with E. oleracea, and removing canopy trees that shaded, oleracea and may limit fruit production (Anderson 1990a, b). In Chapter 2, 1 examine the effects of these management activities on forest structure and bird communities by comparing managed and non-managed forests. In particular, I measure canopy height, canopy density, stem density, basal area, and stem composition, in five managed and five non-managed forests. In addition, I census understory birds with mist-nets in the same ten forests. The differences in the structure and composition of both the vegetation and the bird communities between the two forest types are compared to vegetation and birds in forests subjected to other types of management, such as selective logging and agroforestry. Forest management, by definition, alters forest structure and composition. Fruit extraction, on the other hand, is frequently viewed as having no ecological impacts (e.g., Peters 1992). This view is not supported, however, by studies demonstrating that fruit and frugivore abundance are linked across space and time (Levey 1988, Loiselle and Blake 1991). These correlative studies, however, have limited power to inform

PAGE 13

4 predictions of how faigivores may respond to fruit harvest. In Chapter 3, 1 take advantage of the E. oleracea system high-density production of easily-removed fruit that is eaten by a diverse community of frugivores to conduct what is apparently the largest experimental manipulation of fruit abundance, to date. In replicated plots I mimic subsistenceand commercial -level fruit extraction and compare the species richness, the number of individuals, and visit lengths of fruit-eating birds and mammals. I also monitor the oviposition behavior of seed-feeding beetles. The results of this experiment not only confirm that fruit abundance drives fruit-eating bird abundance, but also that fruit extraction by humans affects the abundance, distribution, and behavior of fruiteating animals. While fallen fruits and seeds of trees growing in non-inundated forests become available to terrestrial fruitand seed-eaters (DeSteven and Putz 1984, Bodmer 1989, 1990, Levey and Byrne 1993, Chapman and Chapman 1996), fallen fruits and seeds in inundated forests also become available to fruitand seed-eating fish and aquatic invertebrates (Goulding 1980). Fallen fruits and seeds not consumed by these organisms may decompose and release nutrients back into their cycles. In Chapter 4, 1 investigate these two aquatic aspects of E. oleracea fruit-frugivore interactions. First, to evaluate fruits as nutrient sources in floodplain forests, I determine the nutrient content of fresh E. oleracea fruit, the loss of nutrients over time from submerged fruits, and the total amount of N and P in E. oleracea fruits per area of forest. Second, to test the effect of human harvest of E. oleracea fruit on aquatic animals, I compare the species richness and abundance of aquatic invertebrates and fish in the same experimental plots used in Chapter 3.

PAGE 14

5 In many fleshy fruit-producing plants, the dispersal of seeds by fruit-eating vertebrates is only the first of several dispersal events (Wenny 1999). For species that grow in inundated forests, subsequent dispersal of seeds is often accomplished by water In Chapter 5, 1 explore the dispersal of seeds by water in tidally-flooded forests, where water levels fluctuate according to daily and lunar cycles. I measure water depth as a function of lunar phase and distance to streams, and quantify the effect of water depth on dispersal of seeds of different species. Based on the results, I argue that water, like other agents of dispersal, is unpredictable over time and space and that this unpredictability selects for multiple mechanisms of seed dispersal within and among fruits of individuals and species.

PAGE 15

CHAPTER 2 RESPONSES OF VEGETATION, BIRDS, AND MAMMALS TO MANAGEMENT FOR THE NON-TIMBER FOREST PRODUCT EUTERPE OLERACEA Introduction Extraction and marketing of non-timber forest products (NTFP) can promote forest conservation and human well-being in several ways (Feamside 1989, Peters et al. 1989, Panatoyu and Ashton 1992, Peters 1992, Plotkin and Famolare 1992, Arnold and Perez 1998). First, harvest of NTFP, such as fruits, seeds, resins, barks, and fibers, tends to be less destructive to forests than logging. Second, harvest of NTFP provides income for local, forest-dwelling people, furnishing a profitable alternative to forest clearing. Third, the economic benefits of NTFP harvest are realized primarily in the long-, rather than the short-term, term, thereby motivating long-term forest management and planning. Such benefits of NTFP harvest have been embraced by conservationists and form the foundation of extractive reserves, such as those in the Brazilian Amazon (Feamside 1989, Alegretti 1990, Nepstad and Schwartzman 1992, Mattoso and Fleischflesser 1994, Alves 1995). One limitation to profitable harvest of many tropical NTFPs is the low density at which source species often occur. Low densities of source plants increase the time required to search for individuals, to travel among individuals, and to transport goods (Peters 1992, Salafsky et al. 1993). To decrease these collecting costs and to increase the economic value of their property, many landowners increase productivity through so6

PAGE 16

7 called enrichment management "increasing the density of crop-producing species by planting in their natural habitats" (Schulze et al. 1994, p. 582). Such enrichment has apparently been practiced for millenia (Balee 1989, Gomez-Pompa and Kaus 1992, Pinedo-Vasquez and Padoch 1996) and may have created extant "oligarchic forests" (sensu Peters et al. 1989), in which NTFP species occur at high densities Today, enrichment management is encouraged in various types of forests, including indigenous lands (McCann 1999), extractive reserves (Kainer et al. 1998), privately owned areas (Anderson and loris 1992, Peters 1996), and others (Schulze et al. 1994, Pena 1996, Ricker etal. 1999). The effects on forests of NTFP enrichment are distinct from the effects of NTFP harvest, yet have received much less attention from those interested in the development of sustainable NTFP use (Bawa 1992). Enrichment often occurs in secondary forests and often accompanies other management activities, such as logging, liberation treatments, and planting of timber species (Schulze et al. 1994, Salick et al. 1995, Bawa and Seidler 1998). Enrichment management can significantly augment the density of species used by humans. For example, Salick et al. (1995) studied the responses of both timber and NTFP species to logging and liberation (i.e., canopy opening) treatments in primary rainforest in Nicaragua. NTFP and timber species richness and density were higher both one year and nine years post-logging, suggesting that logged forests allowed to regenerate can be enriched for economically-valuable species. This is also the case in the Amazon River floodplain, long inhabited by humans who removed valuable timber trees and enriched for a variety of marketable species (Anderson 1990b, Anderson et al. 1995). Similar systems are known from diverse forests in Asia, Europe, and Latin America

PAGE 17

8 (Pinedo-Vasquez and Padoch 1996, Shankar et al. 1996, Freese 1997, McShane and McShane-Caluzi 1997, WoUenberg and Ingles 1998). While these studies illustrate that enrichment management increases the density of harvested species, less is known about the capacity of NTFP -enriched forests to conserve biological diversity and ecosystem functions. The viability of NTFP enrichment and harvest as a conservation tool depends not only upon sustaining production of forest products, but also upon sustaining biodiversity in enriched forests. I evaluate the ecological impacts of enriching Amazonian floodplain forests with Euterpe oleracea, known locally as a9ai. This native palm is the source of two important NTFPs: palm heart and fruit. The E. oleracea system represents a complex interaction among palm trees, humans, and animal communities. Enrichment with E. oleracea likely alters forest structure, similar to management for timber species; however, it also creates high density stands of a fruit resource used by a suite of frugivorous animals. I focus my comparisons o^E. oleracea-QxmdhQd and control forests on vegetation structure and composition, and understory bird and mammal communities. Birds and mammals are known to respond to variation in vegetation structure and associated microclimatic fluctuations across natural environmental gradients and human disturbances (MacArthur and MacArthur 1961, Karr and Freemark 1983, Johns 1988, Thiollay 1992, Mason 1996. Greenberg et al. 1997, Bawa and Seidler 1998, Putz et al. 2000). I chose to study the E. oleracea system for several reasons. In contrast to upland areas in the Amazon basin, where conversion of forest to pasture or agriculture captures the highest short-term profits, management of this floodplain for forest products is considered the most rational and profitable land-use option (Anderson 1990b, Anderson

PAGE 18

and Ions 1992, Anderson et al. 1995, Hiraoka 1995). Indeed, demand for floodplain forest products, especially palm fruit and palm heart, is high and increasing (Pollack et al. 1995). Approximately 40% of the 25,000 km^ Amazon estuary is enriched primarily for E. oleracea production and several areas within Brazil's 22,000 km^ system of extractive reserves feature E. oleracea as an NTFP on which residents' incomes are based (Mattoso and Fleischflesser 1994, Alves 1995). Methods Study System The range of the pinnately leaved, multi-stemmed E. oleracea includes the Brazilian States of Para, Amapa, Tocantins, and Maranhao, the Pacific coast of Colombia, and northern Ecuador, Trinidad, Venezuela, and the Guianas (Henderson 1995). Throughout the Amazon River estuary, E. oleracea , which I also call by the colloquial name "afai", grows in monodominant stands in floodplain forests, known as varzea and igapo. Individual trees contain up to 25 stems (Henderson 1995), which become reproductive at heights of 4 30 m, depending on growing conditions (Strudwick and Sobel 1988). E. oleracea produces 1 cm diameter fleshy -fruited drupes (Roosmalen 1985), which are consumed by many species of frugivores (Chapter 3). E. oleracea is among the most widely-used forest species in the Amazon River estuary, where its products are central in the diet, culture, and economy of the inhabitants (Bates 1864, Anderson 1988, 1990b, Strudwick and Sobel 1988, Anderson et al. 1995, Hiraoka 1995, Muniz-Mirit et al, 1996). From E. oleracea people harvest fruits, which are produced into a popular beverage, and the apical meristem, from which heart-of-palm

PAGE 19

10 is extracted. E. oleracea fruits have been harvested for centuries (Bates 1864), but largescale commercial harvest of E. oleracea heart-of-palm began relatively recently, and coincided with overharvest of populations of E. edulis in southern Brazil (Hiraoka 1995). During the late 1960s, as populations of E. edulis declined to the point of low returns, palm heart factories moved from the Atlantic forest region to the Amazon estuary. At the same time, the market for sugar cane, grown for decades by the estuary's rural inhabitants, collapsed. Rural workers turned to extracting palmito from wild, highdensity a9ai stands and managing their own forests for afai production. This process was fueled by the growing number of palmito processing factories in the region and by the expanding market for a9ai fruit, particularly in Belem. Thus began what Hiraoka (1995) has termed the "a9aization" of the estuary, in which management that favored E. oleracea became the dominant land use strategy. Selling heart-of-palm and fruit from E. oleracea are profitable ventures. On the 15 km^ Combu Island, near Belem, for example, E. oleracea palm heart and fruit sales comprise 85% (US$3,400) of the income for the majority of the island's 97 households (Anderson and loris 1992). Families selling a9ai fruit can earn up to US$235 per hectare per year (Peters 1992). During the months of low a9ai fruit production (November March), income is supplemented by selling palm hearts. Throughout the region, palm heart sales generate approximately US$300 million per year and employ nearly 30,000 people (Clay 1997). The low suitability of estuarine floodplain for other uses and the relative profitability of a9ai products make a9ai management and harvest the preferred land use in the region (Hiraoka 1995).

PAGE 20

11 A^ai-enriched forest stands originate in one of two ways. Plots formerly devoted to swidden agriculture or sugar cane can be transformed to a9ai agroforests, after a fallow period of 5 7 years (Hiraoka 1995), or mature mixed forest can be gradually converted to a?ai -dominated forest. Both processes involve planting of a?ai seeds and seedlings, protection and thinning of afai plants, and removal of non-harvested plants that might compete for light and nutrients (Anderson and loris 1992). In addition, plants that impede walking through the forest, or in which snakes are thought to hide, are removed. According to Hiraoka (1995), a(?ai production reaches peak levels 6-10 years after this "tolerant" (sensu Anderson 1990b) management begins. Acai -enriched forests contain other native and exotic harvested species such as tapereba {Spondias mombin), rubber {Hevea brasiliemis), andiroba (Carcpa guiammis), bacuri (Rheedia macrophylld), cacao {Theobroma cacao), coconut (Cocos nucifera), guava (Psidium guajava), and mango {Mangifera indica) (Anderson 1990b, Anderson et al. 1995). Management for a^ai and other land uses results in a mosaic of forest types that includes "home gardens", afai-dominated forest, and mature, non a^ai-enriched forest (hereafter "control") from which other forest products are gathered. Streams often form the boundaries between enriched and control forests, and forests dedicated to a?ai management are easily distinguished from control forests. For example, in enriched forests a9ai genets almost always contain remnant stumps of stems harvested for heartof-palm. Another clue is discarded, harvested infructescences from which fruits have been removed. Tree species dominant in control forests include A/awr/7/a Jlexousa, Virola surinamemis, Pterocarpus officinalis, and Rhizophora mangle (Anderson et al. 1995).

PAGE 21

12 This research took place during 1997 in the southeastern portion of the Amazon estuary, Para State, Brazil (r45'56" S, 48°57'41" W; Fig. 2-1). The hundreds of islands comprising this region are characterized by extremely low-elevation, low-pH Entisol soils, low plant species diversity, and average temperature and precipitation of approximately 25 C and 3000 mm, respectively (Hiraoka 1995). Rainfall is seasonal, with highs from January to May, and lows from June to November. — Study Islands Figure 2-1. Study site within Brazil, and study islands within site.

PAGE 22

13 Study sites were located in floodplain forest near the mouth of the Tocantins River, which reverses its flow twice per day with tidal influx. During the full and new moons, and more regularly during the rainy season, the tidal influx swells the river enough to overflow its banks and flood the forests. The restriction of agricultural possibilities by these floods has led to management for a?ai production as the primary land use in the area. Study sites lie within 20 km of the nearest city, Abaetetuba, and within 80 km of the capitol city Bel em (pop. -169,807,000). Most afai fruit and palm heart extracted from this area are sold in Abaetetuba, with occasional shipments selling in the Belem market. Afai fruit is the primary commodity sold by the rural people, with occasional sales by most producers of palm heart. Production ranges from 8,640 13,734 kg fruit/ha/year, which totals approximately US$6,594 14,936 per ha (Muniz-Miret et al. 1996). Data were collected in ten forest stands: five enriched primarily for a^ai tree production through enrichment (hereafter "enriched"), and five not enriched for afai (hereafter "control"). These forests were identified through conversations with local people. Enriched forests had been so managed for > 20 yr and were approximately 4-5 ha in size. The sizes of the control forests ranged from 10 100 ha. Although some of the enriched and control stands were adjacent to each other (Fig. 2-2), the locations at which data were collected within the forests were all > 500 m apart. Because afaienriched stands tend to be close to homes, which tend to be close to rivers, most enriched stands were closer to rivers than were control stands. Aside from differing in management activities, the ten forests were assumed similar in soils, topography, and vegetation because of their proximity to one another in the floodplain. Other studies have shown

PAGE 23

Figure 2-2. Relative locations of five a9ai-enriched and five non-enriched stands along the Maracapacu-Mirim River.

PAGE 24

15 high similarity in forest composition among floodplain forest sites in this region (Anderson et al. 1995). Control forests have long been subject to subtle management, in which people selectively remove trees for timber and remove other products, such as palm leaves for thatch, bark for medicines, and fruits for food. Nevertheless, they differ from a9ai-enriched forests by not being managed specifically for a9ai production. Vegetation and Microclimate Sampling Vegetation structure and composition were evaluated in each of the 10 stands along a 50 X 1 m transect. Finding homogenous patches in the varzea, which is riddled with streams and permanently inundated low-lying areas, and where management activities themselves are patchy, proved difficult. Fifty meters was chosen as the minimum transect length that allowed a characterization of the vegetation while remaining within a homogeneous stand of either enriched or control forest. Within such stands, transect starting points and orientations were located randomly. Along each transect, data were collected on vegetation structure and composition that could be directly affected by management activities and that might affect vertebrate communities. I collected two types of data at 5 m intervals, for a total of ten points per transect. Canopy cover above 1.5 m was measured using a spherical densiometer, an instrument for measuring forest overstory density (Lemmon 1957). Canopy height was visually estimated after practicing with a clinometer at known distances from trees. Along the entire length of each transect, I counted and measured the dbh (diameter at breast height) of all stems above 1 .5 m. Stems were recorded as belonging to one of the following categories: E. oleracea, Mauritia flexuosa (Palmae); Raphia taedigera (Palmae); Montrichardia linifera (Araceae), non-woody vine; woody liana; or hardwood stem. The

PAGE 25

16 palms and M linifera were counted separately from the other types of stems because they lack woody bark and may therefore represent different foraging substrates for vertebrates than do vines, lianas, or hardwood trees. Hardwood stems include trees, treelets, and shrubs. Differences in vegetation structure between forest types can lead to differences in microclimate. In turn, microclimatic differences may affect forest use by vertebrates. To test for microclimatic differences in my forests, I recorded temperature and relative humidity during the first day of mist netting (see below) in each of the 10 forests at 0600, 1200, and 1800 hrs. In each of the ten forests all three measurements were taken in the same location, which was always shaded understory. Measurements were taken with a thermometer and a hygrometer. Bird Censusing I censused the avian communities in the same ten a^ai-enriched and control stands using mist nets during September November 1997, which coincided with the a9ai fruiting season. These data were supplemented with casual observations of birds seen or heard during mist netting or vegetation sampling. Although the use of nets introduces biases associated with differences in net visibility among habitats being sampled, and differences among species in their probability of being captured (Karr 1981, Remsen and Good 1996), I chose to use them for several reasons. Specifically, nets allow the detection of species that are more secretive or nocturnal, do not vocalize consistently (e.g., some hummingbirds and cotingas), or that are not vocalizing during the period of sampling. In addition, mist nets were used to sample birds in the only other comprehensive studies of

PAGE 26

17 birds in floodplain forest in this region (Novaes 1970, Lovejoy 1974), so their use in this study allows a direct comparison to other studies. In each stand, 12 nets (36 mm mesh, 12 x 2.6 m) were placed in three groups of four, each group comprising a net "lane" of approximately 50 m. Lanes within each stand were 50 100 m apart. An attempt was made to place the lanes parallel to each other, but this was not always possible owing to the interruption of vegetation cover by streams and trails, which were avoided. Nets were opened at 05:45 (dawn) on two consecutive days at each site They were left open for 12 hours on the first day and 6 hours on the second (216 net-hours per stand). To avoid a seasonal bias, I alternated between sampling enriched stands and control stands. Nets were checked at least hourly. All birds were removed, placed in a cotton bag, and carried to a central station for identification and measuring before being released at the point of capture. The first time each bird was captured I clipped a feather on the left wing in order to identify re-captures, which were not included in analyses. Avian Perch Use and Availability Availability of perches can be an important mechanism driving differences among bird communities in areas with differences in vegetation, because foraging birds tend to select perches of a given size (Kendeigh 1945, Morse 1976, Robinson and Holmes 1982). The availability of perches of certain dimensions may be altered by management, which removes some types and sizes of trees, vines, lianas, and other understory vegetation (hereafter "stems"). Selection for perch sizes by birds can be tested by comparing the size distribution of perches used by birds with that expected if birds were selecting perches based on the availability of stems of different sizes in the forest. To do this, I determined

PAGE 27

18 the perch sizes and types used by four antbird (Formicariidae) species and compared them with stem sizes and types available in the five enriched and five control forest stands. The four focal species were chosen because they were commonly seen and forage almost exclusively in the understory, in contrast to other species that utilize both understory and canopy strata and would therefore be difficult to observe. During the same period as mist-netting, I observed Thamnophilus nigrocinereous, T. punctatus, Cercomacra tyramia, md Myrmotherula axillaris, and recorded the diameters of perches they used while foraging. I observed birds opportunistically as I encountered them, following individuals as long as possible. At least five individuals of each species were followed, but some individuals were followed multiple times. I determined the availability of stems that birds could use as perches along the transects I used to sample vegetation. I measured the diameters of all perch-like stems (including vines, liamas, tree branches, palm leaf petioles) that touches my outstretched arms as I walked along the transect. Mammal Trapping As with birds, mammals were censused to evaluate their use of afai-enriched versus control forest. Mammals were trapped with Sherman (6.6 x 7.7 x 19.8 cm) and Tomahawk (52.8 x 15.4 x 15.4 cm) live traps during December 1996 February 1997. Because this was a period of low E. oleracea fruit production, differences in mammal captures between enriched and control forests would represent responses of mammals to differences in factors other than fruit availability (e.g., vegetation structure, microclimate, nest sites). Traps were set at 10 m intervals on a grid system in two, one-hectare plots one in an a9ai-enriched stand and one in a control stand. The number and position of

PAGE 28

19 traps fluctuated throughout the study, but each forest typically contained 20 small and 4 large traps. Traps were moved periodically from ground level to 4 m, in response to changing flood levels and in order to sample both terrestrial and understory habitats. Understory traps were tied onto tree limbs or, occasionally, placed on platforms that were hung next to trees. Traps were never opened for more than four consecutive nights. Trap effort totaled 521 small and 172 large trap-nights in control forest, and 509 small and 151 large trap-nights in afai-enriched forest. Depending on availability, bait included bananas, papaya, pineapples, and com, but in any given night the same bait was used in all traps. I used fruit for several reasons. First, most small, non-volant neotropical mammals include fruit in their omnivorous diets (Emmons 1997). Second, fruit is often used as bait in studies of non-volant neotropical mammals (e.g., Adler 1998). Furthermore, fruit was consistently available to me during the trapping period. Bait was placed in the traps in the late afternoon and traps were checked the following morning. Animals were removed from the traps, identified, measured, marked and released at the trap site. Henna hair dye was used to place a unique mark on each animal's flank, which aided in identifying re-captures. Data Analysis Vegetation and microcl imate sampling Canopy height and canopy density were compared between enriched and control forests with nested ANOVAs on transformed data. Total basal area was compared with a t-test. Because understory vertebrates might be sensitive not only to the total density of stems but also to the types of stems available, I separately compared the densities and dbhs of palms, vines, lianas, hardwoods, andM linifera with t-tests. To assess the

PAGE 29

20 availability of small perches to vertebrates, I compared the distributions of small (0-5 cm dbh) tree stems, vines, and woody lianas between the two forest types with a chi-square test of homogeneity. Finally, temperature and relative humidity were compared with Wilcoxon Rank Sums tests. Bird censusing Bird communities in the two forest types were compared with three methods commonly used in community comparisons. First, to quantitatively describe the communities, I calculated the total number of captures, species richness, and species diversity for all birds captured in nets, and species richness for all birds captured and observed. Species diversity was estimated with Simpson's Index: i=l where pi = proportion of species /, and 5 = the number of species. The second comparison of the bird communities evaluated species composition using Dice's Index (1945). This index, which estimates the similarity of two communities, is calculated as: 2a {2a + b + c)' where a is the number of species common to both habitats, and b and c are the number unique to the two habitats. A value of 1 indicates complete species overlap. I used this index to answer two questions. First, how similar is the bird community in enriched

PAGE 30

21 forest to that in control forest? I pooled all species caught in a^aienriched forest and pooled all those caught in control forest, then calculated the index. Second, I asked the question: how similar are enriched forests to one another, and how similar are control forests to one another? To answer this, the index was calculated for all pairs of sites within each forest type, which resulted in 10 index values for a9ai-enriched forest and 10 for nonenriched forest. I used a Wilcoxon Rank Sums test on these values to evaluate which forest type was most similar among sites. Responses of bird species to forest management may differ among members of different dietary guilds and among those that utilize different strata of the forest 1988, Thiollay 1992, Mason 1996, Greenberg et al. 1997). In particular, understory insectivores tend to decline dramatically in enriched forests. I tested the hypothesis that bird species would differentiate between enriched and control forest based on diet and forest strata preferences. I used correspondence analysis, which is an ordination procedure useful for testing hypotheses about joint relationships between variables with categorical data (James and McCulloch 1990). I first assigned species to guilds (using Hilty and Brown 1986, Levey and Stiles 1992, Ridgely and Tudor 1994, Mason 1996, and Greenberg et al. 1997, and personal observations of food consumption) and strata preferences (using Parker et al. 1996). Some species, generally considered omnivorous, were classified as granivorous or frugivorous if they were observed primarily eating seeds or fruit during this study. Correspondence analysis (SAS 1996) was based on the number of captures of each species in the two forest types.

PAGE 31

22 Avian perch use and capture times I tested if the four antbirds that I observed showed preferences for perches of certain sizes, and if the abundance of those perch sizes differed between forest types. To test for perch size preferences, I compared the size distribution of perches used by birds with the distribution that would be expected if birds were simply using stems according to their availability with a chi-square test. Use of a forest type might also depend upon favorable microclimatic conditions. I predicted that, if temperature and relative humidity influence the activity of birds, then this may be reflected in the time of day at which I captured them. Therefore, for all species that were captured in both forest types (= 1 5 species), I compared capture times with a one-way ANOVA. Mammal trapping Capture rates of mammals were standardized for the number of trap-nights (= number of traps x number of nights open), which was higher in control than in a9aienriched forest. To compare trap success between stands, I pooled all individuals of the three species that I captured. Captures were dominated, however, by one species {Marmosa murina), so it alone was used to compare the number of individuals between forest types using a chi-square test. Results Vegetation and Microclimate Vegetation structure differed substantially between a9ai-enriched and nonenriched stands. Forest canopy averaged 16.2 m in control stands and 10.2 m in enriched

PAGE 32

23 stands (F,,„ = 1 1 1.50, p < 0.01). Canopy density was 86.6% in control stands, but only 77.4% in enriched stands (F,,„ = 96.70, p < 0.01). Stem density (t = 3.91, df = 8, p < 0.05) and basal area (t = 3 .72, df = 8, p < 0.01) were similarly greater in the control stands (density: 4680+ 1197, area: 16.4 + 2.0) than in the enriched stands (density: 1640 ± 1267; area: 3.4 + 2.9). A?aienriched and control stands also differed in vegetation composition (Fig. 23). A(;aienriched stands contained six times the number of reproductive-sized a^ai adults (t 3.86, df = 8, p < 0.01), eleven times the number of juveniles with trunks (t = 2.97, df = 8, p < 0.02), and seven times as many juveniles without trunks (t = 3 .96, df = 8, p < 0.01) than did control stands. On the other hand, control stands contained four times more small trees (< 10 cm dbh; t = 3.84, df = 8, p < 0.01), five times more non-woody vines (t = 2.95, df = 8, p < 0.02), and 84 times more woody lianas (t = 4.51, df = 8, p < 0.01). Stems of a^ai trees were larger in enriched stands, while vines had greater diameters in control stands (Fig. 2-4). These differences, particulariy the virtual lack of vines and lianas in the a? aienriched stands, led to significantly different distributions of small-diameter trees, vines, and lianas (together "stems") in the two forest types (Figs. 2-5a & b, = 20.27, df = 9, p < 0.001). While most small stems in the control stands measured less than 1 cm dbh, stems were found in all size classes between 0 5 cm (Fig. 2-5a). In contrast, most small stems in afai-enriched stands were either 0-1 cm or 4 5 cm, with relatively few stems at all, and no vines or lianas, between 2 4 cm (Fig. 2-5b).

PAGE 33

24 9000 n Stem types Figure 2-3. Numbers of stems of different types (mean + s.e.) in five stands enriched with E. oleracea and five non-enriched stands. Bars show standard error and asterisks indicate significant (p < 0.05) differences between stand types.

PAGE 34

25 Stem types Figure 2-4. Diameters (dbh) of different types of vegetation (mean + s.e.) in five forest sites managed fori:, oleracea and five non-managed sites. Bars show standard error and asterisks indicate significant (p<0.05) differences between forest types.

