3~ \\

INVESTIGATING SEED DISPERSAL AND SEED PREDATION IN A HAWAllAN

DRY FOREST COMMUNITY

IMPLICATIONS FOR CONSERVATION AND MANAGEMENT

A THESIS SUBMITTED TO THE GRADUATE DIVISION OF THE UNIVERSITY OF HAWAI'I IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF

MASTER OF SCIENCE

IN

BOTANICAL SCIENCES

(BOTANY - ECOLOGY, EVOLUTION, AND CONSERVATION BIOLOGY)

DECEMBER 2004

By Charles G. Chimera

Thesis Committee:

Donald Drake, Chairperson Gerald Carr David Duffy Lloyd L. Loope ACKNOWLEDGEMENTS

I would like to thank the EECB Program, the Hawai'i Audubon Society, the Dai Ho

Chun Fellowship, and the Charles H. Lamoureux Fellowship in Plant Conservation for providing funding for my research. I would like to thank Sumner Erdman and Tony

Durso ofUlupalakua Ranch for facilitating access to my study site through ranch lands. I would like to thank the staffofthe Natural Area Reserve System, particularly Betsy

Gagne, Bill Evanson and Bryon Stevens, for providing access to the Kanaio Natural Area

Reserve as well as for logistical support. Lloyd Loope and Art Medeiros were extremely generous with their knowledge and guidance, factors that greatly influenced my decision to conduct research in Hawaiian dry forests. Lloyd Loope also provided vehicle support during the initial stages ofmy research. I thank my thesis committee chair, Don Drake, and my committee members, Gerry Carr, Dave Duffy and Lloyd Loope for providing me with guidance and assistance during all stages ofmy thesis research. Finally, I would especially like to thank my wife, Melissa, for providing me with unlimited patience and support during my entire graduate experience.

111 TABLE OF CONTENTS

ACKNOWLEDGEMENTS .111

LIST OF TABLES VI

LIST OF FIGURES VIII

CHAPTER 1. INTRODUCTION 1 Loss ofNative Seed Dispersers 4 Rodents and Seed Predation 5 Influence ofVegetation on Seed Dispersal and Predation 6 Hypotheses 8

CHAPTER 2. PATTERNS OF SEED DISPERSAL: VEGETATION STRUCTURE, FRUGIVORY AND SIZE EFFECTS 10 Introduction 10 Loss ofNative Seed Dispersers 12 Influence ofVegetation on Seed Dispersal 13 Methods 15 Study Site 15 Measurement ofSeed Rain Under Trees Versus in Exposed Areas 16 Seed Dispersal Analysis 19 Bird Observations 20 Seed Size Measurements 21 Characterization ofthe vegetation 22 Results 22 Seed Rain 22 Bird Observations 33 Seed Size and Bird Dispersal 35 Discussion 36

CHAPTER 3: POST-DISPERSAL SEED PREDATION: LOCATION AND EFFECT ON SPECIES 46 Introduction 46 Methods 49 Study Site 49 Measuring Seed Predation Levels 50 Treatments 51 Analysis 53 Results 54 Discussion 62

IV CHAPTER 4: CONCLUSION 68

APPENDIX A. TREE DIMENSIONS (BASAL DIAMETER, HEIGHT, CROWN DIAMETER) OF STUDY SPECIES AND DEAD TREES 73

APPENDIX B. PHENOLOGY OF STUDY TREES AND COMMON FLESHY- FRUITED SHRUBS OF THE KANAIO NATURAL AREA RESERVE 75 Methods 75

APPENDIX C. RELATIVE FREQUENCY AND ABUNDANCE OF NON-NATIVE BIRDS IN KNAR STUDY SITE 82 Methods 82

APPENDIX D. CHARACTERIZATION OF THE VEGETATION OF KNAR STUDY SITE 87 Methods 87

LITERATURE CITED 91

v LIST OF TABLES

Table 2.1. Mean seed density by dispersal vector and location at KNAR from April 2003 to March 2004 23

Table 2.2. Mean density under trees and in exposed sites, and percentage oftotal seed rain for all species collected in 180 seed traps in KNAR from April 2003 to March 2004 24

Table 2.3. Mean species richness per trap offleshy-fruited seed rain under trees and in exposed sites for six dry forest trees 25

Table 2.4. Mean density ofseeds dispersed by birds and fleshy-fruited seed rain under trees and in associated exposed sites for six dry forest trees in KNAR. 27

Table 2.5. Mean density ofseeds dispersed by birds for 11 species under six dry forest trees in KNAR 28

Table 2.6. Relative frequency offleshy-fruited species in the seed rain under six dry forest trees and all exposed seed traps at KNAR 29

Table 2.7. Frequency ofseedlings offleshy-fruited species under six dry forest trees at KNAR 33

Table 2.8. Mean visits per hour and mean fruits consumed per foraging visit by seven non-native bird species to five dry forest trees at KNAR. 35

Table 2.9. Mean seed length and width of 14 seeds collected in the seed rain at KNAR...... 36

Table 3.1. Means ofseed length and width and mean removal rate under trees and in exposed sites for species used in the seed predation trials 55

Table 3.2. General linear models for number ofseeds and/or fruits remaining versus block, treatment, location, and the interaction oftreatment and location after 15 days for four dry forest trees in KNAR 56

Table 3.3. Two-way ANOVA results for the effects oftreatment and location on removal rates for Pleomele auwahiensis 60

Table A.I. Tree dimensions ofstudy species used in sampling offleshy-fruited seeds and seed removal trials 73

Table B.1. Monthly percentage offleshy-fruited trees and shrubs with mature fruit...... 76

VI Table B.2. Mean monthly percentage ofstems with immature fruits by tree species...... 76

Table B.3. Mean monthly percentage ofstems with mature fruits by tree species 77

Table B.4. Mean monthly number ofimmature fruits and mature fruits on branches by shrub species 77

Table C.l. Relative frequency and abundance ofbird species recorded at 10 sampling stations over a l2-month period in KNAR. 83

Table D.l. Abundance ofground vegetation at KNAR study site 88

Table D.l (Continued) Abundance ofground vegetation at KNAR study site 89

Table D.2. Abundance ofcanopy vegetation at KNAR study site 89

Table D.3. Abundance