PAGE 35

26 Diameter (cm) Figure 2-5. Distribution of small (0-5 cm) trees, vines, and lianas among size classes in two different stand types. a) E. o/emcea-enriched forest b) Non-enriched forest

PAGE 36

27 Measurements of temperature and relative humidity showed that microclimate fluctuated more throughout the day in enriched stands than in control stands. Dawn and dusk temperatures were similar in the two forest types, but noon temperature was 33.7 C in the enriched stands, but only 31.2C in the control stands (Table 2-1). Relative humidity did not differ significantly between the two forest types. Table 2-1. Temperature (C) and relative humidity (%) in afai-enriched and control stands. Significant difference at p < 0.05 is indicated by '. A(;ai -enriched Control Dawn temperature 25.6 26.5 Dawn relative humidity 88.5 86.5 Noon temperature * 33.7 31.2 Noon relative humidity 69.25 75.5 Dusk temperature 28.9 29.3 Dusk relative humidity 82 81.5 Bird Censusing I captured 54 bird species in nets and observed an additional 16, for a total of 70 species among all ten sites (Table 2-2). Species accumulation curves indicate that most species were detected after the first 15 of the 18 hrs of netting; few new species accumulated in the final three hours (Fig 2-6). Although 70 species is a relatively low species richness for an Amazonian site (Terborgh et al. 1990, Thiollay 1994), it is not especially low for floodplain forests in eastern Amazonia, which are known to have lower species diversity than interfluvial and western Amazonian sites (Lovejoy 1974). Two eariier studies (Novaes 1970, sampling for 12 months; Lovejoy 1974, sampling at two sites for 22 months) in similar habitat found between 62 and 87 bird species, indicating that my results fall within the expected species richness range for the region.

PAGE 37

28 o c CT3 CO H •d duals ;d stand t Strata. detecte CLI CO CO '> indi nric CO o -ed .2 a c3 «> c . O diet ivel ands nets ies' pec CO CO 0) t) a. Ui O T3 CO CO U( •a 1 3 ca an Q, CO C3 •4— ( O C T3 ont c/5 ese :he o spec •c a c IT) 1 mns -uou O olu T3 C o o CO •d •o •a ricl ure hir :he af-ei cap and -enri w o T3 c C o 3 o Cm T3 CO O erv( ved 'he ber bs^ u o CO 3 x> C •a o "o CO T3 •o _c O 0) s 3 c cap rese nds CO CO C Bird CO ol St ilum t— o u c o o 2i o 3 D, c o fin H c CO T3 C o U ON 00 CO •o c i3 CO •a •c c o o 2 CO 'o D. CO CM E o E s <3 •S •5 a c-l 3 3 •2 ad a s «3 t a s Co S 3 3 C«! «J 4^
PAGE 38

29 cn o m >n o (N o CO o o o CO CO CO X 1/-) 1/-) X O CO o X X X (S CO CN < Glyphorhyncus spirurus Nasica longirostrus Xenops minutus Cercomacra tyrannia Formicivora grisea Hypocnemoides melanopogon Microhopias quixensis Myrmotherula axillaris Sclateria naevia Thamnophilus nigrocinereous Thamnophilus punctatus Thamnophilus amazonicus aureola Manacus manacus Querula purpurata

PAGE 39

30 o o o 0 0 0 0 0 0 0 0 0 0 rj0 0 o
PAGE 40

31 ^3 s I t2 m m o o o o o >n m fS (N CN CO rCN CO X u u/c u/m u/c u/m u/m 3 m/c u/c DC DC DC Thraupis palmarum Ramphocelus carbo Eucometis penicillata Tachyphonus rufus Oryzoborous angolensis Sporophila americana Sporophila nigricollis Saltator maximus Saltator orenocensis 2 o _> 'C o c II c o > 'c II .2 oa is 2 o > '5b O _> 'c E o II o oT V-i O > CO 1> O TD C 3 O o •a c 3 CO II o > II E >. a, o

PAGE 41

32 1 2 3 4 5 6 7 8 9 101112131415161718 Hours netting Figure 2-6. Bird species accumulation curves during 216 mistnet-hours in ten stands. Squares represent £. oleracea-ennched stands, while x's represent non-enriched (control) stands. Species richness was nearly twice as high in enriched stands than in control stands when only netted birds were considered (43 vs 25 species, = 4.76, df = 1, p < 0.05) and when both netted and observed birds were considered (57 vs. 36 species; = 4.74, df = 1, p < 0.05). I found no differences between enriched and control stands in the number of individuals captured (194 vs 180; x' = 0.52, df = 1, p > 0.50) nor in species diversity (Simpson's Index: 8.2 vs 9.4; Mann-Whitney U test, U = 14, p > 0.10).

PAGE 42

33 The difference in species richness between the two forest types reflects their nearly completely different sets of species. Of the 70 species observed overall, only 24 were common to both forest types, and Dice's Index of Similarity of the two forest types was low (0.5567, 1.0 indicates complete overlap). Values of Dice's Index were also significantly (Mann-Whitney U test, U = 1 1 .063, p < 0.01) greater among the nonenriched forests (0.6004 ± 0.080) than among the enriched forests (0.3966 ± 0. 123). This higher degree of similarity among control sites suggests a core set of species typical of control forest. Indeed, six species {Glyphorhynchus spirurus, Xenops minutus, Myrmotherula axillaris, Thamnophilus nigrocinereous, Pipra aureola, Eucometis penicillata) were caught in every control forest. In contrast, only one species, the silverbeaked tanager (Ramphocelus carbo), was caught in every a^ai-enriched forest. Moreover, some species were observed in only one of the two forest types. For example, I observed nine species in the control sites only, and these tended to be smallbodied, nectarivorous, piscivorous, or insectivorous species that inhabit forest understory [four hummingbirds (Campylopterus largipennis, Phaethornis ruber, Thalurania sp., Threnetes leucurus), pygmy kingfisher (Chloroceryle aenea), black-chinned antbird (Hypocnemoides melanopogon), slaty antshrike {Thamnophilus punctatus), scale-crested pygmy-tyrant (Lophotriccus pileatus), river warbler {Basileuterus rivularis)]. Of the species observed but not caught, the raptor i5u/eo magnirostris and the parakeet Brotogeris versicolurus were seen only in control forest. On the other hand, 32 species were observed in a? ai-enriched forest only, and 20 of these were represented in only one of the sites (#4). These species tended to be larger-bodied frugivores or omnivores, many of which occupy the forest canopy (Table 2-2).

PAGE 43

34 Correspondence analysis revealed significant relationships among diet, forest strata preference, and forest type preference (X = 170.24, df 105, p < 0.001; Fig. 2-7). Most notably, insectivores were associated exclusively with the understory of control stands, while frugivores were associated with the understory and midstory of both enriched and control stands. Omnivores and piscivores, on the other hand, were associated almost exclusively with enriched stands; omnivores in the canopy, and piscivores in the midstory and understory. Nectarivores and granivores showed few preferences, being caught in all strata of both stand types. In addition to comparing the species composition of the bird communities, I also compared capture rates of bird species, as an indication of forest type preferences. Most of the frugivorous, granivorous and omnivorous species were caught with greater frequency in the a^ai-enriched forest, whereas most of the non-frugivorous species (insectivores, nectarivores, and piscivores) were caught more in the control forest (Fig. 28, X' = 6.37, df= l,p<0.01). Avian Perch Use and Capture Times Numerous mechanisms, including understory perch availability and microclimate, may underlie the observed differences in avian communities. The four antbird species all used perches between 0.5 and 3.0 cm, although they differed in their perch use according to body size (Fig. 2-9). The smallest of the observed species, M axillaris, used mostly <2.5 cm perches, while the larger C. cinerescem used mostly 0.5 4.0 cm perches. The larger T. nigrocinereous and T. puntatus both used perches of all diameters between 0.5 5.0 cm. Perch use by all four species differed significantly from that expected if birds used perches according to their availability in the environment, indicating perch size

PAGE 44

35 0.8 0.60.4 0.2-0.0granivores nc 1 . lie uc u "nectarivores -0.2-0.4' -0.6 -0.8. e/um lie ne. urn -0.8 -0.6 -0.4 -0.2 -0.0 c2 .2 -r.4 .8 Figure 2-7. Results of correspondence analysis between dietary guild, forest type, and forest strata preferences of 56 bird species captured in E. o/eracea-enriched and nonenriched (control) stands. Circled letters represent guilds: f=frugivore; i=insectivore; n=nectarivore, g=granivore; o=omnivore; and p=piscivore. Other notation indicates forest type/strata. Forest types are e=enriched, ne=non-enriched. Strata are: c=canopy; m=midstory; u=understory; t=terrestrial Plot shows that insectivores were associated exclusively with the understory of non-enriched stands, omnivores were associated with the canopy of enriched forests, and frugivores were associated with the understory and midstory of both enriched and non-enriched forests. Piscivores, granivores, and nectarivores showed fewer forest type and strata preferences, being associated with both forest types and all strata.

PAGE 45

36 2000 more captures in a9U-enriched stands granivores JU fragivores m -500more captures in non-enriched stands piscivoies -lOOO^j U species abbreviations Figure 2-8. Percent difference in bird species abundance, estimated from mist-net captures, in five forest sites enriched for oleracea and five non-enriched sites. Change is shown relative to non-enriched forest, so that positive values indicate more captures of that species E. o/eracea-enriched forest, and negative bars, more captures in non-enriched sites. Species are grouped according to dietary guilds. Abbreviations are as follows: Ep = Eucometis penicillata. Mm = Manacus manacus. Pa = Pipra aureola Mo = Mionectes oleaginea, Tr = Tachyphonus nifus, Rc = Ramphocelus carbo, Sn = Sporophila wgricollis, Sm = Saltator maximus. So = Saltator orenocensis, Oa = Oryzoborous angolensis, Lr = Leptotila rufaxilla, Sa = Sporophila americana, Pc = Pachyramphus castaneus, Pp = Pachyramphus polychopterus, Tf = Tolmomyias flaviventris, Ms = Myiozetes similis, Pm = Phaeomyias murma„Ga = Geothlypis aequinoctialis, Ct = Columbina talpacoti. Cm = Crotophaga major, C{= Celeus flavus, Ec = Eleania chiriquensis, Mg = Myiopagis gaimardii, Mc = Myiozetetes cayanensis, Cof = Coerebaflaveola, Ta = Turdus albicollis, Tuf = Turdus fumigatus, Tl = Threnetes leucurus, CI = Campylopterus largipennis, Pr = Phaethornis ruber, Ts = Thalurania sp., Thf = Thalurania Jurcata, As = Amazilia sp., Gh = Glaucis hirsuta, Ps = Phaethornis sp, Gs = Glyphorhyncus spirurus, Br = Basileuterus rivularis, Hm = Hypocnemoides melanopogon, Tp = Thamnophilus punctatus, Lg = Lophotriccus galeatus, Xm = Xenops minutus. Ma = Myrmotherula axillaris, Cet = Cercomacra tyrannia, Tn = Thamnophilus nigrocinereous. Sen = Sclateria naevia, Xg = Xiphorhynchus guttatus, Xp = Xiphorhynchus picus, Mq = Microhopias quixensis, Tm = Todyrostrum maculatum, Fg = Formicivora grisea, Ca = Chloroceryle aenea, Ci = Chloroceryle inda, Cha = Chloroceryle americana.

PAGE 46

37 selection (M axillaris, = 78.1, df = 2, p < 0.001; C. tyrannia. x = 40.6, df = 2, p < 0.001; T. nigrocinereous: = 42.3, df = 2, p < 0.001, T. punctatus. X = 42.4, df = 2, p < 0.001). Small-diameter perches, especially small-diameter vines and lianas, were extremely rare in afai-enriched forests (Fig. 5). For the 15 bird species that were captured in both forest types, I compared the mean capture times as a way to assess activity levels over the course of the day. Mean capture time was significantly later (11:01 hrs) in the control forest than in the afaienriched forest (10:06 hrs; Fi.^is = 7.04, p < 0.008), implying that activity is concentrated earlier in enriched forest, perhaps because greater mid-day temperatures there deter midday activity of birds. Mammal Trapping The same three species of marsupials were caught in both a9ai-enriched and control stands (Table 2-3). There were no differences in capture rates between stand types ofMarmosa murina, the smallest species captured {y^ = 2.37, df = 1, p > 0.05). Too few individuals of the other species (Didelphis marsupialis, Caluromys philander) were captured to statistically compare. Capture success in small traps was nearly double in enriched stands (6.3 vs. 3.8%), but capture success in large traps was greater in control stands (2.3 vs. 1.3%). Table 3. Summary of small mammal captures in the two forest types. species # a^ai# control mean mass mean mass enriched + s.e. male + s.e female Marmosa murina 31 20 57+11.01 48.62±5.42 Didelphis marsupialis 2 3 510 775 Caluromys philander 1 1 149 200

PAGE 47

38 e u o a o M. axillaris S C. cinerescens l' O p p p p 1 1 >n p in p •A 1 p 1 p 1 0.61 0.50.40.30.2o.r A. 0.0^ ElU, 9 T. nigrocinereous T.punctatus I* o p >o p O p I CN 1 1 1 1 i 1 p >n P p o >o CO Stem diameter (cm) Figure 2-9. Small tree branches, vines, and lianas used by the antbird species Myrmotherula axillaris, Cercomacra cinerescens (a), Thamnophilus nigrocinereous, and T. punctatus (b). Perches between 1 5 cm diameter are largely absent from E. ' oleracea-ennched stands.

PAGE 48

39 Discussion My purpose in this study was to compare the understory vegetation, birds, and mammals in afai-enriched and non-enriched Amazonian floodplain forests. Enrichment management of tropical forests is a recommended strategy to increase profits derived from forests by increasing the density of economically-valuable species (Schulze et al. 1994, Kainer et al. 1998, Ricker et al. 1999). Because enrichment with species that yield nontimber products can result in long-term increases in profits, this type of management is promoted in tropical forests in many regions (Schulze et al. 1994, Kainer et al 1998, Ricker 1999). This study revealed large differences between enriched and control forests in the composition and structure of both vegetation and bird communities. The mammal community, on the other hand, showed equally low diversity in both enriched and control forests. The apparent effects of enrichment management on forests have not before been appreciated, but should be considered wherever this strategy is employed. Changes in Flora and Fauna Most of the changes I found in vegetation structure and composition are not surprising given that the goal of management is to increase afai production. As expected, enriched forests have higher densities of a?ai (Anderson and Jardim 1989, Pliraoka 1995, Pollack et al. 1995). Lower overall stem density and abundance of non-palm stems reduce perceived competition for nutrients and light; removal of vines and lianas also reportedly make enriched forests safer places to work. Thinning and canopy opening are recommended practices for increasing palm growth rates and fruit production (Anderson and Jardim 1989, Jardim and Rombold 1994) and are common throughout the Amazon

PAGE 49

40 estuary (Anderson et al. 1995, Muniz-Miret et al, 1996). By all measures, a?ai management succeeds in augmenting palm productivity (Anderson and Jardim 1989). This study reveals, however, that management is also associated with a substantially altered bird community. Afai -enriched stands contained a number of species not usually found in floodplain forests, and lacked others that are typical of floodplains of the region (Novaes 1970, Lovejoy 1974). Furthermore, the community associated with enriched stands was biased towards fruitand seed-eating frugivores, granivores, and omnivores, whereas understory insectivores were underrepresented. As suggested for forests enriched for other non-timber products (Greenberg et al. 1997), afai-enriched forest may serve as "secondary habitat" for some bird species. For example, frugivores may forage in afai patches but may nest and conduct other nonforaging activities in control habitat. I found no nests in enriched forest, but found four nests of T. nigrocinereous and one of M axillaris in control forest. Moreover, frugivore use of agai-enriched forest likely increases during afai fruit production (when mist netting occurred). More detailed observational studies are required to determine the relative importance and seasonal use of different forest types for birds, many of which likely use more than one type. In any study using mist nets, some caution is warranted in drawing inferences about bird communities because of potential biases (Karr 1981, Van Remsen and Good 1996). For example, differences in vegetation structure between enriched and control forest could differentially influence the capture of birds. The lower canopy in enriched sites, for instance, could result in a higher probability of capture of a canopy-dwelling bird. There was no evidence for this, however, as canopy-dwellers found in enriched

PAGE 50

41 forest (e.g., Myiozetetes cayanensis, Pachyramphus castaneus, P. polychopterus, Eleania chiriquensis) were neither caught nor observed in control forest. Furthermore, several of the canopy species, such as Formicivora grisea and Coerba flaveola, were not caught in control forest in this study nor by Novaes (1970) or Lovejoy (1974), and are considered edge or disturbance species (Hilty and Brown 1986). Surprisingly, only three species of mammals were captured, and I saw no evidence, such as partially eaten fruits, feces, or tracks, of other species (except domesticated pigs). If the mammal community consists only of the three marsupials that were captured, then it is very depauperate compared with other Amazonian sites. People currently hunt D. marsupialis (S. Moegenburg pers. obs.), and it is possible that past hunting, over the long period during which humans have occupied the region (Hiraoka 1995, Roosevelt et a. 1996), reduced the species diversity of mammals. The three mammal species that were captured showed no differences in abundance between forest types. This may be due to several factors. First, these three species are habitat generalists and omnivores and may not discriminate between the two forest types sampled in this study. As has been shown for marsupials in other habitats (Malcomb 1988), they tend to be less sensitive to habitat alteration than are other types of mammals, such as rodents. A second explanation may be the timing of trapping, which took place during the winter, when a? ai fruit is scarce. During the summer, when a^ai fruit is abundant, marsupials may utilize a9ai-enriched forest to a greater extent. Marsupials are known to respond to phenology of fruit production in other neotropical forests (Charles-Dominique et al. 1981). The mammalian community was undoubtedly

PAGE 51

42 more diverse in the past, but has probably suffered from hunting and the high human population density in the region. Mechanisms of Bird Community Change The differences in the bird communities between enriched and control forests likely result from the differences in vegetation structure in these two forest types. In particular, the abundance of fruit-producing afai could explain the dominance by fruitand seed eating species in enriched forests. Six of the 28 fruit-eating species caught in enriched forests showed evidence of a^ai fruit consumption (defecation of a? ai seeds or purple-stained bills). While the bounty of a^ai may be the lure for some frugivores, others may respond to an overall increase in fruit production resulting from the more open canopy in enriched forests. A similar result has been found in selectively logged forests (Johns 1988, ThioUay 1992, Mason 1996) and in coffee agroforests (Greenberg et al. 1997). In general, frugivores respond to spatial and temporal variation in fruit resources (Levey 1988, Loiselle and Blake 1991, Rey 1995). Nectar and fruit tend to be ephemeral resources, so may only function to attract animals to a habitat during the season of production (Levey and Stiles 1992). A second indirect effect of a^ai management on avian communities could be the change in relative abundance of palm and non-palm trunks and leaves. The high abundance of a9ai trunks and leaves, and the low abundance of bark-producing plants, in the enriched understory could render the habitat less suitable for bark foraging (e.g., Glyphorhyncus spirurus, Xenops minutus) and foliage gleaning (e.g., Basileutenis rivularis, Hypocnemoides melanopogori) insectivorous species. Palm leaves are structurally well defended against herbivores, and may support fewer or different

PAGE 52

43 arthropods than non-palm leaves. Likewise, barkless a9ai trunks provide poor substrates for bark-dwelling invertebrates and may therefore discourage use of the habitat by insect feeders. Another mechanism for the differences in the bird community could be the lower abundance of small diameter perches in enriched forest. The four species of antbirds displayed preferences for small perches 0.5-3 .0 cm in diameter, which were relatively and absolutely more common in control forest, despite the abundance of a9ai leaves in the former. Some understory insectivores, such as Cercomacra cinerescens and Xenops minutus, specialize their foraging on small vines and lianas (Ridgeley and Tudor 1994). In addition, understory frugivores such as Manacus manacus and Pipra aureola use small understory perches for displaying (Snow 1962a,b). Indeed, the lack of small vines, lianas, stems, and branches in the understory could explain why, despite the higher abundance of fruit in enriched forest, four species of frugivore (M manacus, P. aureola, Eucometis penicillata, and Mionectes oleaginea) were more abundant in the control forests. Owing to the dependence of many birds on small perches, the management practice of methodically removing vines and lianas may have one of the most detrimental effects on the understory bird community. Finally, the higher temperature and lower relative humidity at noon, and the overall greater fluctuation in temperature and humidity in the enriched forests, could affect bird activity and community composition. For the 15 species that were caught in both forest types, mean capture time was significantly later (by one hour) in the control forest, suggesting that bird activity extends later into the day there, perhaps due to more favorable microclimatic conditions (Karr and Freemark 1983). Some species that were

PAGE 53

44 captured primarily in enriched forest, such as Formicivora grisea, are considered secondary forest or scrub dwelling birds (Ridgeley and Tudor 1994). To the contrary, the river warbler, Basileuteris rivularis, is typical of forest interior species that seem sensitive to microclimatic changes associated vwth opening of the forest canopy and understory (Thiollay 1992, Mason 1996), further indicating that the abiotic differences that I recorded may indeed influence bird distribution and behavior. Such impacts of microclimate on birds have been implicated in other enriched forests (Karr and Freemark 1983, Johns 1988, Stouffer and Bierregaard 1995) and coffee agroforests (Bawa and Seidler 1998). Comparison to Alternative Forest Management Systems The differences in bird community structure between a9ai-enriched and control forests are comparable to those between agroforests, plantations, and logged and control forests. In particular, canopy dwelling and fruit-eating species tend to remain or become more abundant in modified forest, while forest interior, insect-eating species decline (Bawa and Seidler 1998). These patterns have emerged from agroforests in Sumatra (Thiollay 1995), shade coffee plantations in Mexico (Greenberg et al. 1997), cacao plantations in Brazil (Alves 1990), and logged forests in Venezuela (Mason 1996), French Guiana (Thiollay 1992), Malaysia (Johns 1988), Borneo (Lambert 1992), and Indonesia (Marsden 1998). As in this study, likely mechanisms for these bird community responses include higher flower and fruit production (Greenberg et al. 1997), greater canopy openness (Thiollay 1992, 1995, Mason 1996), lower understory stem densities (Mason 1996) warmer understory temperatures (Thiollay 1992), and fewer understory microhabitats such as vine tangles (Thiollay 1995) in modified forest. Forest management

PAGE 54

45 has substantial ecological impacts, whether it is for timber, coffee, or non-timber forest products like a^ai. Recommendations and Conclusions My results indicate that the role of forests enriched for NTFPs in conservation should be re-evaluated. In particular, enriched forests such as those in this study can be valuable complements to, not replacements for, forests dedicated to conservation Based on this study, I offer several recommendations to reduce the effects on bird communities of forest enrichment with E. oleracea. Further research is required on other non-timber forest product yielding species to assess the generality of both the results of this study and these recommendations. Wherever enrichment takes place, steps should be taken to maintain characteristics, such as canopy height and density, of control forests. Moreover, as little modification as possible should be done to the understory vegetation of enriched stands. Removal of small trees, vines, and lianas decreases perching and foraging substrates for understory animals. Furthermore, maintenance of both understory vegetation complexity and a dense canopy can help prevent large fluctuations in understory microclimate, to which understory animals can be sensitive. Research is needed to identify ways to successfully enrich forests with economic species without substantially modifying overall forest structure and composition. My results suggest that control forests may play an important role in conserving regional biodiversity by supporting certain species, especially understory insectivores, not found in enriched forests. In my study area, the area of forest left control is shrinking, however, because market demand for a9ai fruit is growing, which encourages

PAGE 55

46 landowners to increase a9ai enrichment in forests (Hiraoka 1995, Muniz-Miret et al. 1996). Encouraging landowners to maintain a portion of control forest in the present state could be accomplished in several ways. One method would be to promote markets for other non-timber forest products, such as andiroba, which are harvested from control forests. A second method might be a community-wide certification program (Kiker and Putz 1997), in which people or families that maintain a certain percentage of their forest as control become certified producers of a^ai and earn more per volume of fruit than do non-certified producers. Higher prices captured for both afai fruit and palm heart or improved market access could balance any profits lost by leaving some forest control. Such incentive programs could be viable, because the landscape itself promotes the maintenance of the existing system, in which not all forest is intensively enriched. The estuary region consists of islands surrounded by rivers; most households and enriched forest lie along the river margins. Because the islands are roughly circular in shape, the individual land holdings take the shape of pie slices, joining at their points near island centers. Island centers are difficult to access and therefore usually remain non-enriched. The aggregation of the control areas of various families' forests means that if a certification program is implemented, then all the parcels of control forest would remain grouped. This may be the most effective manner in which to promote both forest utilization and conservation in the a?ai management system.

PAGE 56

CHAPTER 3 LINKING FRUIT AND FRUGIVORE ABUNDANCE: EXPERIMENTAL EVIDENCE FROM AMAZONIAN BRAZIL Collecting the fruits produced by oligarchic forests is one of the most benign forms of resource exploitation practiced in Amazonia. If harvests are properly controlled and conducted in a non-destructive fashion, fruit collection has minimal impact on forest structure and function. Canopy cover and floristic composition are maintained, the constituent fauna is preserved, and the cycling of water and nutrients remain essentially unaltered. The fact that many oligarchic forests occur in habitats subjected to seasonal flooding and sediment deposition suggests that nutrient losses resulting from fruit removal are likely to be quite small Peters, 1992, pg. 17. Introduction The ubiquity of fruit-eating animals and fleshy fruit-producing plants in neotropical forests (Gentry 1982, Karr et al. 1990, Robinson and Redford 1986, Levey and Stiles 1992) suggests that fruit-frugivore interactions influence the ecology and evolutionary history of the organisms involved (McKey 1975, Snow and Snow 1980, Bodmer 1989, Fleming 1991, Levey and Stiles 1992). One prediction stemming from this idea is that the abundance and distribution of fruits and frugivores should be linked (McKey 1975). Indeed, several studies demonstrate correlations between fruit and frugivore abundance both within and among habitats (Snow 1962a, b, Crome 1975, Worthington 1982, Wheelwright 1983, Levey 1988, Sargent 1990, Loiselle and Blake 1993), and others show that frugivores apparently track fruit resources over space and time (Loiselle and Blake 1991, Powell and Bjork 1995, Rey 1995, Kinnaird et al. 1996, Bjork 2000). Moreover, fruit availability seems to regulate populations of some highly47

PAGE 57

48 frugivorous tropical animals (Foster 1982a, Terborgh 1986b, Snyder et al. 1987, Abramson et al. 1995, Adler 1998, Wright et al. 1999). Another series of studies, however, calls into question the significance of fruitfrugivore relationships. Very few vertebrates, for example, are completely dependent on fruit; indeed many of them switch to other foods when fruits become scarce (Loiselle and Blake 1993, Pina 1999). In addition, fruit-frugivore interactions vary greatly over space (Bronstein and Hoffmann 1987, Fleming 1991, Chapman in prep.) and time (Herrera 1998), so their role in selecting for traits such as fruit morphology or frugivore behavior might in fact be limited (Herrera 1992). These results have led some (Levey and Benkman 1999) to question the ecological and evolutionary significance of fruitfrugivore relationships. That these relationships are more complex than was previously thought indicates the need for more refined research questions, such as; which frugivores are sensitive to fruit abundance and which are not? At what scale do frugivores detect variation in fruit abundance? At what hierarchical level (e.g., populations, communities) do frugivores respond? To date, however, answering these types of questions has been hindered by, among other things, the logistical difficulties of experimentally manipulating fruit abundance in forests. Responses of frugivores to fruit abundance may depend upon many factors, including the scale at which frugivores operate in the environment, their degree of frugivory, their body size and coloration, and their social structure (Howe 1979, Pratt and Stiles 1983, Levey 1988, Chapman et al. 1989, Wheelwright 1991, Loiselle and Blake 1993, Westcott and Graham 2000). Large-bodied tropical animals with large ranges may not respond to local variations in fruit availability (Bourne 1974, Jardim and Oliveira

PAGE 58

49 1997, Pina 1999), whereas smaller fruit-feeding mammals, birds, and insects might (Janzen 1970, 1971, Wright 1983, 1991, Levey 1988, Fleming 1992, Loiselle and Blake 1993, Adler 1998). In addition, highly frugivorous animals may display more sensitivity to fruit abundance than partially frugivorous ones (McKey 1975, Pratt and Stiles 1983, Levey 1988, Loiselle and Blake 1991, Chapman and Fedigan 1994, Allen 1997, Peck et al. 1999). Birds that are conspicuous to predators, either through coloration (Howe 1979) or vocalization (Bourne 1974, Snyder et al. 1987) may respond less to fruit abundance than birds that can conceal themselves in treetops. Finally, flocking species, such as parrots, may respond not only to food abundance but also to the presence of conspecifics (Snyder et al. 1987, Chapman et al. 1989). If animals do respond to decreased fruit abundance, it may be on several levels: communities may contain fewer species (Martin and Karr 1986, Develey and Peres 2000); species may be represented by fewer individuals (Karr 1976, Levey 1988, Chapman et al. 1989, Loiselle and Blake 1993, Galetti and Aleixo 1998); and individuals may alter their behavior (Davidar and Morton 1986, Sargent 1990, Sherman and Eason 1998, Shepherd and Boates 1999). Such changes may, in turn, alter processes such as fruit removal and seed dispersal (Pratt and Stiles 1983, DeSteven and Putz 1984, Wheelwright 1991, Strahl and Grahal 1991, Chapman and Chapman 1995, Allen 1997, Hamann and Curio 1999, Loiselle and Blake in prep.). Among the tropical fruits consumed by frugivores, palms are often considered "keystone" resources that maintain populations during periods of fruit scarcity (Terborgh 1986b, Peres 2000). Palms are eaten by a high diversity of birds (e.g., Galetti et al. 1999) and mammals (Terborgh 1986a), often comprising an important component of the diet

PAGE 59

50 (Foster 1982b, Terborgh 1986a, Snyder et al. 1987, Bodmer 1990, Allen 1997, Adler 1998). In southern Brazil, for example, Galetti et al (1999) documented one mammal and 14 bird species feeding on Euterpe edulis palm fruits, which comprise between 8 30% of the diets of these birds. Because of the apparent importance of palms to frugivores, then, palm-frugivore interactions might be the ideal system in which to test responses of frugivores to changes in fruit abundance. In some tropical forests, natural variation in palm fruit availability (Foster 1982a, Terborgh 1986a, reviewed by Fleming 1991, Chapman et al. 1994) is magnified by extraction of wild fruits by people (Peters et al. 1989, Vasquez and Gentry 1989, Allen 1997, see also Galetti and Aleixo 1998). In the Amazon River estuary in eastern Brazil, for example, the palm Euterpe oleracea, which occurs in high-density stands (>300 trees ha ') across ca. 10,000 km^ (Calzavara 1972, Peters et al. 1989, Kahn 1991), produces highly-valued fruit that people harvest by climbing stems and removing infructescences (Anderson 1988, 1990a, Strudwick and Sobel 1988). Many people harvest only a small amount of fruit for household use, but some households practice intensive extraction of up to 9,000 kg fruit ha"' yr' (Muniz-Miret et al. 1996). While it is recognized that many frugivores also consumed, oleracea fruit (Strudwick and Sobel 1988, Sick 1993), impacts of the harvest on frugivores have not been examined. I took advantage of the E. oleracea system to test the effects of different levels of fruit abundance on frugivore habitat selection and behavior. I first monitored the abundance of parrots, parakeets, macaws (the most frequent visitors to E. oleracea trees), and E. oleracea fruit during one season of fruit production in four E. o/eracea-dominated forests and four forests in which other species of palms occurred but were not dominant.

PAGE 60

51 As this demonstrated a correlation between E. oleracea fruit availability and frugivore abundance, I then experimentally removed E. oleracea fruit at two levels of intensity and monitored the community-level responses of mammals, birds, and seed-feeding bruchid beetles. These data were used to evaluate species-level responses of birds to determine which frugivores are most sensitive to fruit harvest and at which level of fruit reduction they cease to visit E. oleracea forest sites. With this information I formulate harvest recommendations that optimize both fruit utilization by people and maintenance of frugivore communities. This research has special relevance to Amazonian Brazil, the forests of which are under increasing pressure from logging, mining, and ranching (Anderson 1990a). Brazil has implemented a strategy to conserve forests through a system of extractive reserves: areas designated for the long-term sustainable harvest of primarily non-timber forest products, including fruit (Feamside 1989, Alegretti 1990, Mattoso and Fleischflesser 1994). E. oleracea and related species are extracted in several of these reserves, so understanding the consequences of its fruit extraction is crucial to developing harvest guidelines that can maintain forest biodiversity (Vasquez and Gentry 1989, Peters 1990). Methods Study Species The multi-stemmed E. oleracea occurs throughout the Brazilian States of Para, Amapa, Tocantins, and Maranhao, along the Pacific coast of Colombia and northern Ecuador, and in Trinidad, Venezuela, and the Guianas (Henderson 1995). Across approximately 10,000 km^ of floodplain forests in the Amazon River estuary, E. oleracea forms monodominant stands, some of which are the result of historical or contemporary

PAGE 61

52 management (Calzavara 1972, Peters et al. 1989, Kahn 1991). Individual genets (hereafter "trees" for simplicity) contain up to 25 slender stems that reach heights of 30 m (Henderson 1995). Reproductive stems produce infructescences bearing several thousand purple-black globose drupes ca. 1 cm in diameter. The ubiquity of E. oleracea across inhabited estuarine regions suggests its importance the diet, culture, and economy of the people in Amazonia (Anderson 1988, 1990b, Strudwick and Sobel 1988, Hiraoka 1995, Muniz-Mirit et al. 1996). E. oleracea yields two of the region's most profitable non-timber forest products: heart-of-palm and fruit. People harvest E. oleracea fruit by climbing stems and removing infructescences with machetes or knives. Fruits are processed into a drink consumed daily by many thousands of people. Where people have access to markets, they often increase fruit production by enriching their forests with E. oleracea, they then sell the fruit not consumed in the household. On Combu Island near Belem, Brazil, several dozen families earn more than US$3,400 per year from the sale of E. oleracea fruit (Anderson and loris 1992). Study Sites and Plots This study was carried out at the 33,000 ha Estafao Cientifica Ferreira Penna (Ferreira Penna Scientific Station; r42'30" S, 5r31'45" W), operated by the Museu Paraense Emilio Goeldi of Belem and located within Caxiuana National Forest in the municipality of Melga90, Para State, Brazil (Fig. 3-1). Average rainfall is 2,500 3,000 mm, mean annual temperature is 26C, and mean annual relative humidity is 85%. The vegetation is evergreen humid rainforest, and the majority of the station is non-flooded, terra firme forest (Lisboa 1997). Approximately 3,300 ha, mostly along the blackwater Bay of Caxiuana, is extremely low-lying forest that inundates when the river levels rise

PAGE 62

53 during the rainy seasons (December May) and high tides. Water depths reach their maxima (ca. 1 m) in May. Much of this floodplain forest is dominated by E. oleracea, with Virola surinamensis (Myristicaceae) and Pterocarpus santalinoides (Fabaceae) also common (Ferreira et al. 1997). Although there are no homes and no current management of this forest, several lines of evidence suggest that the area of floodplain forest dominated by E. oleracea (hereafter "palm forest") resulted from past human management. First, areas of highest £. oleracea density occur closet to the river margin, where human dwellings tend to sit. Second, local people commented that the names of some of the sites (e.g., Moreira, Fazenda) reflect previous uses or names of inhabitants. E. oleracea fruit begins to appear in these sites in May and persists until September. In most years people occasionally extract £. oleracea fruit from these areas, however, in 1997 and 1998 local inhabitants respected my request to refrain from extraction. Data collection occurred during the fruiting seasons of 1997 and 1998. Within palm forest I located four sites, called Plaquinha, Fazenda, Miriti, and Moreira, located approximately 1 km apart along the bay, in which I worked in both years. Within the non-floodplain forest I located an additional four sites, known as Estagao Sur, Estafao Norte, Heliporto, and Inventorio, in which I collected data during 1997 only. In 1997, 1 randomly located 1 ha square plots in each of the eight sites, in which I surveyed parrots (all sites) and monitored E. oleracea fruit abundance (palm forest sites only). These plots were demarcated with colored flagging and were bisected by a 100 m-long transect.

PAGE 63

54 X Field station o study Sites Figure 3-1. Study site within Brazil, and E. o/eracea-dominated study plots and field station within site. Fruit and frugivore surveys in 1998 were conducted in one circular plot per site, with diameter 150 m (area ^ 1.8 ha; Fig. 3-2), located roughly in the center of highest density ofE. oleracea at Plaquinha, Fazenda, Miriti, and Moreira. Each plot was divided into three equal subplots, which were further divided in half by a transect along which all censusing took place. Along each transect I constructed a "footbridge" of tree trunks to facilitate walking during flooded periods. Subplot edges, plot perimeters, and transects were demarcated with colored flagging. Fruit Availability In both years I censused fruit availability by examining E. oleracea trees along transects and recording the number of ripe infructescences by size (small, medium, or

PAGE 64

55 large). During 1997 I estimated fruit abundance during July and August in the palm forest plots by walking the 100 m-long transect in the middle of each and recording infructescences on 50 trees adjacent to the transect. In 1998, fruit surveys were carried out twice monthly during the entire fruiting season (May September) on 20 trees adjacent to transects in each of the three subplots, for a total of 60 trees per plot. The individuals observed were not necessarily the same in all censuses, but rather were chosen randomly from those occurring along the transects. plot area = 1.8ha ____ subplot boundaries . transects Figure 3-2. Schematic map of study plots, divided into three subplots, each with a 75 m transect down the middle. One subplot was randomly designated high-removal, another low-removal, and the third, control. During frugivore censusing, observer 1 began near the edge of the control subplot, while observer 2 started at the plot center. Bold and italicized numbers next to arrows indicate paths and directions walked by observers 1 and 2, respectively.

PAGE 65

56 To convert the number and sizes of infructescences to total available fruit mass, I first weighed the fruits from five harvested small, medium, and large infructescences. These average weights were then multiplied by the total number of small, medium, and large infructescences recorded in each plot or subplot. This total fruit mass was then divided by the number of trees censused (50 in 1997, 20 in 1998) for a per tree estimate of fruit availability. To determine the total fruit available per area, the per tree fruit availability was multiplied by the number of trees in the hectare plots (1997) or subplots (1998). This was done on a per-subplot basis in 1998 due to differences among subplots in^. oleracea tree densities. Parrot Use of Palm Forest Versus Non-palm Forest Before evaluating the community-level responses of frugivores to fruit harvest, I wanted to determine unmanipulated frugivore densities and seasonal fluctuations in palm forest and non-palm forest. I therefore monitored the visitations of the most abundant palm forest frugivores (parrots, parakeets, and macaws) to both forest types during the E. oleracea fruiting season of 1997. I conducted four surveys in the four palm forest sites (Plaquinha, Fazenda, Miriti, and Moreira) and three in the four upland sites (EstafSo Sur, Esta^ao Norte, Heliporto, and Inventorio). Surveys involved slowly walking along the 100 m-long transect that bisected each 1 ha plot from 07001000 hrs and recording all pssitacids within the plot, including species using the area, size of groups, and time spent in palm trees and eating fruit (in palm forest). I estimated the number of minutes that pssitacids spent in the censused areas conservatively as the longest interval between observations of a group. At the end of the season, an average group size was calculated based on all observations in which group size could be ascertained. Over the census period I detected no changes in parrot group

PAGE 66

57 sizes (Chapman et al. 1989), which were: Pyrrhura per lata (12), Pionites leucogaster (4), Deroptyus accipitrinus (3), Amazona amazonicus (2), and Ara spp. (2). For each group observed, I multiplied the number of minutes it spent in the plot by the average group size for that species. The sum of this value for all groups observed during a census was termed the parrot visitation in that plot. Frugivore Responses to Experimental Fruit Harvest The truest experimental design is a pre-test post-test of controls and treatments (James and McCulloch 1995). To maximize the robustness of my experiment in 1998, 1 therefore surveyed both fruit abundance and frugivores during a pre-harvest phase (May and June) and a post-harvest phase (July and August) in three treatments. Survey methods were identical in the two phases. Frugivore survevs While frugivore surveys in 1997 focussed on pssitacids only and were meant to establish a correlation between fruit and frugivore abundance, surveys in 1998 were expanded to include a variety of types of animals that likely perceive fruit abundance on different scales. I therefore censused fruit-eating mammals, fruit-eating birds, and seedfeeding bruchid beetles. Moreover, although their abundance was not expected to change in response to fruit harvest, I censused non-frugivorous birds, as a general gauge of seasonal fluctuations in bird abundance. Animals were classified as non-frugivores, frugivores, partial frugivores, or granivore/frugivores based on personal observations of fruit-eating and the literature (Hilty and Brown 1986, Levey 1988, Ridgeway and Tudor 1994, Emmons 1997, Levey and Stiles 1992, 1994). Only two species of nonfrugivorous mammals were observed in the plots during censuses: southern tamandua

PAGE 67

58 {Tamandua tetradactyla) and grisons {Galactis vittatd). One species of bruchid beetle {Pachymerus sveni) is known to feed onE. oleracea seeds (Johnson et al. 1995). Surveys of birds, mammals, and beetles were conducted on the same days as the fruit surveys during 1998. From 0700 1 130, two observers simultaneously walked transects in a pre-determined pattern, one starting at the plot edge and the other at the plot center (Fig. 3-2). They then slowly walked toward the center or edge, respectively, allowing thirty minutes to walk one-way on a transect. In this way the observers were always in different subplots and walked each transect twice during the 4.5 hr survey. Because surveys were conducted at ca. two week intervals, with few intervening visits to plots, it is unlikely that animals became habituated to our presence and modified their behavior. Each observer recorded on a schematic map, resembling Fig. 3-2, all birds and frugivorous mammals seen and heard, paying particular attention to the subplot in which animals were located. Multiple observations were recorded to estimate visitation time in the plot. The simultaneous observations by two observers allowed for more accurate mapping of animals. Later, final maps of animal locations and visitation times during each survey were compiled. I developed rules to aid in estimating the number of individuals and visitation times of birds. First, the number of individuals of flocking species, if not readily countable, was assumed to be the average number in flocks, as in 1997 (see above). Second, we frequently detected a bird first in one subplot, and later in another subplot. In these cases, the individual was assumed to have spent equal amounts of time in the two subplots. Final summaries of census data therefore included number of species, number of individuals, and visitation times (mins) for frugivorous birds, nonfrugivorous birds, and mammals.

PAGE 68

59 Bruchid beetles were not surveyed directly, but rather by the number of eggs they oviposited. Under five E. oleracea trees in each subplot (N=15 per site) I placed a mesh bag (mesh size 1 cm^ allowed easy passage by beetles) containing \0 E. oleracea seeds. At each bi-monthly sample, seeds were removed, placed in plastic bags, and replaced with fresh ones. After ca. 6 months, seeds were checked for bruchid beetle emergence. These emergence rates were reduced by a factor that controlled for beetles that were in the seeds prior to their use in the experiment (i.e., beetles that had entered seeds while they were still on trees). For each batch of seeds used, therefore, a sample (>100 seeds) was stored in dry, plastic bottles fitted with mesh "windows" and later (after 4-9 months) checked for beetle emergence. An average base infestation rate was thereby obtained, and this was subtracted from the actual rates observed in the experimental seeds. Fruit removal Mid-way through the 1998 fruiting season (early July) I initiated the E. oleracea fruit harvest experiment. In each plot one subplot was randomly selected as highremoval (-100% ripe fruit removed), and one as low-removal (-50% ripe fruit removed). These harvest intensities mimic extraction that is done for both consumption and marketing of fruit, and for household consumption only, respectively. In each plot one subplot was left as a control (0% fruit removed) to mimic the condition of non-harvested forests. All harvests were carried out one day prior to frugivore surveys in that plot. Four local teenagers were contracted as harvesters. They were told from which areas to harvest, but were not told which infructescences to harvest. This allowed the extraction to be carried out in as realistic a manner as possible; that is, harvested infructescences represented those that would actually be selected by local people. To remove fruit, harvesters climbed the palms and cut off the infructescences with machetes.

PAGE 69

60 This involved climbing up to 120 palms per plot. No trees were climbed in control plots. On the ground, fruit was removed from infructescences, put into baskets, and used for consumption. Harvest from some palms was impossible, as they were too thin and fragile to allow climbers to safely reach the infructescence. After the initiation of the harvests, bird, mammal, and bruchid beetle surveys continued as before. In each plot four harvests and four surveys (on the day subsequent to harvest) were conducted during July and August. Statistical Analyses I tested for a correlation between ripe E. oleracea fruit and the number of minutes that parrots spent in plots in 1997 using SAS JMPE^f (1996). The fruit removal experiment was a repeated measures factorial design with two between (site and treatment) and two within (pre-vs post-harvest and census) factors. Each of the eleven response variables (Appendix 3-2; Tables 1-9) was therefore used in a repeated measures ANOVA, conducted with Super ANOVA (Abacus Concepts 1993). If necessary, data were first transformed to achieve normality and equal variances. The experimental design, with only four replicates, had low power to detect true differences between treatments (Zolman 1993). To increase power, and reduce Type n error, I therefore chose 0. 1 as the initial significance value rather than the traditional value of 0.05. This value was adjusted, however, using a Bonferroni sequential technique (Rice 1989), because multiple (three) comparisons were conducted on each data set. In addition to testing, on the community level, the existence of a relationship between fruit and frugivore abundance, I sought to provide data on which to base sustainable fruit harvest guidelines. In other words, I wanted to determine which frugivore species were most sensitive to fruit harvest, and the threshold harvest intensity

PAGE 70

61 beyond which they ceased to visit the plots. To do this, I first used the data fi^om the 1998 fruit and fiugivorous bird surveys in logistic regression analyses using SAS JMPIN Software (1996). In logistic regression, a continuous independent variable (fruit abundance) is used to predict the state of a discreet, dependent variable (presence vs. absence of each frugivore species). I first performed the analyses to evaluate which species' presence or absence was significantly determined by the abundance of fruit ("fruit-sensitive species"). These analyses only included those visitations of > 5 min, to exclude birds that landed in plots but did not stay. I then used the logistic linear model to calculate, for each fruit-sensitive species, how much fruit would be required for a 25, 50, 75, and 99% probability of that species visiting a plot. Second, I tested for relationships between the amount of fruit available and the number of minutes that each species spent in subplots by regressing the number of minutes against kg fruit available, after both variables were log-transformed. In practical terms, managers could use these data to establish the minimum amount of fruit that should remain in the forest to ensure, within a chosen probability, the persistence of a certain species. Results Parrot Use of Palm Forest Versus Non-Palm Forest Pssitacids apparently use E. o/eracea-dominated forest to a greater extent than non-palm forest, and this use is associated with ripe palm fruit abundance (Fig. 3-3a). In 1997, fruit abundance was greatest at the initial survey and declined steadily through late August, as did the length of parrot visits. I found a positive correlation between ripe fruit abundance and the number of minutes parrots spent in palm forest plots (r^O.42,

PAGE 71

62 Fi,i2=8.4, p<0,01). The number of minutes that parrots spent in non-palm forest showed no such temporal trend and was consistently lower than in the E. oleracea forest (Fig. 33b). B 1 o u u 350 300 250200150100 50-1 0early July fruit abundance -90 -80 -70 -60 ta cr c -50 3 a ta -40 3 n n -30 -20 -10 -0 late July early August late August b. , T , late July early August Timing late August Figure 3-3. Censuses in 1997. Shown are mean + s.e. a) Fruit abundance (kg) and parrot visits (min) were positively correlated in 1 ha plots at four E. oleracea -dominated forest sites. b) Parrot visits (min) in 1 ha plots at four non-£. oleracea forest sites.

PAGE 72

63 Frugivore Responses to Experimental Fruit Harvest Fruit availability As in 1997, fruit production in 1998 peaked in July and declined through August (Fig. 3-4a). There were no significant differences in ripe fruit availability among treatments in the pre-harvest phase (F 2,18=0.95, p=0,44; Fig. 3-4, Table Al). Fruit harvest significantly reduced fruit abundance (Figs. 3-4a, b & c). Low-removal treatments had on average 41% less fruit than control subplots (Fi,i8=4.15, p=0.09), whereas highremoval treatments had on average 75% less fruit (Fi.i8=10.52, p=0.02). Low-removal and high-removal treatments did not differ significantly in their fruit availability (F,,8=1.46, p=0.27). Birds: community-level responses Forty-one species of frugivorous birds visited plots during 1998 (Appendix 3-1). Their responses to fruit harvest are indicated by significant "harvest x treatment" interaction terms in the repeated-measures ANOVA (Tables A1-A3; Appendix 3-2). Prior to fruit harvest, no differences existed (Fig. 3-5) among treatments in the number of frugivorous species (F2.i8=0.34, p=0.72), individuals (F2,,8=0.13, p=0.88), or their visit lengths (F2.,8=0.04, p=0.96). Once fruit harvest began, however, the number of species declined by 25% (F,,,8=13 .08, p=0.01), the number of individuals declined by 29% (F,,,8=34.96, p=0.001), and frugivores spent 68% less time (F,.,8=22.50, p=0.003), in the high-removal subplots, compared with the controls (Fig. 3-5). Moreover, highremoval treatments had significantly fewer species (F,.,8=l 1.60, p=0.01), individuals (F,,,8=70.61, p=0.0002), and visitation times (F,.,8=22.24, p=0.003) than the low-removal treatments (Fig. 3-5). In contrast, low-removal subplots showed no significant

PAGE 73

64 Figure 3-4, Averaged, oleracea fruit abundance in four 0.6 ha sites, over time. Censuses in May and June are pre-harvest, those in July and August are post-harvest. Shown are mean + s.e. a) Control. b) Low-removal: 4-56% less fruit than controls. c) High removal. 43-75% less fruit than controls.

PAGE 74

65 pre^arvest CO Q) •H u a w o -H > -rH O -H 6 10-1 8 6420 a. 2520151050w •H Q T3 •H 900-1 6003001 1 1 control low Ia post-harvest a > i high control Treatment low b high Figure 3-5. Effects of E. oleracea fruit harvest on frugivorous birds. Shown are mean ± s.e. Columns labeled with different letters differed significantly (p < 0.05). No differences existed pre-harvest. a) Number of species: post-harvest, high-removal treatments had 25% fewer species than controls. b) Number of individuals: post-harvest, high-removal treatments had 29% fewer individuals than controls. Low removal treatments had slightly more individuals. c) Visit lengths: post-harvest, visits to high-removal treatments were 69% shorter than to controls.

PAGE 75

66 differences from controls in the number of species (F,,i8=0.04, p=0.84), or visit durations (F,.,8=0.01, p=0.98), but had more individuals (F,,,8=6.20, p=0.047). I recorded 40 species of non-frugivorous birds in the plots (Appendix 1). Nonfrugivores showed no differences in number of species, number of individuals, or visitation times among treatments (Fig. 3-6) either before (all p's>0.51) or after (all p's>0.73) fruit harvest (Tables A5-7). The repeated-measures ANOVA's revealed several significant main and interaction effects in addition to the "harvest x treatment" interaction. As these effects are not central to the study questions, they will be reviewed here, but not discussed in detail. Numbers of both frugivorous and non-frugivorous birds varied spatially and temporally, as indicated by the significant effects of site, harvest, and time (Tables AlA6). The significant "site" effect indicates variation among my four replicate sites in the abundance and behavior of birds. Fazenda and Miriti consistently had more individuals of frugivorous birds that spent more time than did Moreira and Plaquinha. The reverse pattern was true for non-frugivorous birds, which regularly were represented by more species at Moreira and Plaquinha than the other two sites. In several of the repeated-measures models "harvest" was a significant effect (Tables A1-A6). This factor compares the responses of birds in May and June (the preharvest phase) with those in July and August (the post-harvest phase), independent of treatment. This effect of harvest on the number of frugivorous bird species and individuals depended, however, on site (significant harvest x site interaction). At Fazenda and Miriti, both frugivorous and non-frugivorous birds were more abundant and

PAGE 76

67 7-| 6543210-pre-harvest fi post-harvest -i140 ^ 120 •| 10080c. Treatment Figure 3-6. Effects ofE. oleracea fruit harvest on non-frugivorous birds. Shown are mean + s.e. No significant differences existed among treatments either pre-or postharvest. a) Number of species. b) Number of individuals. c) Visit lengths.

PAGE 77

68 paid longer visits to sites in July and August than in May and June, whereas at Plaquinha and Moreira, the abundance of birds did not vary much between the preand post-harvest phases. The factor "time" was significant in all of the models. This factor explains variation among the four censuses within each harvest phase (i.e., early May, late May, early June, etc.). In five of the six ANOVA's, the effect of time depended upon "harvest" (significant harvest x time interaction). In these cases, the effect of time was important in the pre-harvest phase, but not important in the post-harvest phase this is likely due to several factors. First, as fruit production began and progressed in May and June, the number of frugivores increased. In the post-harvest phase however, fruit abundance was kept at a more constant level by the experimental manipulations, and the number of frugivores overall in plots also stayed more constant. Non-frugivorous bird abundance increased also over time in the pre-harvest phase but did not fluctuate in the post harvest phase. Non-frugivores likely responded to other factors, such as water levels or their breeding cycles. In two of these five ANOVA's, (frugivorous bird species and frugivorous bird individuals) the interaction between harvest and time also depended on site (significant harvest x time x site interaction). This can be explained by the fact that the number of frugivores increased dramatically over time within the pre-harvest phase at Fazenda and Miriti, but not at Plaquinha and Moreira. The increase in bird abundance may be due, in part, to the arrival of species that immigrate from other areas in Amazonia.

PAGE 78

69 Birds: species-level responses In addition to evaluating community-level responses to fruit harvest, I used three methods to identify "fruit-sensitive species" those whose presence and/or visitation times were affected by fruit abundance. First, I simply counted the species that no longer visited high-removal and low-removal subplots, once fruit harvest began. Of the regular visitors, six species ceased visits to the high fruit -removal sites. These included a tanager {Thraupis palmarum), a trogon (Trogon coUaris), a parrot (Deropteus accipitrinus), two parakeets {Pionites leucogaster and Brotogeris versicolorous), and a macaw {Ara macao). Only one species, Trogon viridis, stopped visiting low-removal sites. Second, logistic regression revealed that the presence of six of the 20 species analyzed could be predicted by the abundance of fruit (Table 3-1). The probability of these six highly frugivorous species occurring in plots increased as fruit abundance increased, and these are the species most responsible for the lower species richness in high-removal treatments. Indeed, two of them {Ara and Pionites) are among those that ceased visits altogether. The most sensitive to fruit availability was Rhitipterna simplex, requiring 62 kg and 869 kg fruit for a 25 and 99% chance of occurring, respectively. The least sensitive was Vireo olivaceus, requiring 0 kg and 350 kg fruit for a 25 and 99% chance of occurring, respectively. Finally, I used linear regression to evaluate the relationship between the length of frugivore visits and the abundance of fruit. The length of frugivore visits was less related to fruit abundance than was frugivore presence or absence (Table 3-2) Only three species {Pionites leucogaster, Ramphastos tucanus, Vireo olivaceus) demonstrated a

PAGE 79

70 •c CO > c (U ^ p , TO (U 'C •a nS c > S c ? a, 1^ W CO i-i U IS ^ CO ^ 8 C > to « 4) O c c 00 o a, a — -o "O — OS 0\ 'u u O O >. iH o ;£ O o. C u u u w On 00 (N o. a u •a i>
PAGE 80

71 xn U 3 C 3 Q. <^ .t; ^ C/3 If C/3 nn4 y VI in tn v> tn tn tn c W d 8 CP V V o> <<{< s: 3 •c oc fM 00 NO NO a\ % i?) fo 1/-, oo 00 VO — « oo Tf m so fN oo m d d 00 o ON -* fN so t^' o o 00* •c e<; 0\ fN o fN On so ON fN SO OS o fN 00 00 m 00 Os SO 00 q fN 3 d 00 iri OS OS d fN d 00 fN s: + + + + 00 + + + + -(+1 + -)+ o VO o + + T + ST! m so d cn rso 00 O o On o OO OS m O o if-i OO tN ON Os 0\ fN SO so' r*^ fN (*i r-i On so fN nher 5 «N £! a oo Os o CnI SO fN so ON fN fN f^ 00 (N fN oo rso On Os fN o O so SO f^ fN O d so' d + 1 + 1 +1 + 1 rt' +1 + 1 +1 + 1 +1 + 1 + 1 + 1 +1 ON ON + 1 fN + 1 r-so fN m 00 so so VD in oo o 00 ON v~) r-ON d (N © so oo' O iri vd d d fN d i: c o lorus s c »^ mplex to azoni rsico \>itteli 52 to illis US <^ tes leuc > tn Eleania /7ai nulus 1^ <^ s; .s: US eel 1 Ramphoceh CI a Ara Si Amaz Pyrrh Pioni Broto Ramp t 5S a; Celeu § Rhitip 5S Turdu Cacici

PAGE 81

72 significant response to fruit abundance. Because these three were among the most frequent visitors, the community as a whole showed a strong response to fruit harvest (Fig, 3-5c). Mammals Five species of mammals known to eat E. oleracea fruit were observed in the plots: Guianan squirrel {Sciurus aestuans). South American coati (Nasua nasua), brown capuchin monkey (Cebus apella), red-handed howler monkey {Alouatta belzebul), and golden-handed tamarin (Saquinus midas). As with birds, a significant response of mammals to fruit harvest is indicated by a significant "harvest x treatment" interaction. Similar to the observed responses of birds, mammals showed no differences in the number of species, individuals, or visitation times among treatments in the pre-harvest phase (Fig. 3-7, Tables A7-A9). Post-harvest, however, the number of mammal species was 58% lower in the both high-removal (F,.,g=9.59, p=0.02) and the low-removal (F,,,8=9.59, p=0.02) treatments, compared with the controls; the lowand high-removal treatments did not differ (F,,,8=0.01, p=0.99). The lower species richness in high-removal subplots reflects the lack of visits there by both howler monkeys and tamarins once fruit harvest began. Howler monkeys, but not tamarins, also stopped visits to low-removal treatments after fruit harvest; squirrels and capuchin monkeys made less frequent visits post-harvest to low-removal treatments, which resulted in the lower species richness. Too few observations of coatis were made to detect a pattern for that species. Unlike species richness, the number of individuals and their visit times of mammals did not differ among the treatments after fruit harvest (Fig. 3-7, all p's>0.09). As with birds, mammal species (Fi.ig=18.78, p=0.005) and individuals (Fi.i8=29.26, p=0.043) showed an overall significant response to harvest, independent of

PAGE 82

73 1-| 0.8•3 0.6^ 0.4:| 0.20-pre-harvest P a post-harvest 1 5-1 43210 1 ^ 80!5 60> 40200-Treatment Figure 3-7. Effects of £. oleracea fruit harvest on fruit-eating mammals. Shown are mean + s.e. Columns labeled with different letters differ significantly (p < 0.05). No differences existed pre-harvest. a) Number of species, post-harvest, both highand low-removal treatments had 58% fewer species than controls. b) Number of individuals. c) Visit lencths.

PAGE 83

74 treatment (Tables A7 & A8). I found both more species and individuals in all three treatments post-harvest (Fig. 3-7), when overall fruit availability in plots was greater. This effect varied, however, among the sites, being stronger at Fazenda and Miriti than at Plaquinha and Moreira. Bruchid beetles Beetle emergence showed no differences among treatments, post harvest (Fig. 38, F2.150 = 0,346, p=0.71). The number of eggs oviposited was most affected by site (F3, 15c = 2.90, p=0.036), with the greatest emergence rates at Plaquinha and the lowest rates at Fazenda. control low high Treatment Figure 3-8. Effect ofE. oleracea fruit removal on the number of eggs oviposited by bruchid beetles on E. oleracea seeds. No significant differences were found among low-removal, high-removal and control sites. Shown are mean + s.e.

PAGE 84

75 Discussion Frugivores responded strongly to both natural and experimental declines in palm fruit abundance. When intensive fruit extraction by people was mimicked (75% ripe fruit; high-removal treatment), frugivorous bird species richness, abundance, and visits, and frugivorous mammal species richness, were significantly reduced. Removing 41% of the ripe fruit (low-removal treatment) had no such effects. That non-frugivorous birds showed no differences among treatments is additional evidence that frugivores responded to fruit abundance rather than to other factors. Although these effects were tested on a small scale relative to the home ranges of most frugivores, they may occur at larger scales since human fruit harvest is both extensive and intensive throughout the region. Frugivore Responses to Fruit Harvest Birds communitv level Spatial and temporal fluctuations in tropical frugivore species richness, abundance, and visit lengths have long been noted (Davis 1945, Fogden 1972, Karr 1976, Terborgh 1977). These have been linked to fruit abundance at various scales (Snow 1962a,b, Levey 1988, Loiselle and Blake 1993, Wright et al. 1999), and several studies have demonstrated that frugivores apparently track fruit abundance across space and time (Loiselle and Blake 1991, Powell and Bjork 1995, Rey 1995, Kinnaird et al. 1996, but see Herrera 1998). Nevertheless, many other factors (e.g., weather, breeding cycles) also influence the abundance and distribution of birds in forests (Karr 1976, Terborgh 1977, Karr et al. 1990), and these factors are often not considered in studies of fruits and frugivores. While fruit abundance is frequently implicated as the factor driving frugivore changes in abundance and movements, this idea has not yet been tested experimentally in a natural system.

PAGE 85

76 This study is apparently the largest controlled manipulation of arboreal fruit abundance, to date. Along with manipulations of fruit abundance on the forest floor (Adler 1998, Sherman and Eason 1998), it provides strong evidence that not only are abundances of fruits and fnigivores linked, but that fruit is the mechanism behind frugivore responses. If the strong influence of fruit abundance found in these studies represents a general pattern, then fruit abundance may help explain frugivore species diversity, abundance, and behavior across many scales: in foraging flocks (Chapman et al. 1989, Develey and Peres 2000) and individual fruiting trees (Howe and Vande Kerckhove 1980), within (Loiselle and Blake 1993) and among (Levey 1988) understory habitats, across regions (Rey 1995, Levey and Stiles 1992), and through time (Martin and Karr 1986, Bronstein and Hoffmann 1987). Frugivore communities not only had lower species richness and number of individuals in areas of high fruit removal; the species comprising the communities also spent 68% less time there. The length of individual visitations among all species ranged between 0. 13 65 min. These visit lengths are similar to those found for frugivores in individual trees in both Papua New Guinea (Pratt and Stiles 1983) and Costa Rica (Wheelwright 1991). The amount of time that frugivores spend in fruiting patches has rarely been examined directly at levels above individual trees, although Sargent (1990) found that as the amount of fruit on both fruiting plants and in fruiting neighborhoods declined, so did visitations by both flocking and non-flocking bird species. My results differ from those of Galetti and Aleixo (1998) who studied the impacts of reduced Euterpe edulis palm fruit abundance resulting from harvest of trees for heartof-palm. These authors found no difference in the abundance of frugivorous birds in

PAGE 86

77 harvested vs. unharvested forest, and argue that because £. edulis produces fruit concurrent with many other plant species, the 96% reduction in palm density in harvested forest represented a small overall loss of fruit. Clearly, such a large reduction in fruit abundance would induce large responses from frugivores in my Amazonian site. One explanation for the differences in outcome between these two studies could be differences in the spatial and temporal scales of frugivore censusing:Galetti and Aleixo censused birds in a much larger area from which palms had been harvested 5 10 yrs previously . In addition, fruit fall rather than fruit production was measured; fruit fall does not always represent the amount of fruit available to frugivores (Chapman et al. 1994). The lag time between harvests and censuses, and the lack of data on fruit production, may have made finding a relationship between fruit and frugivores inherently unlikely. Alternatively, the larger scale at which they censused may indicate that the small-scale differences I measured do not hold at larger scales (but see Rey 1995). Although I consider it unlikely, frugivores may have responded to the human activity associated with my fruit harvests rather than to the reduction of fruit. In reality, there were two differences between the experimental treatments and the controls: tree climbing and fruit removal. Many of the frugivores observed in this study certainly are sensitive to human activity and would not remain in an area while fruit harvest was occurring. Others, such as Amazona parrots, are relatively insensitive to human activity at my sites. Because fruit harvest occurred a full day prior to the frugivore surveys, however, I believe that any disturbance, aside from the odor of humans (which could have affected mammals but not birds) caused by the activity would no longer have been present.

PAGE 87

78 Birdsspecies level The lower species richness in high-removal treatments has at least three components. First, ten "fruit-sensitive" species, defined as those that either discontinued visits to high-removal sites (six species) or whose visits depended, according to logistical regression, on fruit abundance (4 additional species), account for the 25% reduction in species richness in high-removal sites. A second factor, however, is the lower consistency with which other frugivores visited those subplots. Some species, such as Turdus albicollis and Pipra rubrocapilla, continued to visit high-removal sites, but did so more sporadically once fruit harvest began. Lowered visitation rates could result if reduced fruit availability caused by harvest forced frugivores to search a wider area for food (Chapman et al. 1989, Fleming 1991). A third component of lower species richness in high-removal treatments is the fact that fewer individuals were visiting those sites. With such a diverse frugivore community (41 species), the observed 29% reduction in frugivorous individuals would almost certainly lead to the loss of some species, just by chance alone. What determines which birds visit particular areas and how long they spend there? The most obvious factor is degree of frugivory; both Levey (1988) and Loiselle and Blake (1993) showed that the abundance of manakins, the most heavily fmgivorous birds in their study sites, was correlated with small-scale fruit abundance. In Papua New Guinea, Pratt and Stiles (1983) found that highly frugivorous fruit pigeons spent more time m fruiting trees than did partially frugivorous bowerbirds, which spent more time than did birds of paradise, which eat fruit only occasionally. I found that the presence of six species, and the visit lengths of one additional species {Ramphastos tucanus), were predicted by the abundance of fruit in 0.6 ha subplots. All of these species include

PAGE 88

79 substantial amounts of fruit in their diet, but probably not more so than the frugivorous species that did not respond to fruit abundance. In other words, the responses of individual species do not appear to be determined solely by their degree of frugivory, a result also found by Wheelwright (1991). As Levey (1988) suggested, fruit abundance may not be equivalent to fruit availability, because competetive interactions may limit access to fruit for some species. Toucans, for example, are known to supplant other bird species feeding on the same trees (Bourne 1974). In addition to the abundance of their primary food source, the abundance, distribution, and behavior of birds may be influenced by alternative food abundance, crypsis, body size, and social systems (Howe 1979, Pratt and Stiles 1983, Chapman et al. 1989, Wheelwright 1991). Of the birds recorded in this study, only five qualify as highly frugivorous {Pipra rubrocapilla, Gymnoderus foetidua, Cotinga cayana, Cotinga cotinga, Querula purpuratd), the others commonly consume alternative foods, especially insects. I did not monitor insect abundance, but in other tropical forests it tends to be low in the dry season (Davis 1945, Karr 1976, Develey and Peres 2000), which is when my study took place. If insect abundance varied over the course of my study and among my treatments, inducing the differences among treatments in frugivores, then similar or stronger responses should have been detected for non-frugivorous bird species. Because neither the species richness, number of individuals, nor visitation times of non-frugivores differed among treatments, it seems unlikely that insect abundance played a role in frugivore abundance and distribution. Howe (1979) suggested that crypsis determines, in part, the amount of time frugivores spend in fruit patches, because colorful birds are more obvious and vulnerable

PAGE 89

80 to predation. In support Howe's hypothesis, Pratt and Stiles (1983) found that less conspicuously colored fruit pigeons spent more time in fruit trees than did more colorful bowerbirds and birds of paradise. On the other hand, Wheelwright (1991) found no such pattern in Costa Rica; time spent in fruiting trees by quetzals, bellbirds, toucanets, and robins was unrelated to plumage coloration (Wheelwright 1991). My results agree with those of Wheelwright; fruit-sensitive species included the dull-colored red-eyed vireo (Vireo olivaceus) and grayish mourner {Rhitipterna simplex) as well as the conspicuous white-fronted parakeet {Pionites leucogaster) and scarlet macaw {Ara macao). Body size may also affect time spent in fruiting trees, for two reasons. First, large birds, like colorful birds, may be more vulnerable to predation (Howe 1979, Pratt and Stiles 1983). Second, large birds generally require greater amounts of food, which are unlikely to be provided in a small area of forest (Fleming 1991). Larger birds might therefore be expected to move more and spend less time per tree or fruiting area. My results do not, however, indicate that body size greatly affected which species were sensitive to fruit harvest; fruit-sensitive species ranged from the 16 g red-eyed vireo Vireo olivaceus to the 1250 g red-and-green macaw Ara chloroptera. The reasons for this result may lie in the experimental design; despite being a relatively large-scale manipulation of fruit abundance, the area of my experimental treatments (0.6 ha) likely encompassed only a small portion of the home ranges of many or most frugivores in the community (Fleming 1991). Recent work (Westcott and Graham 2000) indicates that even small Mionectes (1 1 g) flycatchers can have home ranges >28 ha, and larger birds such as parrots and toucans are known to range over many kilometers per day (Bourne 1974, Snyder et al. 1987, Bjork 2000).

PAGE 90

81 The reasons underlying the 68% decrease in visitation times in high-removal sites may relate most to the behavioral attributes of the most numerous group of birds in the plots parrots, parakeets, and macaws. Eleven of the 41 frugivore species are members of the Pssitacidae, and their visits frequently accounted for >50% of the total communitylevel visit minutes. Throughout the neotropics, parrots rely heavily on palm fruit (Snyder et al. 1987, Abramson et al. 1995, Galetti et al. 1999), so their high abundance in E. oleracea forests is not surprising. Also, parrots move and respond to fruit abundance in groups (Chapman et al. 1989); a group of eight parakeets visiting a plot translated into many more visitation minutes than did a visit by a single individual of another species. Furthermore, pssitacids frequently vocalize upon arrival in a feeding area (Snyder et al. 1987). This behavior may serve to relay information about food resources to other individuals, attracting them to the sites of high fruit abundance and leading to higher visit times there. Mammals Both high and low intensities of fruit removal reduced the species richness of fruit-eating mammals in plots. The number of individuals and the time they spent, however, were unaffected by fruit harvest. Howler monkeys, tamarins, and squirrels showed the most significant responses to fruit harvest, apparently avoiding the areas from which fruit had been harvested. Of the three primate species observed in plots, howler monkeys are the most frugivorous at my study site, composing up to 71% of their diet with fruit (Jardim and Oliveira 1997, Pina 1999), whereas tamarins and capuchin monkeys have more mixed diets of insects and fruit (Chapman and Fedigan 1990, Peck et al. 1999). That these large-bodied mammals responded to fruit harvest treatments on a scale of 0.6 ha indicates their sensitivity to small-scale differences in food availability

PAGE 91

82 within their much larger home ranges. A similar result was found by Allen (1997), who found that agouti and tayra responded to differences in availability of Mauritia flexuosa (Palmae) fruit by removing a significantly greater proportion of fruit from areas with low fruit availability. Fruit abundance not only affects fruit-eating mammal behavior, it can also regulate fruit-eating mammal populations (Foster 1982b, Terborgh 1986b, Bodmer 1989, 1990, Adler 1998, Wright et al. 1999). Intensive harvest oiE. oleracea fruit over a large area could therefore have population-level impacts on frugivorous mammals. Bruchid beetles I presumed that bruchid beetles would have relatively small home ranges and thus would be sensitive to my experimental reduction in their sole developmental food source. Beetles could have responded to reduction in fruit abundance with either an increase or a decrease in the number of eggs oviposited. Increased oviposition would have been expected if competition for seeds existed, so that beetles had fewer seeds on which to oviposit in harvested areas (Siemens and Johnson 1996). Decreased oviposition would have been expected if fruit harvest reduced beetle populations, so that fewer females were available to oviposit (Wright 1990). Contrary to these predictions, beetle oviposition did not vary among treatments. This lack of response may have several explanations. First, bruchids may actually have larger ranges than initially suspected. Wright (1983) showed that bruchid oviposition rates on fallen fruits were similar at distances within 16 m of the trunk of fruiting trees; only at 100 m did rates decline. These results imply that bruchids search over wide areas for seeds beetles may thus have traversed the treatment subplots within my plots. A second potential reason that oviposition did not differ among treatments is the timing of the experiment. If fruit harvest limited the number of ovipositing females, this effect

PAGE 92

S3 would only begin to take place after 2 months of harvest, because beetles take >60 days to emerge from seeds (S. Moegenburg unpubl. data). As I only monitored the number of eggs on experimental seeds for 2 months, I may have missed a decline in the number of eggs oviposited due to a decline in the number of females, which my have been detectable later. A final reason for the lack of bruchid response may be the possibility that to them, E. oleracea fruit abundance did not decrease. The fruit harvest clearly reduced fruit availability to arboreal frugivores, but not necessarily to terrestrial granivores. Despite the fruit harvest, many seeds were dropped by frugivores who removed the pulp but left the seed intact. These seeds remained available to bruchid beetles, and may have been sufficiently available to prevent seed limitation for ovipositing females. Implications of Fruit Harvest for Frugivores and Fruiting Plants Few studies have evaluated the ecological impacts of non-timber resource harvest. Extraction is not limited to tropical fruits; the harvest (and associated ecological impacts) of non-timber resources occurs worldwide. One study in northeastern North America, for example, found that harvest of worms for fishing bait significantly reduced the foraging efficiency and prey base of semipalmated sandpipers (Shepherd and Boates 1999). Even fewer studies have attempted to determine harvest levels of non-timber resources that minimize ecological impacts. Using matrix models and life table analysis, Peters (1990) estimated that 80% of the fruit produced by Grias peruviana (Lethycidaceae) trees in Peru could be harvested without affecting regeneration of the species. Like E. oleracea, G. peruviana occurs in high-density stands and produces fruits

PAGE 93

84 eaten by a variety of frugivores. In Peter's study, however, the effects of an 80% fruit harvest on frugivores were not considered. As shown in the present study and by Shepherd and Boates (1999), resource extraction may impact animal communities in a variety of situations. In the case of tropical fruit, what are the implications of these impacts for frugivores, fruiting plants, and conservation strategies based on fruit extraction? Of the birds most sensitive to E. oleracea fruit harvest, at least one (scarlet macaw, Ara macao) is considered vulnerable to extinction from other causes (Parker et al. 1996). Several other species recorded in this study, or known to eat E. oleracea fruit in other regions, such as blue and yellow macaws {Ara araruana), green and red macaws (Ara chloroptera), and golden parakeets (Guarouba guarouba) may also be affected by fruit harvest (Sick 1993). This last species which, according to Sick (1993), favors the fruits from E. oleracea, is one of the most threatened pssitacids in the Brazilian Amazon (Oren and Novaes 1986). Parrots, parakeets, and macaws throughout the neotropics are threatened by habitat destruction and capture for the pet trade (Snyder et al. 1991); harvest of their foods by people should be added to the list of threats. Pssitacids were not the only group of frugivores to respond to fruit harvest, however. Contrary to conventional wisdom (Anderson 1990b, Peters 1992, Anderson et al. 1995), fruit harvest can have impressive impacts on biodiversity, reducing it, in the case of birds, by 25%, and in the case of mammals by 58%. Moreover, harvest substantially affected bird behavior; the birds that did visit high-removal treatments spent 68% less time there. These substantial effects may reflect the importance of £. oleracea

PAGE 94

85 in their diets if, as has been proposed for other palm fruits, it constitutes a "keystone" resource for them (Terborgh 1986b, Peres 2000). Reduction of frugivore diversity and visitations may cascade into indirect effects on other organisms. In particular, reduction in frugivore activity may affect the plants in this system, such as E. oleracea and Virola surinamemis, whose seeds are dispersed by frugivorous birds (Bourne 1974, DeSteven and Putz 1984, Strahl and Grahal 1991, Hamann and Curio 1999, Loiselle and Blake in prep.). On the one hand, less time spent in fruiting patches may actually increase dispersal of the seeds that do get ingested, because frugivores are forced to move farther in search of food (Pratt and Stiles 1983). On the other hand, less frequent and shorter visits can mean that fewer total seeds are ingested (Davidar and Morton 1986, Sargent 1990), and that many may fall directly below parent trees (Wheelwright 1991), In the case of oleracea, frugivores perform a critical function by removing pulp, because seeds cannot germinate without at least some pulp removal (S. Moegenburg unpubl. data). Likewise, dispersal of seeds by frugivores is critical for £. oleracea because lack of dispersal results in high mortality from insect seed consumers under the parent plant (S. Moegenburg unpubl. data). Most of the frugivores in this study range over many hectares, so fruit harvest at the level of this experiment likely had no, or few, impacts on them other than the visitation patterns observed Harvest of E. oleracea fruit occurs, however, over a very large area (up to 10,000 km^ of the Amazon estuary, Calzavara 1972) by many thousands of people. Furthermore, extraction is increasing in response to higher market prices for fruit and the emergence of a forest conservation strategy known as extractive reserves. These reserves, encompassing over 22,000 km^ mostly in the Brazilian Amazon, are

PAGE 95

86 collectively managed by local, forest-dwelling residents for the long-term sustainable use of forest resources, including fruits and other non-timber products (Feamside 1989, Allegretti 1990). What might be the effects on frugivores across this larger scale at which E. oleracea fruit is harvested? While my study shows that intensive fruit harvest at a small scale (0.6 ha) negatively affects frugivores, it has limitations in terms of predicting the impacts of E. oleracea fruit harvest at the larger scales at which people currently practice extraction. Fruit harvest across the home ranges of frugivores could have several types of effects. Specifically, in response to decreased fruit availability throughout their ranges, frugivore numbers may decrease through a functional response, if frugivores vacate the area, or through a numerical response, if frugivores suffer lower reproductive success. Alternatively, they may show neither a functional nor numerical response but rather simply switch diets. Evidence suggests that all three responses may occur, each by different members of the frugivore community. In other systems, for example, frugivorous birds apparently respond functionally to fruit availability by tracking ripe fruit over space and time. Such tracking behavior has been shown across altitudinal gradients (e.g., Loiselle and Blake 1991), latitudinal gradients (Martin and Karr 1986), and agricultural landscapes (Rey 1995). In systems in which frugivores cannot track ripe fruit availability (e.g., non-volant mammals on Barro Colorado Island, Panama), response to decreased fruit abundance is numerical, with frugivores suffering either mortality (Foster 1982a, Wright et al. 1999) or lowered reproductive success (Adler 1998). Finally, some partial frugivores and omnivores (usually the majority of fruit-eaters in the

PAGE 96

87 community) may switch to other diets when fmit becomes scarce (Loiselle and Blake 1993, Pina 1999). How might the responses of frugivores to large-scale fruit harvest be ascertained? Clearly, a controlled experiment on an adequate scale would be difficult. A potential route might be in an extractive reserve in which fruit is harvested and in which the level of human community organization would allow the orchestration of a large-scale quasiexperiment. In the Cajari River Extractive Reserve in Amapa State, Brazil, for example, several dozen families extract E. oleracea fruit from several thousand hectares. In such a setting it may be possible to request families to harvest from designated areas and not from others, which would establish areas of high and low fruit availability, in which frugivores could be monitored. As in this study, frugivores should be surveyed both preand post-harvest to increase the power of such a quasi -experiment. Such research needs to be done, on E. oleracea and other species of harvested fruit. Until it is, however, the data available from my study and others indicate that fruit harvest does affect frugivores, so it should be limited if biodiversity and ecological processes are to be conserved. Forests devoted to fruit extraction are apparently not equivalent, from the point of view of frugivorous birds and mammals, to forests from which fruit is not extracted. Fruit extraction can potentially play an important role in tropical forest conservation, as a long-term renewable source of income. However, forests dedicated to fruit extraction should not be seen as substitutes for non-harvested forests; rather, they should be part of conservation plans that also include non-harvested forests.

PAGE 97

CHAPTER 4 FALLING FRUITS AND FEEDING FISHES: EUTERPE OLERACEA FRUITS IN AMAZONIAN FLOODPLAIN FOREST NUTRIENT AND FOOD CYCLES Introduction Approximately 2 7% of the Brazilian Amazon exists as floodplain forest, which inundates via two hydrological processes of Whitewater, clearwater, and blackwater rivers (Ducke and Black 1953, Sioli 1966, Pires 1974, Prance 1978, Goulding 1980). The first process occurs throughout the basin during the wet season, when rainwater swells rivers, causing them to overflow their banks as much as 13 meters. This high water phase in socalled "igapo"^ forests can persist from 3-11 months per year (Pires 1974). The second, shorter-phase process is caused by tides from the Atlantic Ocean, which push river water upstream, causing it to spill into forests to depths of 2 3m. These flood pulses in "varzea" forests last for about six hours and occur approximately twice per day. When rivers flow into forests, fishes follow, finding detritus, invertebrates, and fruit, on which to feed. Fruits serve as food for at least 200 of the 2,500 3,000 species of Amazonian fish (Mariier 1967, Goulding 1980, Waldhoffet al. 1996), some of which consume little else during the rivers' high water phase. Some species, such as Colossoma macropomum, have evolved special dentition that allows them to crush even very hard pericarps (Goulding 1980). Top-swimming fishes gulp floating fruits and seeds, such as ' Alternative definitions for varzea and igapo exist. According to Pires (1974), for example, varzea forests are those flooded by turbid, or sediment-laden, water, while igapo is forest flooded by blaclcwater and Clearwater nvers. Neither set of definitions is more correct, either ecologically or biogeographically (Goulding 1980). My preference for the one used above reflects the definition used by Furch (1997) and many local people in the Amazon. 88

PAGE 98

89 those from the rubber tree, Hevea brasiliensis, while bottom dwellers, such as the catfish Lithodoras dorsalis, eat fruits and seeds that sink (Kubitzki and Ziburski 1994, Goulding 1980). Small fish that cannot ingest whole fruits bite off pieces of pulp from submerged fruits (S. Moegenburg pers. obs ). Just as in unflooded forests (Hladik and Hladik 1969), not all fruit that falls in flooded forests is immediately consumed. Rather, uneaten fruits remain floating or submerged (Kubitzki and Ziburski 1994) and begin to decompose. During this phase, fruits may play several roles in the floodplain ecosystem. First, they may continue to serve as food for fruit-eating fishes (Araujo-Lima et al. 1986, Henderson and Crampton 1997). Second, decomposing pulp and seeds may furnish nutrition for scavenging and detritous-feeding fish and invertebrates (Irmler 1975, Henderson and Walker 1986, Junk and Robertson 1997), Third, decaying fruits and seeds may release nutrients into the water to be taken up by phytoplankton or plants. Blackwater rivers and the forests they flood are extremely nutrient poor (Sioli 1968, Setaro and Melack 1984, Furch 1997), presumably recycling most nutrients through the 5,000 kg ha"' of leaf litter that falls per year and decomposes (Franken et al. 1979, Furch and Junk 1997b). Nutrient cycling through the 469 kg ha ' yr ' of fruits that fall in igapo forests (Waldhoffet al. 1996) has not been well studied. In igapo forests of the eastern Amazon, fruit production at many sites is dominated by the palm, Euterpe oleracea, which produces one-seeded, spherical drupes approximately 1 cm in diameter (Roosmalen 1985). E. oleracea is hydrophilic, occurring primarily along river margins and in both Whitewater and blackwater varzea and igapo. Unlike many other floodplain diaspores, however, neither the fruits nor seeds of E

PAGE 99

90 oleracea float (Roosmalen 1985). Many animals, including primates (Roosemalen 1985), birds (Moegenburg and Levey in prep ), and fish (Goulding 1980, WaldhofF et al. 1996) consume E. oleracea fruits. In addition, people harvest the fruits to make a nutritious drink (Strudwick and Sobel 1988). In intensively managed and harvested areas, fruit removal can approach 13,000 kg ha ' yr' (Muniz-Miret et al. 1996). Fruit extraction may thus remove a significant source of food for fish and invertebrates, and a source of nutrients for floodplain forests. I studied two aspects of E. oleracea fruits in floodplain systems. First, to evaluate fruits as nutrient sources in floodplain forests, I determined the nutrient content of fresh E. oleracea fruit and the total amount of N and P in E. oleracea fruits per area of forest. Second, to test the effect of human harvest of E. oleracea fruit on aquatic animals, I compared the species richness and abundance of aquatic invertebrates and fish in areas subjected to high-intensity fruit harvest, low-intensity fruit harvest, and no fruit harvest in an experimental manipulation. Methods Study Species and Sites The cespitose E. oleracea ranges throughout Amazonian South America, along the Pacific coast of Colombia and Ecuador, and in Trinidad (Henderson 1995). Individual genets (hereafter "trees" for simplicity) contain up to 25 slender stems that reach heights of 30 m (Henderson 1995). Reproductive stems produce infructescences bearing several thousand purple-black fruits. Fruits contain approximately 37% water (Waldoffet al. 1996). E. oleracea pericarp contains 8.1 and 15.3 % crude protein and fat, respectively. E. oleracea "seed" (presumably seed + endocarp) contains 7.2 2.5, and 8.1% crude protein, crude fat, and cmde fiber, respectively (Waldhoff et al. 1996).

PAGE 100

91 This study was carried out from June September, 1998 at the 33,000 ha Estafao Cientifica Ferreira Penna (Ferreira Penna Sciemific Station; r42'30" S, 51°31'45" W), operated by the Museu Paraense Emilio Goeldi of Bel em and located within Caxiuana National Forest in the municipality of Melgafo, Para State, Brazil (Fig. 4-1). Average rainfall is 2,500 3,000 mm, mean annual temperature is 26C, and mean annual relative humidity is 85%. The vegetation is evergreen humid rainforest, and the majority of the station is non-flooded, terra firme forest (Lisboa et al. 1997). Approximately 3,300 ha, mostly along the blackwater Bay of Caxiuana, is extremely low-lying forest that inundates when the river levels rise during the rainy seasons (December May) and high tides. Water depths reach their maxima (ca. 1 m) in May. Much of this floodplain forest is dominated by E. oleracea, with Virola surinamensis (Myristicaceae) znd Pterocarpus santalinoides (Fabaceae) also common (Ferreira et al. 1997). Although there are no homes and no current management of this forest, several lines of evidence suggest that the area of floodplain forest dominated by E. oleracea (hereafter "palm forest") resulted from past human management. First, areas of highest £. oleracea density occur closet to the river margin, where human dwellings tend to sit. Second, local people commented that the names of some of the sites (e.g., Moreira, Fazenda) reflect previous uses or names of inhabitants. E. oleracea fruit begins to appear in these sites in May and persists until September. In most years people occasionally extract £. oleracea fruit from these areas; however, in 1998 local inhabitants respected my request to refrain from extraction. I worked in four blackwater river flooded forest sites (hereafter "palm forest") located approximately 1 km apart along the bay. The water level in all of the sites

PAGE 101

Figure 4-1. Study site within Brazil, and study plots within sites.

PAGE 102

93 fluctuated, but the magnitude of the flood differed among sites. One site, "Moreira", was permanently flooded with 8 40 cm water. The other three sites ("Plaquinha", "Fazenda", and "Miriti") flooded during high tide phases (new and full moons) but often dried out near the first quarter and last quarter of the moon. Water depths in these three sites varied between 0 30 cm. In each site I set up a 1 .8 ha circular plot (diameter 1 50 m), located roughly in the center of highest density of E. oleracea (Fig. 4-1). Each circular plot was divided into three equal-sized subplots, which were demarcated with colored flagging. Fruit Production To determine the amount of nutrients released from decomposing fallen fmits, one first needs to know total fruit production. This requires estimates of E. oleracea tree density, the number of reproductive stems per multi-stemmed tree, the number of infructescences produced per stem, the number of fruits per infructescence, and the mass of individual fruits. The density of palm trees and reproductive stems were counted along a 20 X 60 m belt transect in each subplot (Fig. 4-1). These belt transects began at outer plot edges and bisected each subplot stopping 15 m short of the plots' center to avoid transect overlap. Reproductive stems were identified either by the presence of flowers or fruits or by scars left from previous infructescences. To determine the number of infructescences produced per year per stem, I monitored fruit production by six multiple-stemmed trees each at Plaquinha, Fazenda, and Miriti, for a total of 18 trees. These trees were randomly selected from among those occurring in "control" subplots (see below). Each reproductive stem in each tree was marked with colored flagging and all inflorescences and infructescences recorded. The

PAGE 103

94 18 trees contained a total of 57 reproductive stems, with a range of 1-7 reproductive stems per tree. The total number of reproductive stems per plot was calculated as: 7 ^[i^ trees/ plot * pt] where /?, = the proportion of trees with i reproductive stems. Not only doE. oleracea trees vary in their number of reproductive stems, but also in the number and sizes of infructescences produced per stem. The 57 stems on the 18 monitored trees produced 0-4 infructescences during 1998. Using these data I calculated the proportion of reproductive stems producing 1, 2, 3, and 4 infructescences per year as 0.44, 0.39, 0.11, and .06. The number of infructescences produced per plot was then calculated as: #reproductivestems/ tree * pt] i=I where/?, = the proportion of stems producing i infructescences. All infructescences on the 57 stems were classified as small, medium, or large, based on a predetermined visual scale. In addition, I counted fruits from harvested small, medium, and large infructescences, which contained 950 + 590, 2500 + 969, and 3050 + 1039 (mean + Is.e.) fruits, respectively. These known quantities of fruit were then weighed to obtain a conversion factor between infructescence size and wet mass of fruits. The proportion of infructescences of each size was calculated, and the number of each size was then calculated as: 3 ^ [# inf ructescences I stem * /?.]

PAGE 104

95 where p, = the proportion of small, medium, and large infructecences. Nutrient loss from fruits I evaluated nutrient content of fresh fruits (including exocarp, mesocarp, endocarp, and seed; hereafter "fruits") and rates of nutrient loss from decomposing fruits. I focussed on N and P, because these tend to be the most limiting nutrients in tropical forests (Vitousek 1984, Furch 1997) and because they are most often quantified in leaf decomposition studies, to which I wished to compare my results. To quantify initial N and P content, freshly-harvested fruits were dried to constant weight under heat lamps and preserved in sealed plastic bags. Prior to analysis, fruits were ground using a coffee grinder fitted with a stainless steel blade and cup. For both nitrogen and phosphorus analysis, samples were digested using a modification of the aluminum block digestion procedure of Gallaher et al. (1975). Sample weight was 0.25 g, catalyst used was 1 .5 g of 9: 1 K2S04:CuS04, and digestion was conducted for at least 4 h at 375°C using 6 ml of H2SO4 and 2 ml H2O2. Nitrogen and phosphorous in the digestate was determined by semiautomated colorimetry (Hambleton, 1977). To evaluate nutrient loss from decomposing fruits I conducted a fruit decomposition study modeled after leaf decomposition studies (Dickinson and Pugh 1974). Batches of 50 freshly harvested fruits were weighed and then placed inside 13 x 13 cm bags made of 1 mm nylon mesh, which were then sewn shut. All bags were submerged under approximately 25 cm water at Moreira on June 26, 1998. Water flow through the site was not perceptible, and it remained inundated for the duration of the study. Bags were scattered haphazardly on top of litter and did not overlap each other. Every two days, for 20 days, and on August 6 and 22 and September 7, three bags.

PAGE 105

96 serving as replicates, were retrieved and opened. Recovered fruits were preserved and analyzed as above. Responses of aquatic animals to fruit harvest bv people To evaluate responses of aquatic invertebrates and fish to fruit extraction by people, I conducted a fruit removal experiment at the sites used for fruit production estimation. At each of the four study sites, subplots were randomly assigned an experimental treatment. High fruit removal treatments mimicked fruit harvest carried out by people who both consume and sell fruit, and therefore harvest all ripe fruit within reach of people who climbed the stems (43 75% fruit removed). The low fruit removal treatment, on the other hand, mimicked harvest for subsistence use only (4 56% fruit removed). At each site one subplot was left unmanipulated as a control. I assume that reduced abundance of fruit in the canopy resulted in reduced abundance of fruit on the submerged forest floor. Twice in July and twice in August, 1998, fruit was harvested and aquatic animals surveyed, using two methods. On the day of fruit harvest, two locally made shrimp traps (known as "matapi") were submerged near the center of each subplot. Traps were made of long strips of arum fibers lashed together to form a cylinder approximately 50 cm in length and 25 cm in diameter. Fibers attached to the cylinder ends formed inverted cones, with openings of ca. 5c m through which animals could enter. The lengthwise fibers were spaced ca. 0,5 1.0 cm apart, allowing small animals to escape. One trap was baited with E. oleracea fniit; the other with a locally-used bait of powdered Orbignyaphalerata (Palmae) seeds. After two days contents were sorted and later identified. The second method used to capture aquatic animals was a 25 x 45 cm dip net made of 500 Tim nylon netting. On the day following fruit harvest, the net was used to

PAGE 106

97 scoop water and litter four times near the center of each subplot. The area sampled in each subplot was approximately 4,500 cm'. To scoop, the net was held at an angle (ca. 45°), plunged into the water and litter, and then quickly lifted. Net contents were washed into buckets, sorted, and later identified. Aquatic invertebrates were sorted into morphospecies. Captured fish were either collected, if their identity was uncertain, or recorded and released if their identity was known. Voucher specimens of both invertebrates and fish were preserved in a 10% formalin solution. Fish were identified at least to genus and to species whenever possible, whereas invertebrates were identified to order. To evaluate their responses to fruit harvest, I included only herbivorous, omnivorous, and detritivorous species and excluded carnivorous species. Statistical analyses For both invertebrates and fish, neither the raw nor transformed data were distributed normally with equal variances, precluding the use of parametric statistics. Variances were high both among replicates and among sampling periods, due to the apparent patchy distribution of animals and the fluctuating water levels. On one occasion, for example, 20 electric eels (Electrophoridae) were caught in one fish trap; on other occasions few or no fish were caught. The question of interest was whether the experimental subplots had similar or different numbers of invertebrate and fish species and individuals, that is, whether the numbers in one treatment were consistendy higher, lower, or the same as the numbers in the other treatments. To analyze the data, I therefore used a Friedman's analysis of variance by ranks (Marascuilo and McSweeney 1977, Zar 1996), in which the four replicates (sites) are treated as blocks, and the data are ranked within blocks. For both invertebrates and fish, I therefore had AT = 3 treatments, B

PAGE 107

98 = 4 blocks, and n = 4 samples per treatment per block. The statistic (with df = K-1) was then calculated as: X'= 12 XRk^-35(n^+ 1). BKn^-inK+\) where Rk are the ranks of the K treatments. Results E. oleracea fruit production Euterpe oleracea tree densities varied from 228, 272, 330, and 486 trees ha"' at Moreira, Miriti, Plaquinha, and Fazenda, respectively (mean + s.d. = 329 + 98). The number of reproductive stems per tree varied from 1 7, and the number of infructescences per stem ranged from 1 4. Most (ca. 51%) infructescences were small, approximately 31% were medium, and 18% were large. Total annual production of palm fruit varied from 2,456 kg ha ' (4,421 kg plot ') at Moreira, the site with the lowest density of£. oleracea, to 5,236 kg ha ' (9,425 kg plot ') at Fazenda, the site with the highest density of palms. Fresh E. oleracea fruits contain ca. 37 41% water (Waldoff et al. 1994, S. Moegenburg unpubl. data), so the dry mass of fruit produced at Moreira and Fazenda was 1,456 and 3,104 kg ha ', respectively. Nutrient loss from fruits Ripe, freshly-harvested E. oleracea fruits have a dry mass of 0.87 + 0.03 gm, of which approximately 77% is seed. Fresh fruits, including seeds, are 0.77+0.02% N, 0.05+0.003% P, and 98.29% organic matter (dry mass). Both mass and nutrients are lost rapidly from submerged fruits (Fig. 4-2). After 73 days, submerged fruits lost 22.1% of their dry mass, 20.0% of their N, and 7.3% of their P. If all fruits at Moreira, the site with

PAGE 108

99 Days since submersion Figure 4-2. Loss of mass (a), N (b), and P (c) from submerged E. oleracea fruit over 3.5 months. Plotted are means ± Is.e. of three replicates, each containing 47 fruits.

PAGE 109

100 lowest E. oleracea tree density and fruit production, decomposed, this would recycle 1 1 .22 kg N and 0.76 kg P ha"' yr ' within the system. At Fazenda, the site with highest E. oleracea tree density and fruit production, total N and P cycled through fruit would be 23.93 and 1.62 kg ha"' yr"', respectively. Responses of aquatic animals to fruit harvest by people I identified 704 invertebrates of 52 morphospecies in approximately 17 orders (Table 4-1). Most invertebrates (41%) were detritivorous shrimp in the families Palaemonidae and Euryrhynchidae, followed by larvae of beetles (Coleoptera) and bugs (Hemiptera), many of which are omnivorous (Junk and Robertson 1997). In addition to these animals, I captured predatory Trichoptera, Plecoptera, and Odonata larvae. Invertebrates were apparently unaffected one day after fruit harvest (Fig. 4-3). The number of non-carnivorous species did not differ among treatments (Fig. 4-3a; = 1 .39, df = 2, p > 0. 10), nor did the number of non-carnivorous individuals (minus shrimp, Fig.4-3b; f = 0.87, df = 2, p > 0.05), or the number of shrimp (Fig. 4-3c; = 5.93, df = 2, p > 0.05). I captured eleven species of fish (Table 4-2), three of which I saw eating E. oleracea whole fruit (Electrophorus electricus and Haplosternum thoracatum) or exocarp/mesocarp (Pyrrhulina sp ). Most species were omnivorous detritivores, although two carnivorous characids were also captured. The herbivore Pyrrhulina sp. comprised 85% of the captures.

PAGE 110

101 Table. 4-1. Number of carnivorous and noncarnivorous morphospecies of invertebrates captured (the question mark denotes uncertainty about the diet). Taxonomic number of non-carnivorous number of carnivorous Group morphospecies morphospecies Molusca 4 Crustacea 7 Ephemeroptera 1 Odonata 2 Orthoptera 1 Plecoptera 1 Hemiptera 7? Neuroptera 2 Coleoptera 13 Mecoptera 1 Diptera 3 Megaloptera 2 Table 4-2. Diet classification of fish species captured in E. oleracea forest. Species diet Characiformes Characoidea Characidae Tyttobrynchon sp. o/d Hyphessobrynchon o/d Erythrinidae Hoplias malabaricus c Erythrinus erythrinus c Lebiasinidae Pyrrhulina brevis or P. laeta h Copella sp. o/d Rivulidae Rivulus 0 Siluriformes Callichthyidae Hoplosternum = Callichthys thoracatum 0 Gymnotiformes Gymnotoidei Electrophoridae Electrophorus electricus o Perciformes Cichlidae Acaronia sp. o/d Aequidens sp. o/d * c-camivore, h=herbivore, o=omnivore, o/d=omnivore/detritivore

PAGE 111

102 control low-removal high-removal Treatment Figure 4-3. Herbivorous, omnivorous, and detritivorous invertebrate responses to high fruit removal, low fruit removal, and no removal (control). Bars represent averages ± s.e. of 16 samples (four sampling periods in each of four replicates). Treatments did not differ significantly (all p > 0.10). a) all species; b) individuals minus shrimp; and c) shrimp individuals.

PAGE 112

103 Non-carnivorous fish responded to fiuit harvest. The number of non-carnivorous fish species was 76 and 57% lower in the low-removal and high-removal treatments, respectively, than in the control (Fig. 4-4a; y; = 177.50, df = 2, p < 0.001). The number of species did not differ between the lowand high-removal treatments {y; = 5 .78, df = 2, p > 0.05). Of those species, minus Pyrrhulina, the number of individuals did not differ among treatments (Fig. 4-4b; = 0.82, df = 2, p > 0.50). When only Pyrrhulina is considered, however, there were 61% fewer (%^ = 189.00, df = 2, p < 0.001), and 42% fewer (%' = 203.00, df = 2, p < 0.001) individuals in the lowand highremoval treatments, respectively, than in the controls (Fig. 4-4c). Discussion Tropical blackwater rivers and the forests they inundate in the Amazon basin are notoriously low in nutrients (Sioli 1968, Janzen 1974, Furch and Junk 1997a), yet support one of the most diverse and productive fish communities in the world (Goulding 1980, Waldhoff et al. 1996, Junk et al. 1997). Goulding (1980) and others (Gottsberger 1978, Kubitzki and Ziburski 1994) revealed the central role of fruit in maintaining high fish diversity and biomass in Amazonian floodplains. In this study, fruit harvest significantly reduced the abundance of Pyrrhulina, an omnivorous fish, providing further support that fruit is a critical resource for fish in this system. In contrast, invertebrates apparently did not respond to fruit harvest. This study also shows that decomposition of fallen fruit recycles substantial quantities of nutrients, which are removed from the system when humans harvest fruits. Harvest of floodplain fruits by people, a common practice in Brazilian Amazonia (Muniz-Miret et al. 1995), may therefore affect several levels of the flooded forest ecosystem.

PAGE 113

104 a Figure 4-4. Herbivorous, omnivorous, and detritivorous fish responses to high fruit removal, low fruit removal, and no removal (control). Bars represent averages ± s.e. of 1 6 samples (four sampling periods in each of four replicates). Letters over bars indicate statistical differences. a) all species: controls had significantly more species than the high-removal treatments; b) individuals minus PyrrhuUm. no differences among treatments; c) Pyrrhulina individuals: the number of Pyrrhulina sp. was significantly lower in both the low-removal and high-removal treatments than the controls.

PAGE 114

105 Fruit production, nutrient content, and nutrient loss The amount of nutrients that fruits recycle through forests depends, in part, on the amount of fruits produced, which varies widely among forests and years. My estimates of 2,456 5,236 kg ha"' yr' of just E. oleracea fruit fall within the upper range of estimates from more species-rich neotropical forests. Terborgh (1986b), for example, estimated annual fruit production to range between 91 .25 5,840 kg ha ' yr"' at Cocha Cashu, Peru. On Barro Colorado Island, Panama, fruit fall ranges from approximately 930 kg ha ' yr ' in a "normal" year to 646 kg ha"' yr ' in a "famine" year (Foster 1982a). Franken et al. (1979) found that 469 kg ha"' yr' of fruits and flowers fell in a central Amazonian floodplain forest. In a comparison of several forest types in Puerto Rico, Lugo and Frangi (1993) found that total fruit fall was lowest in mature forest (376 kg ha"' yr '), highest in secondary forests (2,946 kg ha"' yr'), and intermediate in forests dominated by the palm Euterpe globosa (1,133 kg ha"' yr'). Because fruit fall does not measure the amount of fruits consumed and carried off site, fruit production in these forests must actually be much higher. Bannister's (1970) estimate of fruit production in the same^. globosa forest, for example, was 14 times (-15,000 kg ha ' yr') the amount calculated from fruit trap data by Lugo and Frangi (1993). In Caxiuana, many E. oleracea fruits and seeds are not eaten directly from the infructescence, but instead fall into the water to be eaten by aquatic frugivores, decompose, or germinate. If all fruit produced annually in my sites fell in the water and decomposed, then 1 1.22 23.93 kg N ha ', and 0.76 1.62 kg P ha ', would leach from them. The percentage of fruits eaten by frugivores, and the percentage that fall and decompose are, unfortunately, not known. If vertebrates remove 50% of the fruit, and

PAGE 115

106 50% of the remaining fruit is left to decompose, then N and P leaching would total 1 .61 3.44 and 0.1 1 0.23 kg ha' yr', respectively. Is recycling through fruits important in the nutrient budget of floodplain forests'^ Franken (1979) found that fruits and flowers together returned 14% of the total N, and 25% of the total P, respectively in a central Amazonian floodplain forest (10.3 kg ha ' yr' N and 0.35kg ha"' yr"' P). These high percentages of nutrients contributed by fruits were disproportional to their abundance in total litter fall, which was only 7% (Franken et al. 1979). That fruits comprise such a small fraction of litter fall leads some researchers to overlook their role in nutrient cycling (Furch and Junk 1997), which clearly disregards an important component of the nutrient budget. Harvest of fruits by people likely removes substantial amounts of nutrients from the system that would otherwise be recycled. Picking 75% of the ripe fruit, as is done regulariy in many forests used for commercial E. oleracea harvest (Muniz-Miret et al. 1995) and as was mimicked by my high-removal treatment, would remove 8.4 17.9 kg N and 0.57 1 .2 kg P ha ' yr'. In one of the few other studies to examine the loss of nutrients through harvest of non-timber resources, Witkowski and Lamont (1996) found that picking of Banskia blooms in Australia reduced N and P contents of plants by 30% over a period of 9 years. Moreover, the soil surrounding picked populations contained 3,103 and 152 g ha ' less N and P, respectively, than did the soil in unpicked populations. Thus, not only plants but also the surrounding system can be affected by removal of nutrient-rich plant reproductive structures. The removal of nutrients via fruit extraction may affect floodplain forests in several ways. First, the removal of such an important fraction of the P cycle could

PAGE 116

107 significantly reduce productivity of the forest and the floodplain. Phosphorous is often the most limiting nutrient in tropical forests (Vitousek 1984); this is especially true in nutrient-poor blackwater systems (Furch 1997). In addition, P has been found to be more limiting than N to phytoplankton production in Amazonian floodplains (Setaro and Melack 1984). Not only the amount of P, but also the timing of its release could affect floodplain forests. Because fruit production is seasonal and fairly synchronized among individuals, decomposing, fallen fruits may supply a "pulse" of nutrients to the otherwise nutrient-poor water (Furch 1997, Furch and Junk 1997a, Lodge et al. 1994). Nutrient pulses can be critical in tropical forests, where competition between microbes and plants for nutrients is often strong (Lodge et al. 1994). Pulses of nutrients, especially those that are most limiting, can temporarily satiate microbial communities, allowing increased nutrient uptake by plants (Lodge et al. 1994). Aquatic animal responses to fruit harvest by people Fruit removal significantly reduced fish species richness and the abundance of the most numerous fish (Pyrrhulina sp.) in this study. This result is not surprising given the important role that fruit plays in supporting Amazonian fish communities. Goulding (1980) and others (Gottsberger 1978, Araujo-Lima et al. 1986, Kubiztki and Ziburski 1994, Waldhoff et al. 1996) have shown that over 200 species offish include fruits in their diets, either as major (> 89% of total food volume consumed) or minor components. Junk et al. (1997) report that fruit was found in the diets of 13 22% of the fish species in various central Amazonian communities. In consuming fruit, fish can function either as seed predators or seed dispersers in floodplain forests (Gottsberger 1978, Goulding 1980, Kubitzki and Ziburski 1994).

PAGE 117

108 The reduction of fruit in the high-removal and low-removal treatments in this study likely affected Pyrrhulina, considered a generalist herbivore (Knbppel 1970, Henderson and Walker 1986), in several ways. I saw this species nibbling E. oleracea exocarp and mesocarp (=pulp) on several occasions; while the importance of pulp in its diet is not knovm, reduction of pulp via harvest could be partly responsible for its lowered abundance in high-removal treatments. In other studies, Pyrrhulina was found to consume mostly algae and fungus growing on submerged litter (Knoppel 1970, Henderson and Walker 1986). Algae and fungus likely colonize decomposing fruits as well, so harvest may have removed an important substrate for algal and fungal growth. Furthermore, the production of algae in floodplains is P-limited, so the removal of P-rich fruits through harvest may have reduced the overall growth of algae, to which Pyrrulina may have responded. That other fish species, and the invertebrates, did not respond to fruit harvest may indicate their greater reliance on other food items. Shrimp dominated the invertebrate community in this study, as is typical in blackwater floodplains (Irmler 1975, Henderson and Walker 1986, Junk and Robertson 1997). Henderson and Walker (1986) found that shrimp in the families Palaemonidae and Euryrhynchidae eat high proportions of not only algae and fungi, but also leaf litter, sponges, and arthropods. Thus, the reduction of fruit may not have severely reduced their food base. Similarly, the other omnivorous fish species captured in this study apparently rely on other food resources to a greater degree than does Pyrrhulina. Knoppel (1970), for example, found that Copella, Hyphessonbrynchon, Acaronia, and Aequidens all consumed detritus and plant remains; Copella was the only one in this group listed as eating algae, but also consumed many

PAGE 118

109 arthropods. Due to my sampling methods, most of the fish that were captured were too small (<12 cm) to ingest E. oleracea fruits whole. If they consume E. oleracea pulp at all, it is likely in the manner of Pyrrhulina, by nibbling off small pieces opportunistically It is possible that larger fish that inhabit these forests may be more sensitive to fruit harvest. Alternatively, the apparent lack of response of other fish and invertebrates to fruit harvest may reflect the overriding importance of dissolved oxygen availability, hydrogen sulfide, or other water chemistry parameters, in determining animal distribution patterns in floodplains (Henderson and Crampton 1997, Junk and Robertson 1997). Pronounced oxygen deficiency is common in floodplain waters, due to lack of mixing with river currents, high rates of decomposition, and low rates of photosynthesis (Furch and Junk 1997a, Junk et al. 1997). Although dissolved oxygen was not measured in this study, it may have indeed been low if fruit decomposition rates were high. Some fish species have apparent adaptations to low oxygen levels (Junk et al 1997). Catfish (e.g., Hoplostemum thoracatum), for example, posses a modified intestine that stores air for breathing. Electric eels {Electrophorus electricus), on the other hand, persist in low oxygen environments by breathing air through a buccal cavity. Similarly, some invertebrates can tolerate low-oxygen conditions (Junk and Robertson 1997). Nevertheless, when oxygen levels become too low, fish typically seek more oxygenated areas in open waters or in places with aquatic macrophytes. In addition, when water levels decrease, as occurred during low tides during this study, fish leave the floodplains for the rivers. In central Amazonia, severely hypoxic lakes tend to have no benthic organisms at all during most of the year (Junk and Robertson 1997). Thus the oxygen

PAGE 119

110 levels may have been more important than fruit abundance in determining the observed patterns of some omnivorous aquatic organisms. In conclusion, this study shows that harvest of fruit from Amazonian floodplains by humans can affect aquatic organisms and nutrient cycling. Others (Goulding 1980, Waldoff et al. 1994) have pointed out the critical role of floodplain fruits in maintaining Amazonian fisheries, the primary protein source for millions of people. That fruit produced in floodplains forms the basis for Amazonian fisheries has even been used as an argument to preserve floodplain forests (Henderson and Walker 1986, Henderson and Crampton 1997). These forests' productivity is based to a large extent on tight cycling of nutrients between the abiotic and biotic components (Furch 1997). My results indicate that nutrient cycling might be disrupted by removal of nutrient-rich fruits by humans and therefore that harvest of fruits by humans can have both communityand ecosystem-level impacts. Caution should therefore be used in implementing fruit harvest as a sustainable forest use strategy.

PAGE 120

CHAPTERS SPATIAL AND TEMPORAL VARIATION IN HYDROCHORY IN AMAZONIAN FLOODPLAIN FOREST Introduction Hydrochory is a common means of seed dispersal in flooded habitats, but its effectiveness at dispersing seeds remains poorly understood. Like other methods of seed dispersal, hydrochory may be advantageous to plant fitness in several ways. Hydrochory may release seeds from the size constraints associated with endozoochory, myrmecochory, or anemochory (Williamson et al. 1999, Williamson and Costa 2000). Larger seeds produce more resistant, vigorous seedlings (Fenner 1985). In addition, hydrochory may convey some of the same advantages as do other means of dispersal (i.e., colonization, escape, directed dispersal; Howe and Smallwood, 1982, Thompson 1981, Dirzo and Dominguez 1986, Bennett and Krebs 1987, Schuup 1993, Wenny 1999). Hydrochory aids colonization of remote oceanic islands (Guppy 1917, Ridley 1930, Murray 1986) and downstream habitats along river corridors (Roth 1987). Water can also help seeds escape from under parent plants (Schneider and Sharitz 1988, Jones et al. 1994), where mortality is often high (Janzen 1970, Hubbell 1980). In addition, water may direct seeds to favorable establishment sites, such as the upper reaches of flooded shorelines, emergent substrates, or tank bromeliads (Schneider and Sharitz 1988, Kubitzki and Ziburski 1994, Scarano et al. 1997). Despite being a common means of dispersal in flooded habitats, however, few studies have quantified movement of seeds by water. Ill

PAGE 121

112 Plants, fruits, and seeds of some species have phenological and morphological mechanisms that promote both primary and secondary dispersal by water (Prance 1978, Goulding 1980). In both North American and Amazonian floodplains, for example, trees ripen and release seeds during maximum flood levels (Schneider and Sharitz 1988, Kubitzki and Ziburski 1994). Many of these diaspores float via mechanisms including hairs (e.g., Pseudobombax munguba), internal air pockets (e.g., Hevea spruceana, Swartzia polyphylla), and fibrous arils (e.g., Laetia suaveolens, Murray 1986, Roth 1987, Williamson et al. 1999). Some structures that impart buoyancy (e.g., plumes in Ehotheca pentaphylla. Fisher 1997) may also serve other functions, such as wind dispersal or adhesion to substrates (Dirzo and Dominguez 1986, Redbo-Torstensson and Telenius 1995). Moreover, some species possess multiple mechanisms for dispersal; for example fleshy pulp and buoyancy, or explosive capsules and buoyancy (Bulow-Olsen 1986, Kaufmann et al. 1991, Kubitzki and Ziburski 1994, Waldhoff et al. 1996). A number of factors influence the effectiveness of water dispersal (Schuup 1993). How long seeds float and how far they travel depends upon flood duration, water depth, and seed buoyancy (Murray 1986, Kubitzki and Ziburski 1994, Williamson et al. 1999). Buoyancy varies both within and among species; some seeds are not buoyant at all, but are nonetheless moved by running riveror rainwater. Water depth also determines which seeds are dislodged from higher microsites and re-dispersed (Huenneke and Sharitz 1986, Schneider and Sharitz 1988). Water velocity and direction likewise determine how many seeds move, how far they move, and their trajectory (Schneider and Sharitz 1988). In addition, leaf litter and emergent substrates can obstruct movement of

PAGE 122

113 seeds and provide suitable establishment sites (Huenneke and Sharitz 1986, Schneider and Sharitz 1988, Scarano et al. 1997). In Amazonia, approximately 230,000 km^ of forest flood seasonally (Junk 1997). Seeds in these forests must overcome 3-11 months of inundation up to 13 m, accompanied by anoxic conditions on the soil surface (Goulding 1980, Kubitzki and Ziburski 1994). According to Kubitzki and Ziburski (1994), "most diaspores" in these forests have devices that make them buoyant, often for 30 days or more. In addition to dispersal by water, various species in seasonally-flooded forest are dispersed by fish (Goulding 1980, Waldhoff et al. 1996). While prolonged inundation and hypoxic conditions characterize the seasonallyflooded forest, the situation in the remaining -20,000 km^ of tidaliy-flooded forest differs considerably. Tidal forest floods when rivers reverse their flow as freshwater is pushed back upstream by the rising tide of the Atlantic Ocean. Due to the immensity of the Amazon River's mouth, this tidal effect is evident as far as 1,000 km upstream (Goulding 1980), and forests flood as far as 200 300 km upstream (Sioli 1966). These twice-daily tidal inundations pose different challenges than do seasonal floods for dispersal and regeneration. The magnitude of the tidal effect depends on the phase of the moon, but averages between 2.40 2.85 m, with a maximum of ca. 3.5 m (Sioli 1966). Tidal floods persist for only 3-4 hours, in contrast v^th much of the year for seasonal floods. In addition, water flow is bi-directional, sweeping in rapidly as the rivers overflow their banks, and flowing out just as swiftly several hours later. Thus, seeds not only flow out of the forest and downstream, but also flow in from dovmstream to colonize the upstream soils. During these short floods, soil conditions likely do not become anoxic, and

PAGE 123

114 submerged seeds are re-exposed merely hours later. These tidal inundations have likely been a unique selective force on the plants that live and reproduce in this system, but have received little attention as vectors of seed dispersal. I studied water dispersal of seeds in the tidally flooded Whitewater forests of the eastern Amazon basin. The specific objectives were to evaluate the spatial and temporal changes in water depths during high tide events and the effects of water depth on dispersal of seeds of different species. In addition, I tested whether microtopography hindered the dispersal of seeds, also during high tide events. Methods This research took place in the southeastern portion of the Amazon River estuary in Para State, Brazil. Study sites were located on floodplain forest, locally known as "varzea", along a Whitewater tributary to the Tocantins River. The extremely low-lying forest in this region is riddled with streams, many of which have water only during high tides (Sioli 1966). Data were collected along ten such streams. Water depth was measured along one transect per stream. Transects, which all began in the center of streams and extended 20 meters perpendicular to the streamside, were >100 m apart. Streams were < 5 m in width. Beginning at stream centers, sticks were placed in the ground at 2-meter intervals along transects, and the depth of the water was measured from the height of the mud mark left on the stick by the flood water. As the water was extremely sediment-laden, this method, though simple, was reliable. After each reading the sticks were cleaned. Water depth was measured during each of the moon phases of a one-month period in 1997 (new = Jan. 10, first quarte r= Jan 17, full = Jan 24, last quarter = Feb 1). Because streams are extremely abundant in the area, and

PAGE 124

115 because their courses are very winding, the 20 m transects sometimes became inundated by water from other streams as well as water from the focal stream. In the water dispersal experiments I used six species of locally available fruits and seeds, depending on which was the typical dispersal unit for each of the species. For Mauritia flexuosa (Palmae) and Euphorbiaceae sp. I used fruits, as those are what usually fell from trees in primary dispersal. Y ox Euterpe oleracea (Palmae), Virola surmamensis (Myristicaceae), Carapa guianensis (Meliaceae), and Hevea brasiliensis (Euphorbiaceae) I used seeds, as these are what one usually encounters on the forest floor. These six species differ in fruit and seed morphology, in buoyancy, and in their diversity of dispersal mechanisms. The two palms produce drupes with fleshy pericarps, and both are eaten by birds and mammals (see Chapter 3, also Roosmalen 1985, Bodmer 1989, 1990). The fruits, but not seeds, of M. flexuosa, float, but neither the fruits nor seeds of oleracea float. V. surinamensis seeds are buoyant and are partly covered by an oily red aril that attracts primates, toucans, and other birds (Howe and Vande Kerckhove 1980). H. brasiliensis seeds are produced in triplets in dehiscent capsules that open explosively, dispersing seeds over a 10 area (Goulding 1980), Hevea seeds are reported as the favorite food of several commercially-important characin fishes (Goulding 1980) and can float for several months (Huber 1910). The seeds of Euphorbiaceae sp., a subcanopy tree, are encased in buoyant blue dehiscent capsules that fall singly or in groups of 2 6. Finally, C. guianensis produces 8-14 seeds within four-valved, dehiscent globose capsules that break open upon falling, liberating the seeds. All of these species (except possibly Euphorbiacae sp.) also occur in the seasonally-flooded forests of the middle

PAGE 125

116 Amazon (Goulding 1980, Gottsberger 1978, Kubitzki and Ziburski 1994, Waldoff et al. 1996). I tested the movement of these fruits and seeds along the same transects used to measure water depth. At four-meter intervals along the transects I established seed deposition stations marked with sticks. At each station, one diaspore of each species was placed on the ground during low tide and checked after two high-tide events. To check seeds and fruits, I examined the area within a 50 cm radius of the seed deposition station. Diaspores were recorded as either not dispersed (i.e., still at station), dispersed < 50 cm, or dispersed >50 cm. Most of the fruits and seeds in this latter category were not relocated. To evaluate whether different water levels associated with different phases of the moon affected dispersal, I repeated the experiment during each moon phase. Any movement of diaspores from the original location was considered dispersal. In the leaf litter experiments I used 1 cm spherical pseudoseeds made of brightlycolored wax, because their buoyancy was assured and their color facilitated relocation. To test if microtopography affected dispersal, I placed 56 pseudoseeds in obstructions, such as palm leaves or stilt roots, and 56 in unobstructed positions on the forest floor. Seeds were placed at low tide, separated by >1 m, and re-located after 1 high tide flood, when I recorded which of them moved and the distance. Results Water depth in the forest varied over space and time (Fig. 5-1). Depth decreased as a function of distance to the nearest stream (r^0.80, F,o.298 = 30.34, p < 0.001), but the rate of decrease depended on the flood intensity, which varied with moon phase (F30.298 = 3 .05, p < 0.001). During the new and full moons the entire forest flooded, whereas during the first quarter and last quarter moons, forest beyond 20 and 10 m from streams,

PAGE 126

117 Distance from stream bed (m) Figure 5-1. Water depth as a function of distance from stream beds and moon phase. Depths were first standardized, so that the depth at each distance is relative to the depth at distance 0, and then averaged for ten different transects. See text for actual depths.

PAGE 127

118 respectively, remained unflooded. Overall, water levels were greatest during the new moon (46-92 cm), followed by the full (15-80 cm), first quarter (2-69 cm), and last quarter (0-48 cm) moons. Whether or not seeds and fruits were dispersed by water during floods depended on water depth and diaspore type (Fig 5-2 ). I used logistic regression (Trexler and Travis 1993) to test whether the continuous variable, depth, was a significant predictor of seed dispersal. In addition, I used inverse predictions (SAS JMP 1996) to determine the water depth at which dispersal of seeds was predicted with a given probability. For all buoyant seeds together and for each species individually, water depth predicted dispersal (Table 5-1). The effect of water depth on dispersal varied, however, among species. For example, for the smallest buoyant seeds (Euphorbiaceae sp.) 50% dispersal was predicted with only 10.7 cm water. This depth occurs throughout nearly the entire forest during the new and full moons, and extends to 4 m during the first quarter moon. The larger seeds of V. surinamemis, however, required 16.0 cm for 50% dispersal, and those of M.Jlexuosa require 25.7 cm. These depths again are common during the new and full moons but are rare otherwise, suggesting that dispersal by water takes place sporadically throughout the month for these species. Water depth also predicted dispersal of non-buoyant £. oleracea seeds. The logistic regression model predicted that 50% of E. oleracea seeds would disperse only with 100 cm water. The predicted depth for 25% dispersal was 75.2 cm. These depths only occur during the new and full moons. At these times the seeds are likely dragged along the forest floor during the rising and falling tides.

PAGE 128

{HSJddsip JO X;i|iq6qojd

PAGE 129

120 to e o (N IT) \0 CN On VO >0 m m >/~i r~m CS ^ (N — r~ 00 o ^ o rlO rn 'O iri O O (S — O — ' O T}ON (N 0\ i/S 'O oci m m fi e JS ' O *G a !Z5 * * * ON rr-00 CO CO o (N On vO m On O ^ m — ON O On in IT) cs (s) IT) o o o o o o (N On — « O m 00 fS O (N NO ON o u-i m o m Tf ' CSi (N — o X X X X x: X o i/-^ rcN o t-^ iri f«-i ,— i ^ o M M CO cn <4-l 5u C ?5 >< .S -S g ^ Q 1^ ?; b P «u 2 U kj w NO <3On m 00 00 NO ON in 00 m •"3, V o > 3 = > x: cn -C T3 u. CO (4-1 « o 12 3 _ o u ^ o cn c (U o c/} u ^ CO o o ts d ^ cd c S3 O 3 o ^ V D. Q.
PAGE 130

121 The influence of water depth on buoyant-seed dispersal also suggests temporal variation in the quantity of seeds dispersed. Indeed, the overall proportion of seeds that dispersed > 50 cm varied according the four phases of the moon (Fj.us = 5.89, p < 0,005, Fig. 5-3a), with almost all seeds moving during the new moon and almost none during the last quarter moon. The number of seeds moving during the new moon was 58% and 94% greater than the number moving during the first quarter (F,. 19 = 0,67, p < 0.021) or last quarter (Fi.,9 = 1.07, p < 0.001) moons, respectively. The number dispersing during the last quarter moon was also 91% less than that moving during the full moon (Fi, ,9 = 0.69, p < 0.014). Despite the differences in water depth, the number of seeds dispersing during the new and full moons did not differ. Among the seeds dispersed < 50 cm, there were no differences among moon phases in the proportion of seeds dispersing, which was always low (Fig. 5-3b; 7.5% during the full moon, 4-5% during the other phases). Forest floor litter did not affect which seeds moved; 71% of obstructed seeds and 89% of unobstructed seeds dispersed (Fig. 5-4a). Litter did, however, reduce the distance moved by seeds. Unobstructed seeds moved 45% farther than did obstructed seeds (Kruskal-Wallis Rank Sums x'=7.32, df=l, p<0.007, Fig. 5-4b). Discussion Water dispersal of both buoyant and non-buoyant fruits and seeds increased with water depth, which varied spatially and temporally. Diaspores on the ground during the full and new moons had a much greater probability of dispersing than diaspores in the same sites during the first quarter and last quarter moons. Also, diaspores nearer to streams had a much greater probability of dispersing than diaspores farther from streams I

PAGE 131

122 1.2-1 Moon phase Figure 5-3. Proportion of five species of buoyant seeds that were dispersed (mean + s.e.). a) Greater than 50cm; b) Less than 50cm.

PAGE 132

123 60-, 5040302010035302520151050obstructed unobstructed Condition Figure 5-4. Dispersal of 56 wax pseudoseeds from obstructed and unobstructed sites. a) The number moved did not differ between conditions; b) Distance moved (mean + s.e.). Unobstructed seeds moved farther than obstructed seeds.

PAGE 133

124 on the same days. The effect of water depth differed greatly, however, among the species that I studied. A 50% probability of dispersal was predicted for C. guianensis, H. brasiliensis, Euphorbiaceae sp., V. surinamensis, M. flexuosa, and E. oleracea with 3 .76, 5.64, 10.73, 15.97, 25.68, and 100.04 cm water, respectively. In addition to water depth, litter also affected dispersal of seeds by significantly reducing the distance that they moved. Such variation among species and over space and time shows that hydrochory is a more heterogeneous system of seed dispersal than previously appreciated. The temporal variation in dispersal found in this study may be typical of dispersal systems with abiotic vectors. Schneider and Sharitz (1988), for example, found temporal "pulses" of dispersal caused by elevated water during short-term floods in a temperate floodplain. Similarly, wind velocity can vary temporally, influencing its effectiveness in dispersing seeds (Greene and Johnson 1997). Such temporal variation in dispersal vectors implies that seeds falling from the same tree experience different probabilities of dispersal. Many of the floodplain forest tree species used in this study produce fruits over several months during the mid and late rainy seasons, which are the periods of highest flood levels in both seasonallyand tidally-flooded forests (Sioli 1966, Goulding 1980, Kubitzki and Ziburski 1994). Seeds falling from these trees would thus experience all four moon phases; some seeds could fall just before or during the new or full moons and have a good chance of dispersal by water. Other seeds could fall just after one of these moon phases and not experience inundation for days or weeks. In the period of this study, for example, 14 days passed between the new and full moons, during which very little flooding occurred. This length of time may be critical to plant fitness. Many seeds

PAGE 134

125 may germinate (Kubitzki and Ziburski 1994) or suffer predation in that length of time (Janzen 1970, Hubbell 1980, Wenny 1999) and therefore not be dispersed by water. The spatial variation in water depth and the influence of litter imply that seeds falling from different trees in the same population experience different probabilities of dispersal by water. The distribution of seeds on the ground is often heterogeneous (Janzen 1970, Wenny 1999), due to the distribution of fruiting trees, and primary dispersal by, for example, ballistic ejection (Biilow-Olsen 1986) or birds (Wenny 1999). In this study, seeds closer to streams had a greater probability of dispersal, because those areas flooded more frequently, more deeply, and probably with more quickly moving water (Schneider and Sharitz 1988, Fisher 1997). Spatial variation in dispersal by water has often been associated solely with the occurrence of elevated microsites, such as logs and buttresses, which trap seeds and prevent their dispersal (Huenneke and Sharitz 1986, Schneider and Sharitz 1988, Redbo-Torstensson and Telenius 1997). In this study, litter did not prevent dispersal but rather resulted in shorter dispersal distances, adding to the spatial heterogeneity of dispersal. In addition to experiencing spatial and temporal variation in flooding regimes within sites, plant species might also experience variation in flooding regimes across their ranges. Many tree species (including M^Zexwo^a, H. brasiliensis, V. surinamensis, C. guianensis, see Anderson et al. 1995) occur in both seasonallyand ti dallyflooded forests. Furthermore, since water can disperse seeds great distances (Guppy 1917, Ridley 1930), seeds produced on a tree in middle Amazonian seasonallyflooded forest might germinate downstream in tidallyflooded forest, where conditions for hydrochory differ greatly.

PAGE 135

126 From the point of view of fruiting plants, then, water is an unpredictable vector for seed dispersal in tidally flooded forests. Such unpredictability may explain why many buoyant fruits and seeds also have apparent mechanisms for dispersal by other means. The nutritious aril and pulp covering V. surinamensis and M flexuosa seeds, for example, attract frugivores such as toucans, primates, tapirs, and fish, which ingest the fruits and (sometimes) disperse the seeds intact (Goulding 1980, Howe and Vande Kerckhove 1980, Bodmer 1990). Hevea seeds are first dispersed explosively when the fruit capsule is exposed to the sun, dries, and splits open (Roosmalen 1985). Seeds of Hevea spp. are also favored by a number of Amazonian fish, which either destroy or disperse them (Huber 1919, Gottsberger 1978, Goulding 1980). Production of buoyant seeds covered in nutritious pulp or aril or enclosed in explosive capsules is thus likely an effective means of increasing the probability of dispersal via one or more mechanisms. This view contrasts with that of Kubitzki and Ziburski (1994), who suggested that many floodplain species possess other means of primary dispersal due to phylogenetic relationships. They demonstrated that in some genera, dispersal mechanisms such as arils, pulp, and membraneous wings are reduced in floodplain species, compared with non-floodplain congeners, implying that their presence is explained more by history than by dispersal agents. The argument of Kubitzi and Ziburski (1994) assumes that plants make use of only one dispersal agent, which is unlikely in light of the apparent importance of secondary dispersal to plant regeneration (Wenny 1999). I suggest that arils, pulp, wings, and other structures for primary dispersal have been maintained or added to buoyancy in species that can take advantage of multiple dispersal vectors.

PAGE 136

127 Floodplain species are not alone in producing seeds with multiple mechanisms for dispersal. Some plant families (e.g., Leguminosiae, Violaceae) contain many diplochorous species that produce elaiosome-bearing ant-dispersed seeds within dehiscent pods or capsules. These are first dispersed ballistically or gravitationally and secondarily dispersed when ants carry the appendage-bearing seed to their colony, consume the elaiosome, and discard the seed (Thompson 1981, Biilow-Olsen 1986, Bennett and Krebs 1987, Ohkawara and Higashi 1994, Espadaler and Gomez 1996). Another example of diplochory is seen is Ficus microcarpa (Moraceae) fruits, which are syconia containing multiple seeds, each of which is partially covered by a fleshy, elaiosome-like exocarp that is highly attractive to ants (Kaufmann et al. 1991). When birds ingest the syconia, they disperse the seeds with the exocarp intact; many of these seeds are then transported by ants and later germinate. While some plant species produce seeds with multiple dispersal mechanisms, others produce multiple types of seeds, each with different dispersal mechanisms, a phenomenon known as heterospory. Seeds of Swartzia polyphylla, for example, vary in the size of the internal air pocket that allows buoyancy (Williamson et al. 1999). Seeds with small pockets have relatively high specific gravity, so they sink. On the other hand, seeds with large pockets have relatively low specific gravity, so they float. Intra-tree and intra-specific variation in seed buoyancy has been found in other tree species in Costa Rica and Amazonia (Kubitzi and Ziburski 1994, Williamson and Costa 2000), implying that heterospory in buoyancy may be common. The presence of diplochory and heterospory in seeds implies that possessing multiple types of dispersal mechanisms increases plant fitness. In the case o^ Swartzia

PAGE 137

128 polyphylla and Pentaclethra macroloba, it was posited (Williamson et al. 1999, Williamson and Costa 2000) that the "floater-sinker" dimorphism exists to take advantage of the high-disturbance nature of the flooded habitat in which these tree species grow. Under conditions of intense erosion and uprooting, it may be advantageous for seeds to sink under parent trees where emerged substrates for establishment may be abundant. At the same time, it may be advantageous for other seeds to float and arrive at other suitable establishment sites. Like heterospory, diplochory may increase plant fitness in several ways. Increased fitaess could result if secondary dispersers place seeds in sites in which germination is favorable or predation is unlikely (Putz and Holbrook 1986, Ohkawara and Higashi 1994, Espalder and Gomez 1996). Increased fitness could also result if a combination of different dispersal mechanisms allows seeds to reach multiple types of suitable sites. In Viola, for example, diplochory may maintain two types of soil seed banks: a short-term seed bank in ant nests, and a longer-term seed bank in the soil outside of ant nests (Bulow-Olsen 1986). When abiotic dispersers act in only one direction, as does water in rivers and gravity on slopes, dispersal by vertebrates may prevent the longterm directional movement of plant populations. For instance, Mack (1995) demonstrated that dwarf cassowaries non-randomly dispersed seeds of Aglaia to level ground uphill from source trees, whereas gravity dispersed seeds downhill only. In the case of Amazonian floodplain species, vertebrates may disperse seeds locally and upstream, while water may disperse seeds distantly and downstream. Advantages of multiple dispersal are not limited, however, to species whose seeds possess mechanisms for multiple dispersal. In most multiple-stage dispersal systems, in

PAGE 138

129 fact, seeds possess mechanisms for only the primary vector. For example, the arillate seeds of Guarea glabra and G. kunthiana are dispersed first by birds, which regurgitate or defecate the seeds intact (Wenny 1999). In these species secondary dispersal by rodents significantly re-arranges the seed shadows, with potential implications for regeneration, but the seeds possess no special attractant for the rodent scatterhoarders. In many species that are secondarily dispersed by rodents or ants, the seed itself is apparently the attractant (e.g.. Forget and Milleron 1991, Levey and Byrne 1993), and the cost to plants of loosing some seeds to predation may be offset by the benefits of having some seeds arrive in favorable establishment sites (reviewed in Crawley 1992). In other cases of secondary dispersal, especially by wind and water (Redbo-Torstensson and Telenius 1995, Greene and Johnson 1997), the membraneous wings or plumes used in primary dispersal also carry the seeds after they have reached the ground, water, or snow. In other species, the spherical shape of seeds promotes secondary dispersal by wind or water. In these cases, secondary dispersal may be merely serendipitous (Dirzo and Dominguez 1986). In conclusion, hydrochory presents interesting opportunities to compare and contrast different mechanisms of seed dispersal. Ecologists are only beginning, however, to understand the interesting and, perhaps unique, features of water dispersal systems. To date, for example, we lack knowledge of dispersal shadows for any hydrochorous species. Detailed studies of seed movement by water are needed to understand the quantity, quality, and effectiveness (sensu Schuup 1993) of water as a dispersal agent and the consequences of dispersal by water. Furthermore, rigorous comparative studies of dispersal mechanisms among and within taxonomic groups are needed to clarify

PAGE 139

130 historical versus selective influences on dispersal structures, and the evolutionary history of buoyancy in seeds.

PAGE 140

CHAPTER 6 GENERAL CONCLUSIONS Ecological Considerations Early studies in the field of fi^givory and seed dispersal focussed on "tight coevolutionary" patterns and processes between fi\agivores and fruiting plants (McKey 1975, Snow and Snow 1980). Later workers stressed that "diffuse co-evolution" between suites of frugivore and plant species might better explain characteristics such as fruit color and frugivore behavior (Wheelwright 1988). Several recent papers (Herrera 1992, 1998) show that fruit-frugivore interactions can be highly unpredictable and asymmetrical. For example, patterns once attributed to fruit abundance (e.g., inter-annual frugivore abundance) are sometimes better explained by abiotic factors (e.g., weather). Despite this unraveling of the original tenets of the field, however, general agreement seems to persist that frugivores and fruits do influence each other, in ecological and evolutionary terms. The key question now seems to be not whether fruit-frugivore interactions matter, but how much and in what contexts do they matter? As in all fields, one key to revealing mechanisms responsible for observed patterns of fruits and frugivores is experimentation. In this project, I took advantage of the fruit-harvesting activities of Amazonian people to experimentally manipulate fruit at a scale large enough to affect frugivore activity. High-intensity fruit harvest substantially reduced fruit-eating bird species richness, abundance and visit durations, fruit-eating mammal species richness, and fruit-eating fish abundance. These results demonstrate 131

PAGE 141

132 that, despite changes in abiotic factors (e.g., amount of rainfall, water depth) occurring at the same time, fruit abundance governed the abundance and distribution of frugivores. Why do some studies demonstrate a tight link between fruit and frugivore abundance while others do not? McKey (1975) posited that the abundance of "specialist" frugivores those that derive all of their carbohydrates, lipids, and protein from fruits should be linked to fruit abundance to a greater degree than the abundance of "opportunistic" frugivores, who derive mostly carbohydrates and water from fruits. If this were the case, then I should have detected a higher degree of fruit sensitivity in specialist than opportunistic bird species, which I did not. Degree of frugivory appears to be an unreliable predictor of links between fruit and frugivore abundance in general; Rey (1995) found that some partial frugivores tracked olive fruit availability over space and time, whereas Herrera (1998) found that the interannual abundances of the same bird species (with the exception of one) were not correlated with fruit abundance. Instead, the abundances of most species were associated with autumn weather patterns. Whether or not a link exists between fruit and frugivore abundance may depend more on the type(s) of fruits involved. The two largest-scale studies showing links my study and that by Rey (1995) each focus on the abundance of a single species of fruit that occurs in high density stands. Both are consumed by numerous species of frugivores and, at least in the case of E. oleracea, fruit production takes place during a period of general fruit scarcity in the forest. This last characteristic is shared by species sometimes considered "keystone plant resources" those that maintain the frugivore community during periods of scarcity (Terborgh 1986, Peres 2000). It is possible that a few key fruit species, rather than the entire fruiting plant community, drive relationships between fruit

PAGE 142

133 and fnigivore abundance. Further studies of so-called keystone plant resources could help illuminate their importance for frugivore abundance and distribution. Another disputed area of fruit-frugivore interactions is whether seed dispersal by frugivores influences the abundance and distribution of plants. The role of frugivores in plant regeneration depends on the importance of their activities relative to other factors that affect seeds, including other agents of dispersal. My chapter on water dispersal of floodplain species is one of the first studies to quantify movement of seeds in relation to water depth. It illustrates the important role that secondary dispersal can have in rearranging seed shadows. As more studies address secondary dispersal, it is likely to come to light as a critical process influencing plant populations (Wenny 1999). Conservation Considerations The past decade has witnessed the popularization of non-timber forest products (NTFP's) as sources of income from tropical forests (Anderson 1988, Peters et al. 1989, Panayotou and Ashton 1992, Plotkin and Famolare 1992). Due to its potential income, NTFP harvest has gained a reputation as a means to merge the conservation of forest resources through sustainable use with economic and social human development. Enthusiasm to feature NTFP harvest in conservation and development projects, has, however, outpaced knowledge about the ecological sustainability of such plans. Empirical data on which to base guidelines for ecologically-sustainable harvest have been lacking. In this study, I examine two human activities involving the NTFP a^ai (i.e., forest enrichment and fruit harvest) to evaluate whether they impact biodiversity. I found that

PAGE 143

134 enrichment management to increase 3931 production significantly alters forest structure and composition, which likely leads to observed changes in bird community structure. Second, I found that high-intensity fruit harvest significantly affects the abundance and behavior of fruit-eating birds, mammals, and fish. Furthermore, fruits may be important in the cycling of nutrients through floodplain forests, so their removal via human harvest may disrupt nutrient cycling. Taken together, these results indicate substantial effects of a?ai management and harvest. This interpretation of my results is akin to viewing a glass of water as half-empty rather than half-full. While I did find the above effects, I also found that a^ai-enriched forests contains a high diversity of birds (albeit not many forest interior species), and that harvest of -40% afai fruit did not affect frugivorous bird activity (although frugivorous mammals were affected). In other words, enrichment and harvest do not affect all aspects of biodiversity at all levels. Furthermore, my results must be viewed in the context of alternative, economically viable forest management options, which are limited to various types of timber extraction, which have their own suite of ecological impacts (Frumhoff 1995, Putz et al. 2000). Some effects of timber extraction on forests (e.g., increased probability of fire) may not apply in forests managed for non-timber products. What is the future role of non-timber forest products in tropical forest conservation? Based on my results, and after consideration of alternative forest uses, it seems that non-timber forest product management and harvest still do have as much or more potential to successfully merge forest use with biodiversity conservation as do other forest management strategies. I therefore continue to offer cautious support to nontimber forest product development. To better understand the potential of non-timber

PAGE 144

135 forest products, research is needed on more species that supply forest products, on systems in which multiple species are harvested from one forest, and on the larger spatial and temporal effects of NTFP management and harvest. Until then, forests managed for NTFP production and harvest should continue to serve as complements to, not substitutes for, strictly protected forests.

PAGE 145

APPENDIX 1 DIET CLASSIFICATION OF SPECIES OBSERVED IN E. OLERACEA FOREST. NOMENCLATURE FOLLOWS SICK (1993). Species diet Ciconiformes Ardeidae Butorides striatus c Accipritridae Accipiter bicolor c Morphnus guianensis c Falconidae Daptrius amehcanus c Craciformes Cracidae Crax sp. g Columbiformes Columbidae Columba speciosa g Columbina talpacoti g Pssitaciformes Pssitacidae Ara ararauna g Ara chloroptera g Ara macao g Ara manilata g Ara sever a g Amazona amazonicus g Pionites leucocephala g Pyrrhura per lata g Amazona farinosa g Brotogeris versicolorus g Deropteus accipitrmus g Cuculiformes Cuculidae Paiya cayana i Trochiliformes Trochilidae Glaucis sp. n Phaethornis sp. n 136

PAGE 146

137 Species diet Trogoniformes Trogonidae Trogon collaris p Trogon melanocephalus p Trogon viridis p Galbuliformes Galbulidae Galbula albirostris i Bucconidae Bucco sp. i Coraciiformes Cerylidae Chloroceryle aenea c Piciformes Picidae Campophelis rubricollis i Celeus undatus p Celeus grammicus p Piculus Jlavigula p Ramphastidae Ramphastos tucanus p Ramphastos vitellinus p Pteroglossus bitorquatus p Passeriformes Dendrocolaptinae Lepidocolaptes albolineatus Dendrocincla fulginosa Xiphorhynchus guttatus Xiphorhynchus obsoletus Glyphorhynchus spirurus Fumariinae Automolus rufipileatus i Xenops minutus i Formicariidae Cymbilaimis lineatus Hypocnemis cantator Hypocnemoides melanopogon Myrmotherula axillaris Cercomacra cinerascens Sclateria naevia Thamnophilus punctatus Pipridae Pipra rubrocapilla

PAGE 147

138 Species Cotingidae Lipaugis vociferans Gymnoderus foetidua Atilla cinnamomeus Cotinga cotinga Cotinga cayana Querula purpurata Rhitiptema simplex Tyrannidae Todyrsotrum maculatum Tolmomyias sulphurescem Tyrannulus elatus Hemitriccus sp. Empidonax euleri Eleania flavogaster Troglodytidae Thryothorus coraya Microcerculus marginatus Muscicapidae Turdinae Turdus albicollis Sylvidae Ramphocaenus melanurus Vireonidae Vireoninae Vireo olivaceus Hylophilus sp. Emberizidae Coerbinae Coerba flaveola Parulinae Psarocolius viridis Icterus cayanensis Cacicus cela Thraupinae Thraupis episcopus Thraupis palmarum Ramphocelus carbo Eucometis penicillata Cyanerpes cyaneus Dacnis cayana diet P f P f f f P P P P f f f f n n diet: c-carnivore, g=granviore/frugivore, f=frugivore, i=insectivore, p=partial frugivore

PAGE 148

APPENDIX 2. RESULTS OF REPEATED-MEASURES ANOVA'S Tables AlA6. Results of repeated-measures ANOVA's with two among (site and treatment) and two within (harvest and time) effects. A significant response to the experimental treatment is indicated by the harvest x treatment interaction. Because multiple (three) comparisons were performed on the data sets for frugivorous birds, nonfrugivorous birds, and mammals, significance levels were adjusted using a Bonferroni sequential method (Rice 1989). The initial p-value was chosen as 0. 10, because the design had inherently low statistical power (only 4 replicates, Zolman 1993). The adjusted significance (adj. sign.) column indicates the level of significance after the Bonferroni method is applied. * significant after Bonferroni sequential adjustment at p<0.033 ** significant after Bonferroni sequential adjustment at p<0.05 *** significant after Bonferroni sequential adjustment at p<0.10 Table Al . frugivorous bird species. source of variation df MS F P adj. sign. site 3 .735 3.194 .1052 treatment 2 .662 2.874 .1332 Subject(group) 6 .230 harvest 1 1.121 17.773 .0056 *** harvest x site 3 .537 8.509 .0140 * harvest x treatment 2 .582 9.229 .0148 harvest x Subject(group) 6 .063 time 3 1.602 15.139 .0001 time X site 9 .220 2.078 .0891 time X treatment 6 .170 1.605 .2030 time X Subject(group) 18 .106 harvest x time 3 1.335 11.156 .0002 ** harvest x time x site 9 .577 4.825 .0022 harvest x time x treatment 6 .256 2.141 .0984 harvest x time x Subject(group) 18 .120 139

PAGE 149

140 Table A2 frugivorous bird inc ividua s. source of variation df M!S r P adj. sign. site 3 3.451 6,0 12 .U/ii treatment 2 3.253 6.420 .0323 * Subject(group) 6 .507 harvest 1 1 1.6/5 '5 O /111 3o,631 AAAO .UOUs harvest x site 3 1.505 4,894 .0472 ** harvest x treatment 2 3.027 9.846 .0127 harvest x Subject(group) 6 .307 time 3 3.979 9,073 .0007 time X site 9 .689 1.572 .1979 time X treatment 6 QQ9 Z.ZO 1 .UoH 1 time X Subj ect(group) 1 o ,439 harvest x time 3 7.146 19.122 .0001 * harvest x time x site 9 2.423 6.483 .0004 ** harvest x time x treatment 6 .806 2.156 .0966 harvest x time x Subject(group) 18 .374 Table A3. frugivorous bird visitations. source of variation df MS F P adj. sign. site 3 1.724 8.290 .0148 * treatment 2 .732 3.521 ,0974 Subject(group) 6 ,208 harvest 1 3.933 18,588 ,0050 ** harvest x site 3 .236 1,113 .4147 harvest x treatment 2 .856 4.045 .0772 harvest x Subject(group) 6 .212 time 3 .613 4,228 .0199 *** time X site 9 .194 1,341 .2841 time X treatment 6 .104 .716 .6418 timex Subject(group) 18 .145 harvest x time 3 1.737 19.480 ,0001 *** harvest x time x site 9 .119 1.333 ,2878 harvest x time x treatment 6 .261 2.928 ,0357 harvest x time x Subject(group) 18 .089

PAGE 150

141 Table A4. non-frugivorous bird species. source of variation df MS F n r adi sipn. 3 1.909 6.342 * treatment 2 .160 .532 .6128 6 301 harvp
PAGE 151

142 Table A6 non-frugivorous bird visitations. source of variation df MS F p adi. sign. site 3 1.909 6.342 .0273 * treatment 2 .160 .532 .6128 6 .301 harvest 1 3.075 9.286 .0226 harvest x site 3 .321 .970 .4662 harvest x treatment 2 .058 .176 .8427 harvest X Subiectr&roun^ 6 .331 time 3 1.653 9.379 .0006 *** time X site 9 .133 .756 .6567 time X treatment 6 .133 .755 .6141 timex Subject(group) 18 .176 harvest x time 3 .401 2.062 .1411 * harvest x time x site 9 .185 .951 .5080 harvest x time x treatment 6 .157 .807 .5779 harvest x time x Subject(group) 18 .194 Table A7. mammal species. source of variation df MS F P adj. sign. site 3 .917 2.839 .1281 treatment 2 .448 1.387 .3198 Subject(group) 6 .323 harvest 1 1.500 18.783 .0049 * harvest x site 3 2.694 33.739 .0004 * harvest x treatment 2 .594 7.435 .0238 * harvest x Subject(group) 6 .080 time 3 .694 2.956 .0602 time X site 9 .426 1.813 .1352 time X treatment 6 .267 1.138 .3806 time X Subject(group) 18 .235 harvest x time 3 .528 2.027 .1462 harvest x time x site 9 .611 2.347 0590 harvest x time x treatment 6 .163 .627 .7070 harvest x time x Subject(group) 18 .260

PAGE 152

143 Table A8. mammal individuals. source of variation df Mb r P adj. sign. site J 2 /.iyy 2.6!)6 treatment 2 2.510 .243 .7914 Sub)ect(group) 6 10.316 harvest 1 29.260 6.546 .U43U harvest x site 3 104.094 23.294 .0010 harvest x treatment 2 16.635 3.723 .0889 harvest x Subject(group) 6 4.469 time 3 23.455 3.927 .0256 time X site y con 23.529 3.939 .0064 UlllC A UCaliilClll /; o J.J 1 ^ 8QQ . J 1 oo timex Subject(group) 18 5.973 harvest x time 3 40.427 6.869 .0028 * harvest x time x site 9 17.483 2.971 .0236 * harvest x time x treatment 6 5.219 .887 .5246 harvest x time x Subject(group) 18 5.885 Table A9. mammal visitations. source of variation MS F P adj. sign. site 3 1.388 1.297 3585 treatment 2 1.079 1.008 .4194 Subiect(group) 6 1.071 harvest 1 .278 .693 .4370 harvest x site 3 4.846 12.071 .0059 *** harvest x treatment 2 1.157 2.883 1326 harvest x Subject(group) 6 .401 time 3 .485 1.636 .2162 time X site 9 .770 2.597 .0406 ** time X treatment 6 .678 2.286 .0813 timex Subject(group) 18 .297 harvest x time 3 .895 1.864 .1718 harvest x time x site 9 .614 1.281 .3120 harvest x time x treatment 6 .192 .400 8692 harvest x time x Subject(group) 18 .480

PAGE 153

LIST OF REFERENCES Abacus Concepts Inc. 1993. Super ANOVA. Berkeley, CA, USA. Abramson, J., B.L. Speer, and J.B. Thomson. 1995. The large macaws: their care, breeding, and conservation. Raintree Publications, Fort Bragg, CA, USA. Adis, J., K. Furch, and U. Irmler. 1979. Litter production of a central Amazonian black water inundation forest. Tropical Ecology 20:236-245. Adler, G. H. 1998. Impacts of resource abundance on populations of a tropical forest rodent. Ecology 79:242-254. Allegretti, M.H. 1990. Extractive reserves: an alternative for reconciling development and environmental conservation in Amazonia, pp. 252-264. In: A.B. Anderson, ed. Alternatives to deforestation. Columbia University Press. New York, NY, USA. Allen, CM. 1997. Frugivores, palms, and conservation in Amazonia. MS Thesis, University of Florida, Gainesville, FL, USA. Alves, A.I. 1995. Reserva extractivista do Rio Cajari. pp. 85-95. In: J. Ruiz Murrieta and R, Pinzon Rueda, eds. Reservas extrativistas. lUCN, Glanz, Switzerland. Alves. M.C. 1990. The role of cacao plantations in the conservation of the Atlantic forests of southern Bahia, Brazil. M.S. Thesis. University of Florida, Gainesville FL USA. Anderson, A.B. 1988. Use and management of native forests dominated by A9ai palm (Euterpe oleracea) in ihe Amazon estuary, pp. 144-154, In: The palm tree of life: biology, utilization, and conservation. Advances in Economic Botany 6. Anderson, A.B. 1990a. Deforestation in Amazonia: dynamics, causes, and alternatives, pp. 3-23. In: A.B. Anderson, ed. Alternatives to deforestation. Columbia University Press New York, NY, USA. Anderson, A.B. 1990b. Extraction and forest management by rural inhabitants in the Amazon estuary, pp 65-85. In: A.B. Anderson, ed. Alternatives to deforestation. Columbia University Press New York, USA. 144

PAGE 154

145 Anderson A.B. and M A G. Jardim. 1989. Costs and benefits of floodplain forest management by rural inhabitants in the Amazon estuary: a case study of a^ai palm production, pp.1 14-129, In: J O. Browder (ed.). Fragile lands of Latin America: strategies for sustainable development. Westview Press, Boulder, CO, USA. Anderson, A.B. and E.M. Ions. 1992. Valuing the rainforest: economic strategies by small-scale forest extractivists in the Amazon estuary. Human Ecology 20:337-369. Anderson, A.B., P. Magee, A. Gely, and M.A.G. Jardim. 1995. Forest management patterns in the floodplain of the Amazon estuary. Conservation Biology 9:47-61 . Araujo-Lima, C.A.R.M., B.R. Forsberg, R. Victoria, and L. Martinelli. 1986. Energy sources for detritivorous fishes in the Amazon. Science 234: 1256-1258. Arnold, J.E.M. and M R. Perez. 1998. The role of non-timber forest products in conservation and development, pp. 17-42, In: E. Wollenberg and A. Ingles, eds. Incomes from the forest: methods for the development and conservation of forest products for local communities. CIFOR, Bogor, Indonesia. Azevedo-Ramos, C, O. de Carvalho, Jr., K. Kalif, P.R.S. Moutinho, and A. Aleixo. 1999. The effect of low and high-impact logging on fauna in eastern Amazonia. Society for Conservation Biology Annual Meeting, College Park, MD, USA. Balee, W. 1988. Indigenous adaptation to Amazonian palm forests. Principes 32:47-54. Bannister, B. 1970. Ecological life cycle of Euterpe globosa Gaert. pp. 299-3 14, In: H.T. Odum and R.F. Pigeon, eds. A tropical rain forest. National Technical Information Service, Springfield, VA USA. Bates, H.W. 1864. A naturalist on the River Amazons. Seventh edition, John Murray, London, UK. Bawa, K.S. 1992. The riches of tropical forests: non-timber products. Trends in Ecology and Evolution 7:361-363. Bawa, K.S. and R. Seidler. 1998. Natural forest management and conservation of biodiversity in tropical forests. Conservation Biology 12:46-55. Bennett, A. and J. Krebs. 1987. Seed dispersal by ants. Trends in Ecology and Evolution 2:291-292. Bjork, R. 2000. Intratropical migration of a lowland parrot: implications for conservation. Paper presented at 1 1th Annual Meeting of the Society for Conservation Biology, Missoula, Montana, USA.

PAGE 155

146 Bodmer, R E. 1989. Frugivory in Amazonian Artiodactyla: evidence for the evolution of the ruminant stomach. Journal of the Zoological Society of London 219:457-467. Bodmer, R.E. 1990. Fruit patch size and frugivory in the lowland tapir (Tapiris terrestris). Journal of the Zoological Society of London 222:121-128. Bourne, G.R. 1974. The red-billed toucan in Guyana. The Living Bird 13:99-125. Browder, J.O. 1992. The limits of extractivism: tropical forest strategies beyond extractive reserves. Bioscience 42:174-182 Bronstein, J.L. and K. Hoffmann. 1987. Spatial and temporal variation in frugivory at a Neotropical fig, Ficus pertusa. Oikos 49:261-268. Bulow-Olsen, A. 1986. Diplochory in Viola, a possible relation between seed dispersal and soil seed bank. American Midland Naturalist 1 12:251-260. Calzavara, B.B.G. 1972. As possibilidades do a^aizeiro no estuario amazonico. Boletim da Fundafao de Ciencias Agrarias do Para 5:1-103. Chapman, C.A. in prep. Selective pressures on tropical trees: variance caused by animal foraging. Frugivores and Seed Dispersal. Chapman, C.A., L.J. Chapman, and L. Lefebre. 1989. Variability in parrot flock size: possible functions of communal roosts. Condor 91:842-847. Chapman, C.A. and L.M. Fedigan. 1990. Dietary differences between neighboring Cebus capucinus groups: local traditions, food availability or responses to food profitability? Folia Primatologia 54: 177-186. Chapman, C.A., R. Wrangham, and L.J. Chapman. 1994. Indices of habitat-wide fruit abundance in tropical forests. Biotropica 26:160-171. Chapman, C.A. and L.J. Chapman. 1995. Survival without dispersers: seedling recruitment under parents. Conservation Biology 9:675-678. Chapman, C.A. and L.J. Chapman. 1996. Frugivory and the fate of dispersed and nondispersed seeds of six African tree species. Journal of Tropical Ecology 12:491-504. Charies-Dominque, P., M. Atramentowitz, M. Charles-Dominique, H. Gerard, A. Hladik, CM. Hladik, and M.F.Prevost. 1981. Les mammiferes frugivores arboricoles nocturnes d'une foret guyanaise: inter-relations plantes-animaux. Rev. Ecol. (Terre et Vie) 35: 341-435.

PAGE 156

147 Clay, J. 1992. Some general principles and strategies for developing markets in North America and Europe for nontimber forest products, pp. 302-309, In; M. Plotkin and L. Famolare, eds. Sustainable harvest and marketing of rain forest products. Island Press, Washington, DC, USA. Clay, J. 1997. The impact of palm heart harvesting in the Amazon estuary. Pp. 283314, In: C.H. Freese, ed. Harvesting vsald species: implications for biodiversity conservation. The Johns Hopkins University Press, Baltimore, MD, USA. Crawley, M.J. 1992. Seed predators and plant population dynamics, pp. 157 191, In: M. Fenner, ed. Seeds: the ecology of regeneration in plant communities. CAB International, Wallington, UK. Crome, F.H.J. 1975. The ecology of fruit pigeons in tropical Queensland. Australian Wildlife Research 2:155-185. Davidar, P. and E.S. Morton. 1986. The relationship between fruit crop sizes and fruit removal rates by birds. Ecology 67:262-265. Davis, D.E. 1945. The annual cycle of plants, mosquitoes, birds and mammals in two Brazilian forests. Ecological Monographs 15:243-295. DeSteven, D. and F.E. Putz. 1984. Impact of mammals on eariy recruitment of a tropical canopy tree, Dipteryx panamensis, in Panama. Oikos 43:207-216. Develey, P.F. and C .A. Peres. 2000. Resource seasonality and the structure of mixed species flocks in a coastal Atlantic forest of southeastern Brazil. Journal of Tropical Ecology 16:33-53. Dice, L.R. 1945. Measures of the amount of ecological association between species. Ecology 26:297-302. Dickinson, C. and G. Pugh. 1974. Biology of plant litter decomposition. Academic Press New York, NY, USA. Dirzo, R. and C.A. Dominguez. 1986. Seed shadows, seed predation and the advantages of dispersal, pp. 237-249, In: A. Estrada and T.H. Fleming, eds. Frugivores and seed dispersal. Dr W. Junk Publishers, Dordrecht. Ducke, A. and G.A. Black. 1953. Phytogeographical notes on the Brazilian Amazon. An Acad. Bras. Cienc. 25:1-46. Emmons, L.H. 1997. Neotropical Rainforest Mammals: A field guide. University of Chicago Press, Chicago, IL, USA.

PAGE 157

148 Espadaler, X. and C Gomez. 1996. Seed production, predation, and dispersal in the Mediterranean myrmtcocoxt Euphorbia characias (Euphorbiaceae) Ecography 19:715. Feamside, P.M. 1989. Extractive reserves in Brazilian Amazonia. Bioscience. 39:387393. Fenner, M. 1985. Seed ecology. Chapman and Hall, London, UK. Ferreira, L.V., S.S. Almeida, C.S. Rosario. 1997. As areas de inunda9ao. pp. 195-211, In: P. L B. Lisboa, ed. Caxiuana. Museu Paraense Emilio Goeldi, Belem, Brazil. Fisher, E.A. 1997. The role of plumes m Eriotheca pentaphylla (Bombacaceae) seed survival in southeastern Brazil. Journal of Tropical Ecology 13:133-138. Fleming, T.H. 1991 . Fruiting plant frugivore mutualisms: the evolutionary theater and the ecological play. pp. 119-144, In: Plant-animal interactions. P.W. Price, T.M. Lewinsohn, G.W. Femandes, and W.W. Bensen. Wiley-Interscience Publications, New York, NY, USA. Fleming, T.H. 1992. How do fruitand nectar-feeding animals track their food resources? pp. 355-391, In: M.D. Hunter, T. Ohgushi, and P.W. Price, eds. Effects of resource distribution on animal-plant interactions. Academic Press, San Diego, CA, USA. Fogden, M.P.L. 1972. The seasonality and population dynamics of equatorial forest birds in Sarawak. Ibis 114:307-343. Forget, P.-M. and T. Milleron. 1991. Evidence for secondary seed dispersal by rodents in Panama. Oecologia 87:596-599. Foster, R.B. 1982a. Famine on Barro Colorado Island, pp. 201-212. In: E.G. Leigh, Jr., A S. Rand, and D M, Windsor, eds. The ecology of a tropical forest. Smithsonian Institution Press, Washington, D C. Foster, R.B. 1982b. The seasonal rhythm of fruitfall on Barro Colorado Island, pp. 151172. In: E.G. Leigh, Jr., A S. Rand, and D M. Windsor, eds. The ecology of a tropical forest. Smithsonian Institution Press, Washington, DC, USA. Fragoso, J.M.V. 1994. The community dynamics of a tropical rainforest: the importance of large mammals. PhD Thesis, University of Florida, Gainesville, FL, USA. Franken, M. 1979. Major nutrient and energy contents of the litterfall of a riverine forest of central Amazonia. Tropical Ecology 2:21 1-224. Franken, M., U. Irmler, and H. Klinge. 1979. Litterfall in inundation, riverine, and terra firme forests of central Amazonia. Tropical Ecology 20:225-235.

PAGE 158

149 Freese, C.H. 1997. Harvesting Wild Species: implications for biodiversity conservation The Johns Hopkins University Press, Baltimore. FrumhofF, P.C. 1995. Conserving wildlife in tropical forests managed for timber. Bioscience 45: 456-464. Furch, K. 1997. Chemistry of varzea and igapo soils and nutrient inventory of their floodplain forests. Pp. 47-67, In: W.J. Junk, ed. The Central Amazon floodplain SpringerVerlag, Berlin, Germany. Furch, K. and W.J. Junk. 1997a. Physicochemical conditions in the floodplains. Pp. 69109, In: W.J. Junk, ed. The Central Amazon floodplain. SpringerVerlag, Berlin, Germany. Furch, K. and W.J. Junk 1997b. The chemical composition, food value, and decomposition of herbaceous plants, leaves, and leaf litter of floodplain forests. Pp. 187-205, In: W.J. Junk, ed. The Central Amazon floodplain. SpringerVerlag, Berlin, Germany. Galetti, M. and A. Aleixo. 1998. Effects of palm heart harvesting on avian frugivores in the Atlantic rain forest of Brazil. Journal of Applied Ecology 35:286-293. Galetti, M., V.B. Zipparro, and P C Morellato. 1999. Fruiting phenology and frugivory on the palm Euterpe edulis in a lowland Atlantic forest of Brazil. Ecotropica 5:115122. Gallaher, R. N., C. 0. Weldon and J. G. Futral. 1975. An aluminum block digester for plant and soil analysis. Soil Science Society America Proceedings 39:803-806. Gentry, A H. 1982. Patterns of neotropical plant species diversity. Evolutionary Biology 15:1-84. Gentry, A. 1992. New non-timber foret products from western South America, pp. 125136, In: M. Plotkin and L. Famolare, eds. Sustainable harvest and marketing of rain forest products. Island Press, Washington, DC, USA. Gomez-Pompa, A. and A. Kaus. 1990. Traditional management of tropical forests in Mexico. Pp. 45-64, In: A.B. Anderson, ed. Alternatives to deforestation. Columbia University Press, New York, NY, USA, Gottsberger, G. 1978. Seed dispersal by fish in the inundated regions of Humaita, Amazonia. Biotropica 10:170-183. Goulding, M. 1980. The Fishes and the Forest. University of California Press, Berkeley CA, USA,

PAGE 159

150 Greenberg, R. 1996. Managed forest patches and the diversity of birds in southern Mexico. Pp. 59-90, In: J. Schelhas and R. Greenberg, eds. Forest patches in tropical Landscapes. Island Press, Washington, DC, USA. Greenberg, R., P. Bichier, and J. Sterling. 1997. Bird populations in rustic and planted shade coffee plantations of Eastern Chiapas, Mexico. Biotropica 29:501-514. Greene, D.F. and E.A. Johnson. 1997. Secondary dispersal of tree seeds on snow. Journal of Ecology 85:329-340. Guppy, H.B. 1917. Plants, seeds, and currents in the West Indies and Azores. Macmillan Press, London, UK. Hamann, A. and E. Curio. 1999. Interactions among frugivores and fleshy fruit trees in a Philippine submontane rainforest. Conservation Biology 13:766-773. Hambleton, L. G. 1977. Semiautomated method for simultaneous determination of phosphorus, calcium and crude protein in animal feeds. J.A.O.A.C. 60:845-852. Henderson, A. 1995. The palms of the Amazon. Oxford University Press, New York, NY, USA. Henderson, P A. and I. Walker. 1986. On the leaf litter community of the Amazonian blackwater stream Tarumazinho. Journal of Tropical Ecology 2:1-17. Henderson, P A. and W.G.R. Crampton. 1997. A comparison of fish diversity and abundance between nutrient-rich and nutrient-poor lakes in the upper Amazon. Journal of Tropical Ecology 13:175-198. Herrera, CM. 1992. Interspecific variation in fruit shape: allometry, phylogeny, and adaptation to dispersal. Ecology 73:1832-1841. Herrera, CM. 1998. Long-term dynamics of mediterranean frugivorous birds and fleshy fruits: a 12-year study. Ecological Monographs 68:51 1-538. Hilty, S.L. and W.L. Brown. 1986. A guide to the birds of Colombia. Princeton University Press, Princeton, NJ, USA. Hiraoka, M. 1995. Land use changes in the Amazon estuary. Global Environmental Change 5:323-336. Hladik, A., and CM. Hladik. 1969. Rapports trophiques entre vegetation et primates dans la foret de Barro Colorado (Panama). Revue d'Ecologie la Terre et la Vie 23 25117.

PAGE 160

151 Howe, H.F. 1979. Fear and frugivory, American Naturalist 1 14:925-93 1 . Howe, H.F. and Vande Kerckove, G.A. 1980. Nutmeg dispersal by tropical birds. Science 210:925-927. Howe, H.F. and J. Smallwood. 1982. Ecology of seed dispersal. Annual Review of Ecology and Systematics 13:201-228. Hubbell, S .P. 1980. Seed predation and the coexistence of tree species in tropical forests. Oikos 35:214-229. Huber, J. 1910. Matas e madeiras amazonicas. Boletim Museu Paraense Emilio Goeldi, Historia Natural Ethnographico 6:91-225. Huenneke, L.F. and R.R. Sharitz. 1986. Microsite abundance and distribution of woody seedlings in a South Carolina cypress-tupelo swamp. American Midland Naturalist 115:328-335. Irmler, U. 1975. Ecological studies of the aquatic soil invertebrates in three inundation forests of central Amazonia. Amazoniana 3:337-409. James, F.C. and C.E. McCulloch. 1990. Multivariate analysis in ecology and systematics: panacea or pandora's box? Annual Review of Ecology and Systematics 21:129-166. James, F.C. and C.E. McCulloch. 1995. The strength of inferences about causes of trends in populations, pp. 40-51, In: T.E. Martin and D M. Finch, eds. Ecology and management of neotropical migratory birds. Oxford University Press, New York, NY, USA. Janson, C.H. and L.H. Emmons. 1990. Ecological structure of the nonflying mammal community at Cocha Cashu biological station, Manu National Park, Peru. pp. 314338. In: A H. Gentry, ed. Four Neotropical Rainforests. Yale University Press, London, UK. Janzen, D.H. 1970. Herbivores and the number of tree species in tropical forests. American Naturalist 104:501-528. Janzen, D.H. 1971. Seed predation by animals. Annual Review of Ecology and Systematics 2:465-492. Janzen, D.H. 1974. Tropical blackwater rivers, animals, and mast fruiting by the Dipterocarpaceae. Biotropica 6:69-103. Jardim, M A G, and J.S. Rombold. 1994. Effects of adubation and thinning on a9ai palm {Euterpe oleracea) fruit yield from a natural population. Boletim do Museu Paraense Emilio Goeldi, serieBot. 10:283-293.

PAGE 161

152 Jardim, M.M.A. & Oliveira, L.F. 1997. Uso do espafo de Alouatta helzehul (Primates, Cebidae) em fun9ao da temporalidade de recursos. pp. 417-436, In: P.L.B. Lisboa, ed. Caxiuana. Museu Paraense Emilio Goeldi, Belem, Para, Brazil Johns, A.D. 1988. Effects of "selective" timber extraction on rain forest structure and composition and some consequences for frugivores and folivores. Biotropica 20;3 137. Johnson, C.E., S. Zona, and J. A. Nilsson. 1995. Bruchid beeties and palm seed hosts. Principes 39: 25-35. Jones, R.H., R.R. Sharitz, P.M. Dixon, D.S. Segal, and R.L. Schneider. 1994. Woody plant regeneration in four floodplain forests. Ecological Monographs 64:345-367. Jordano, P. 1992. Fruits and frugivory, pp. 105-156, In: M. Fenner, ed. Seeds: the ecology of regeneration in plant communities. Cab Press, Wallington, U.K. Junk, W.J. 1997. General aspects of floodplain ecology with special reference to Amazonian floodplains. pp. 3-20, In: W.J. Junk, ed. The Central Amazon floodplain. SpringerVerlag, Beriin, Germany. Junk, W.J. and B. A. Robertson. 1997. Aquatic invertebrates. Pp. 279-298, In: W.J. Junk, ed. The Central Amazon floodplain. SpringerVerlag, Beriin, Germany. Junk, W.J., M.G.M. Soares, and U. Saint-Paul. 1997. The fish. Pp. 385-408, In: W.J. Junk, ed. The Central Amazon floodplain. SpringerVerlag, Beriin, Germany. Kahn, F. 1991. Palms as key swamp resources in Amazonia. Forest Ecology and Management. 38:133-142. Kainer, K.A., M.L. Duryea, N. Costa de Macedo, and K. Williams. 1998. Brazil nut seedling establishment and autecology in extractive reserves of Acre, Brazil. Ecological Applications 8:397-410. Kaufmann, S., D.B. McKey, M. Hossaert-McKey, and C.C. Horvitz, 1991. Adaptations for a two-phase seed dispersal system involving vertebrates and ants in a hemiepiphytic fig (Ficus microcarpa. Moraceae). American Journal of Botany 78:971-977. Karr, J R. 1976. Seasonality, resource availability, and community diversity in tropical bird communities. American Naturalist 110: 973-994. Karr, J.R. 1981. Surveying birds with mistnets. Pages 62-67 in: C.J. Ralph and J.M. Scott, eds. Estimating numbers of terrestrial birds. Studies in Avian Biology 6. Cooper Ornithological Society, Los Angeles, CA, USA.

PAGE 162

153 Karr, J R. and K.E. Freemark. 1983. Habitat selection and environmental gradients: dynamics in the "stable" tropics. Ecology 64:1481-1494. Karr, J.R., S. Robinson, J.G. Blake, and R.O. Bierregaard, Jr. 1990. Birds of four neotropical forests, pp. 237-272. In: A.H. Gentry, ed. Four Neotropical Rainforests. Yale University Press, London, UK. Kendeigh, S.C. 1945. Community selection in birds on the Heidelberg Plateau of New York. Auk 62:418-436. Kiker, C.F. and F.E. Putz. 1997. Ecological certification of forest products: economic challenges. Ecological Economics 20:37-51. Kinnaird, M.F., T O. O'Brien, and S. Suryadi. 1996. Population fluctuation in Sulawesi Red-knobbed Hombills: tracking figs in space and time. Auk 1 13:431-440. Klinge, H. 1977. Fine litter production and return to the soil in three natural forest stands of eastern Amazonia. Geo-Eco-Trop 1:159-167. Knoppel, H-A. 1970. Food of central Amazonian fishes. Amazoniana 2:257-352. Kubitzki, K. and Ziburski, A. 1994. Seed dispersal in floodplain forests of Amazonia. Biotropica 26:30-43. Lambert, F.R. 1992. The consequences of selective logging for Bomean lowland forest birds. Phil. Trans. R. Soc. Lond. B 335:443-457. Leighton, M. and D R. Leighton. 1983. Vertebrate responses to fruiting seasonality within a Bomean rain forest, pp. 181-196. In: S.L. Sutton, T.C. Whitmore, and A.C. Chadwick, eds. Tropical rain forest: ecology and management. Blackwell Scientific Publications, London, UK. Lemmon, P.E. 1957. A new instrument for measuring forest overstory density. Journal of Forestry 55:667-668. Levey, D.J. 1988. Spatial and temporal variation in Costa Rican fruit and fruit-eating bird abundance. Ecological Monographs 58: 251-269. Levey, D.J. and F.G. Stiles. 1992. Evolutionary precursors of long-distance migration: resource availability and movement patterns in neotropical landbirds. American Naturalist 140:447-476.

PAGE 163

154 Levey, D.J. and F.G. Stiles. 1994. Birds: ecology, behavior, and taxonomic affinities, pp. 217-228, In; L.A. McDade, K.S. Bawa, H.A. Hespenheide, and G.S. Hartshorn, eds. La Selva: ecology and natural history of a neotropical rain forest.. University of Chicago Press, Chicago. Levey, D.J. and M.M. Byrne. 1993. Complex ant-plant interactions: rain forest ants as secondary dispersers and post-dispersal seed predators. Ecology 74:1802-1812. Levey, D.J. and C.W. Benkman. 1999. Fruit-seed disperser interactions: timely insights from a long-term perspective. Trends in Ecology and Evolution 14:41-43. Lisboa, P.L.B. 1997. Esta?ao Cientiifica Ferreira Penna. pp. 23-49, In: P.L.B. Lisboa, ed. Caxiuana. Museu Paraense Emilio Goeldi, Belem. Lodge, D.L., W.H. McDowell, and CP. McSwiney. 1994. The importance of nutrient pulses in tropical forests. Trends in Ecology and Evolution 9:384-387. Loiselle, B.A. and J.G. Blake. 1991. Resource abundance and temporal variation in fruiteating birds along a wet forest elevational gradient in Costa Rica. Ecology 72: 180193. Loiselle, B.A. and J.G. Blake. 1993. Spatial distribution of understory fruit-eating birds and fruiting plants in a neotropical lowland wet forest, pp. 177-189, In: T.H. Fleming and A. Estrada, eds. Frugivory and seed dispersal: ecological and evolutionary aspects. Kluwer Academic Publications, Dordrecht. Loiselle, B.A. and J.G. Blake, in prep. Potential consequences of local frugivore extinction for plant populations. Lovejoy, T.E. 1974. Bird diversity and abundance in Amazon forest communities. Living Bird 13:127-191. Lugo, A.E. and J.L. Frangi. 1993 Fruit fall in the Luquillo Experimental Forest, Puerto Rico. Biotropica 25:73-84. MacArthur, R.H. and J.W. MacArthur. 1961. On bird species diversity. Ecology 42 594598. McCann, J.M. 1999. Before 1492: The making of the pre-Colombian landscape. Part II: The vegetation, and implications for restoration for 2000 and beyond. Ecological Restoration 17:107-119. Mack, A.L. 1995. Distance and non-randomness of seed dispersal by the dwarf cassowary Casuarius bennetti. Ecography 18:286-295.

PAGE 164

155 McKey, D. 1975. The ecology of coevolved seed dispersal systems, pp. 159-191, In; L.E. Gilbert and P. Raven, eds. Coevolution of animals and plants. University Texas Press, Austin, TX, USA. Mai comb, J.R. 1988. Small mammal abundances in isolated and non-isolated primary forest reserves near Manaus, Brazil. Acta Amazonica 18:67-83. Marascuilo, L A. and M. McSweeney. 1977. Nonparametric and distribution-free methods for the social sciences. Brooks/Cole Publications, Monterey, CA. Marlier, G. 1967. Ecological studies on some lakes of the Amazon valley. Amazoniana 1:91-115. Marsden, S.J. 1998. Changes in bird abundance following selective logging on Seram, Indonesia. Conservation Biology 12:605-611. Martin, T.E. and J R. Karr. 1986. Temporal dynamics of Neotropical birds with special reference to frugivores in second-growth woods. Wilson Bull. 98:38-60. Mason, D. 1996. Responses of Venezuelan understory birds to selective logging, enrichment strips, and vine cutting. Biotropica 28:296-309. Mattoso, M R. and V. Fleischflesser. 1994. Amapa Reserva extrativista do Rio Cajari. pp. 91-149. In: O Destino da Floresta: Reservas extativistas e desenvolvimento sustentavel na Amazonia. Rio de Janeiro:Relume-Dumara, Brazil. Moegenburg, S.M. and D.J. Levey, in prep. Linking fruit and frugivore abundance: experimental evidence from Amazonian Brazil. Moguel, P. and V.M. Toledo. 1999. Biodiversity conservation in traditional coffee production systems of Mexico. Conservation Biology 13:11-21. Morse, D.H. 1976. Variables affecting the density and territory size of breeding sprucewoods warblers. Ecology 57:290-301. Muiiiz-Mirit, N., R. Vamos, M. Hiraoka, F. Montagnini, and R. Mendelson. 1996. The economic value of managing the afai palm (Euterpe oleracea Mart.) in the floodplains of the Amazon estuary. Para, Brazil. Forest Ecology and Management 87:163-173. Murray, D R. 1986. Seed dispersal by water, pp. 49-85 In: D R. Murray, ed. Seed dispersal. Academic Press, Sydney, Australia. Nepstad, DC, F. Brown, L. Luz, A. Alechandra, and V. Viana. 1992. Biotic impoverishment of Amazonian forests by rubber tappers, loggers, and cattle ranchers, pp. 1-14, In: D C. Nepstad and S. Schwartzman, eds. Non-Timber Products from Tropical Forests. Advances in Economic Botany 9, New York, NY, USA.

PAGE 165

156 Nepstad, D C. and S. Schwartzman. 1992. Non-timber products from tropical forests: evaluation of a conservation and development strategy. Advances in Economic Botany 9, NY, USA. Novaes, F.C. 1970. Distribuigao ecologica e abundancia das aves em um trecho da mata do baixo Rio Guama (Estado do Para). Boletim Museu Paraense Emilio Goeldi 1 VISA. Ohkawara, K. and S. Higashi. 1994. Relative importance of ballistic and ant dispersal in two diplochorous Viola species (Violaceae). Oecologia 100:135-140. Oren, D.C. and F.C. Novaes. 1986. Observations on the golden parakeet (Aratinga guarouba) in Northern Brazil. Biological Conservation 36: 329-337. Padoch.C. 1992. Marketing of non-timber forest products in western Amazonia: general observations and research priorities. Advances in Economic Botany 9:43-50. Panayotou, T. and P. S. Ashton. 1992. Not by timber alone. Island Press, Washington, DC, USA. Parker, T.A. III., D.F. Stotz, and J.W. Fitzpatrick. 1996. Ecological and distributional databases, pp. 132-407, In: D.F. Stotz, J.W. Fitzpatrick, T.A. Parker m., and D.K. Moskovits, eds. Neotropical birds: ecology and conservation. University Chicago Press, Chicago, IL, USA. Peck K.S., O. Henry, and D. Sabatier. 1999. The insectivorous-frugivorous diet of the golden-handed tamarin (Saguinus midas midas) in French Guiana. Folia Primatologica 70:1-7. Pena, M.C. 1996. Ecology and socioeconomics of palm heart extraction from wild populations of Euterpe precatoria Mart, in eastern Bolivia. MS Thesis. University of Florida, Gainesville, FL, USA Peres, CM. 2000. Identifying keystone plant resources from tropical forests: the case of gums from Par^/a pods. Journal of Tropical Ecology 16:287-317. Peters, CM. 1990. Population ecology and management of forest fruit trees in the Peruvian Amazon, pp. 86-98. In: A.B. Anderson, ed. Alternatives to deforestation. Columbia University Press, New York, NY, USA. Peters, CM. 1992. The ecology and economics of oligarchic forests, pp. 15-22, In: D.C Nepstad and S. Schwartzman, eds. Non-timber products from tropical forests. Advances in Economic Botany 9, New York.

PAGE 166

157 Peters, CM. 1996 The ecology and management of non-timber forest resources. World Bank technical paper number 322. Washington, DC, USA. Peters, CM., M.J. Balick, F. Kahn, and A.B. Anderson. 1989. Oligarchic forests of economic plants in Amazonia: utilization and conservation of an important tropical resource. Conservation Biology. 3:341-349. Pina, A.L.C.B. 1999. Dinamica socio-ecologica em uma popula?ao de guaribas-das-maosvermelhas {Alouatta belzebul) na Estafao Cientifica. MS Thesis, Federal University of Para, Brazil. Pinedo-Vasquez, M. and C. Padoch. 1996. Managing forest remnants and forest gardens in Peru and Indonesia. Pp. 327-342, In: J. Schelhas and R. Greenberg, eds. Forest Patches in Tropical Landscapes. Island Press, Washington, DC, USA. Pires, J.M. 1974. Tipos de vegetafao da Amazonia. Bras. Flor. 5:48-58. Plotkin, M. and L. Famolare. 1992. Sustainable harvest and marketing of rain forest products. Island Press, Washington, DC, USA. Pollack, H., M. Mattos, and C. Uhl. 1995. A profile of palm heart extraction in the Amazon Estuary. Human Ecology 23:357-385. Powell, G.V.N. andR. Bjork. 1995. Implications of intratropical migration on reserve design: a case study using Pharomachrus mocirmo. Conservation Biology 9:354-362. Prance, G.T. 1978. The origin and evolution of the Amazonian flora. Interciencia 3(4): 207-222. Pratt, T.K. and E.W. Stiles. 1983. How long do fruit-eating birds stay in the plants where they feed: implications for seed dispersal. American Naturalist 122:797-805. Putz, F.E. and M. Holbrook. 1986. Notes on the natural history of hemi-epiphytes. Selbyana 9:61-69. Putz, F.E., K.H. Redford, J.G. Robinson, R. Fimbel, and G.M. Blate. 2000. Biodiversity conservation in the context of tropical forest management. Wildlife Conservation Society technical paper, Overlays Program, Impact Series. Redbo-Torstensson, P. and A. Telenius. 1995. Primary and secondary seed dispersal by wind and water in Spergularia salina. Ecography 18:230-237. Remsen, J.V. and D A. Good. 1996. Misuse of data from mist-nets captures to assess relative abundance in bird populations. Auk 1 13:381-398.

PAGE 167

158 Rey, P.J. 1995 Spatio-temporal variation in fruit and frugivorous bird abundance in olive orchards. Ecology 76:1625-1635. Rice, W.R. 1989. Analyzing tables of statistical tests. Evolution 43:223-225. Richards, M. 1993. The potential of non-timber forest products in sustainable natural forest management in Amazonia. Commonwealth Forestry Review. 72:21-27. Ricker, M., R.O. Mendelsohn, D.C. Daly, and G. Angeles. 1999. Enriching the rainforest with native fruit trees: an ecological and economic analysis in Los Tuxtlas (Veracruz, Mexico). Ecological Economics 3 1 :439-448. Ridgely, R.S. and G. Tudor. 1994. The birds of South America. Vol. n. University of Texas Press, Austin, TX, USA. Ridley, H.N. 1930. The dispersal of plants throughout the world. Reeve and Company, Ashford, Kent, UK. Robinson, J.G. and K.H. Redford. 1986. Body size, diet, and population density of neotropical forest mammals. American Naturalist. 128:665-680. Robinson, S.K. and R.T. Holmes. 1982. Foraging behavior of forest birds: the relationships among search tactics, diet, and habitat structure. Ecology 63:1918-1931. Roosevelt, AC, M. Lima da Costa, C. Lopes Machado, M, Michab, N. Mercier, H. Valladas, J. Feathers, W. Bamett, M. Imazio da Silveira, A. Henderson, J. Sliva, B. Chemoff, D.S. Reese, J. A. Holman, N. Toth, and K. Schick. 1996, Paleoindian cave dwellers in the Amazon: The peopling of the Americas. Science 272:373-384. Roosmalen, M.G. van. 1985. Fruits of the Guianan flora. Utrecht University, Wageningen. The Netherlands. Roth, I. 1987. Stratification of a tropical forest as seen in dispersal types. Dr W Junk Publ., Lancaster, UK. SAS Institute, 1996. JMP start statistics. Duxbury Press, Belmont, CA, USA. Salafsky, N., B.L. Dugelby, and J.W. Terborgh. 1993. Can extractive reserves save the rainforest? An ecological and socioeconomic comparison of non-timber forest product extraction systems in Peten, Guatemala, and West Kalimantan, Indonesia. Conservation Biology. 7:39-52. Salick, J., A. Mejia, and T. Anderson. 1995. Non-timber forest products integrated with natural forest management, Rio San Juan, Nicaragua. Ecological Applications 5 878895.

PAGE 168

159 Sargent, S 1990. Neighborhood effects on fruit removal by birds, a field experiment with Viburnum dentatum (Caprifoliaciae). Ecology 71 ; 1289-1298. Scarano, F.R., K.T. Ribeiro, L.F.D. De Moraes, and H.C. De Lima. 1997. Plant establishment on flooded and unflooded patches of a freshwater swamp forest in southeastern Brazil. Journal of Tropical Ecology 14:793-803. Schneider, R.L. and R.R. Sharitz. 1988. Hydrochory and regeneration in a bald cypresswater tupelo swamp forest. Ecology 69:1055-1063. Schulze, P.C., M. Leighton, and D R. Peart. 1994. Enrichment planting in selectivelylogged rain forest: a combined ecological and economic analysis. Ecological Applications 4:581-592. Schuup, E.W. 1993. Quantity, quality, and the effectiveness of seed dispersal by animals, pp. 15-29, In: T.H. Fleming and A. Estrada, eds. Frugivory and Seed dispersal: ecological and evolutionary aspects. Kluwer Academic Publishers, Dordrecht. Setaro, F.V. and J.M. Melack. 1984. Responses of phytoplankton to experimental nutrient enrichment in an Amazon floodplain lake. Limnology and Oceanography 29:972-984. Shankar, U., K.S. Murali, R. Uma Shaankar, K.N. Ganeshaiah, amd K.S. Bawa. 1996. Extraction of non-timber forest products in the forests of Biligiri Rangan Hills, India. 3. productivity, extraction and prospects of sustainable harvest of amla Phyllanthus emblica (Euphorbiaceae). Economic Botany 50:270-279. Shepherd, P.C.F. and J.S. Boates. 1999. Effects of commercial baitworm harvest on Semipalmated Sandpipers and their prey in the Bay of Fundy Hemispheric Shorebird Reserve. Conservation Biology 13:347-356. Sherman, P.T. and P.K. Eason. 1998. Size determinants in territories with inflexible boundaries: manipulation experiments on white-winged trumpeters' territories. Ecology 79:1147-1159. Sick, H. 1993. Birds in Brazil: A natural history. Princeton University Press, Princeton, NJ, USA. Siemens, D.H. and CD. Johnson. 1996. Bruchid oviposition patterns beneath Guanacaste trees (Mimosaceae) in Venezuela: probable consequences of extinct seed dispersers. Biotropica 28:96-104. Silva Matos, D.M. and A.R. Watkinson. 1998. The fecundity, seed, and seedling ecology of the edible palm Euterpe edulis in southeastern Brazil. Biotropica 30:595-603.

PAGE 169

160 Sioli, H. 1966. General features of the delta of the Amazon. Pp 381-391, In: Humid tropics research. Scientific problems of the humid tropical zone deltas and their implications. Proceedings of the Daaca Symposium. UNESCO. Sioli, H. 1968. Hydrochemistry and geology in the Brazilian Amazon region. Amazoniana 1:267-277. Snow, D. 1962a. A field study of the Black and White Manakin, Manacus manacus, in Trinidad. Zoologica 47:65-104. Snow, D. 1962b. A field study of the Golden-headed Manakin, Pipra erythrocephala, in Trinidad. Zoologica 47:183-198. Snow, B. and D Snow. 1980. Birds and berries. T&AD Poyser, Calton, UK. Snyder, N.F., J.W. Wiley, and C.B. Kepler. 1987. The Parrots of Luquillo: natural history and conservation of the Puerto Rican Parrot. Western Foundation of Vertebrate Zoology, Los Angeles, CA, USA. Snyder, N.F.R., F.C. James, and S.R. Beissinger. 1991. Toward a conservation strategy for Neotropical psittacines. pp. 257-276, In: S R. Beissinger and N.F.R. Snyder, eds. New World parrots in crisis. Smithsonian Institution Press, Washington, DC, USA. Stouffer, P C. and R. 0. Bierregaard, Jr. 1995. Use of Amazonian forest fragments by understory insectivorous birds. Ecology 76:2429-2445. Strahl, S.D. and A. Grahal. 1991. Conservation of large avian frugivores and the management of Neotropical protected areas. Oryx 25: 50-55. Strudwick, J. and G.L. Sobel. 1988. Uses of Euterpe oleracea in the Amazon estuary, Brazil, pp. 225-253, In: The Palm Tree of Life: Biology, Utilization, and Conservation. Advances in Economic Botany 6. Terborgh, J. 1977. Bird species diversity on an Andean elevational gradient. Ecology 58:1007-1019. Terborgh, J. 1986a. Community aspects of frugivory. pp. 371-384 In: A. Estrada and T.H. Fleming, eds. Frugivores and Seed Dispersal. Dr W Junk Publishers, Dordrecht. Terborgh, J. 1986b. Keystone plant resources in the tropical forest, pp. 330-344, In: M.E. Soule, ed. Conservation Biology: The Science of Scarcity and Diversity. Sinauer Associates, Inc. Sunderland. Terborgh, J., S.K. Robinson, T.A. Parker, III, CA. Munn, and N. Pierpont. 1990. Structure and organization of an Amazonian forest bird community. Ecological Monographs 60:213-238.

PAGE 170

161 ThioUay, J-M. 1992. The influence of selective logging on bird species diversity in a Guianan rain forest. Conservation Biology 6:47-63. Thiollay, J.-M. 1994. Structure, density, and rarity in an Amazonian rainforest bird community. Journal of Tropical Ecology 10:449-481. Thiollay, J-M. 1995. The role of traditional agroforests in the conservation of rain forest bird diversity in Sumatra. Conservation Biology 9:335-353. Thompson, J.N. 1981. Elaiosomes and fleshy fruits: phenology and selection pressures for ant-dispersed seeds. American Naturalist 117:104-108. Trexler, J.C. and J. Travis. 1993. Nontraditional regression analyses. Ecology 74:16291637. Vasquez, R. and A. H. Gentry. 1989. Use and misuse of forest-harvested fruits in the Iquitosarea. Conservation Biology 3:350-361. Viana, V.M., A.R Pierce, and R Z. Donovan. 1996. Certification of non-timber forest products, pp. 145-163, In: V.M. Viana, J. Ervin, R. Donovan, C. Elliot, and H. Gholz, eds. Certification of forest products: issues and perspectives. Island Press, Washington, DC, USA. Vitousek, P.M. 1984. Litterfall, nutrient cycling, and nutrient limitation in tropical forests. Ecology 65:285-298. Waldhoff, D., U. Saint-Paul, and B. Furch. 1996. Value of fruits and seeds in the floodplain forests of central Amazonian as food resource for fish. Ecotropica 2: 143156. Warren, L.A. 1992. ^u/er/?^ palms in northern Brazil: market structure and socioeconomic implications for sustainable management. MS Thesis. University of Florida, Gainesville, FL, USA. Wenny, D.G. 1999. Two-stage dispersal of Guarea glabra and G. kunthiana (Meliaceae) in Monteverde, Costa Rica. Journal of Tropical Ecology 15:481-496. Westcott, D A. and D.L. Graham. 2000. Patterns of movement and seed dispersal of a tropical frugivore. Oecologia 122:249-257. Wheelwright, NT. 1983. Fruits and the ecology of Resplendent Quetzals. Auk 100:286-301. Wheelwright, N T. 1988. Fruit-eating birds and bird-dispersed plants in the tropics and temperate zone. Trends in Ecology and Evolution 3 :270-274.

PAGE 171

162 Wheelwright, N T. 1991. How long do fruit-eating birds stay in the plants where they feed? Biotropica 23:29-40. Whitmore, T. C. 1990. An Introduction to Tropical Rainforests. Clarendon Press, Oxford, UK. Williamson, G.B., F. Costa, and C.V. Minte Vera. 1999. Dispersal of Amazonian trees; hydrochory in Swartzia polyphylla. Biotropica 3 1 :460-465. Williamson, G.B. and F. Costa. 2000. Dispersal of Amazonian trees: hydrochory in Pentaclethra macroloba. Biotropica 32:548-552. Witkowski, E.T.F. and B.B. Lamont. 1996. Nutrient losses from commercial picking and cockatoo removal of Banksia hookeriana blooms at the organ, plant, and site levels. Journal of Applied Ecology 33:131-140. Wollenberg, E. and A. Ingles. 1998. Incomes from the forest: methods for the development and conservation of forest products for local communities. CIFOR, Bogor, Indonesia. Worthington, A. 1982. Population sizes and breeding rhythms of two species of manakins in relation to food supply, pp. 213-225. In: E.G. Leigh, Jr., A.S. Rand, and D.M. Windsor, eds. The ecology of a tropical forest. Smithsonian Institution Press, Washington, DC, USA. Wright, S.J. 1983. The dispersion of eggs by a bruchid beetle among Scheelea palm seeds and the effect of distance to the parent palm. Ecology 64: 1016-1021 . Wright, S.J. 1990. Cumulative satiation of a seed predator over the fruiting season of its host. Oikos 58:272-276. Wright, S.J., C. Carrasco, O. Calderon, and S. Paton. 1999. The El Niiio Southern Oscillation, variable fruit production, and famine in a tropical forest. Ecology 80:1632-1647. Zar, J.H. 1996. Biostatistical analysis. Third edition. Prentice Hall, NJ, USA. Zolman, J.F. 1993. Biostatistics. Oxford University Press, New York, NY, USA.

PAGE 172

BIOGRAPHICAL SKETCH Susan Moegenburg was bom in Milwaukee, Wisconsin, in 1966. She was shepherded through early life by her brother David and sister Julie, and frequently tormented by brother Peter. She lived in Wisconsin for her first 22 years, attending public schools and the University of Wisconsin Madison as an undergraduate. Several events swayed her choice of career. Her awakening to the beauty of birds was spurred on by her brother David, who, when she was 6 or 7, constructed a small paper American Robin that would sit on her shoulder and could, and did, go where she went. In her teenage years she took with friends annual trips to Wisconsin's north woods or Minnesota's boundary waters, where she felt more at home than she did in the more urban haunts of her family. Through these years, she was fortunate to have a number of teachers who recognized, and encouraged, her interest in the natural sciences. A foray into veterinary science in college proved short-lived; after a course in general ecology her path became clear to her and she has never looked back. A lecture by Peter Raven during her junior year in college planted the seed that would later grow into a true goal to travel and study in the Amazon basin. Susan moved to Florida to study tropical ecology. She settled for sub-tropical ecology during her Master's degree research on palm trees along Florida's Gulf Coast. 163

PAGE 173

164 During her PhD tenure she spent a total of 20 months in Brazil and hopes to continue her relationship with that wonderful country for many years to come.

PAGE 174

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. Douglas Levey, unairman Professor of Zoology 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. ; -€iv\/hv^ , Colin Chapman Associate Professor of Zoology 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. rancis Putz Professor of Botany 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. Assistant Professor of Wildlife Ecology and Conservation

PAGE 175

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. Richard Bodmer Assistant Professor of Wildlife Ecology and Conservation This dissertation was submitted to the Graduate Faculty of the Department of Zoology in the College of Liberal Arts and Sciences and to the Graduate School and was accepted as partial fulfillment of the requirements for the degree of Doctor of Philosophy. December, 2000 Dean, Graduate School


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 ED0F387XP_T0BJMC INGEST_TIME 2015-04-14T18:29:40Z PACKAGE AA00029913_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES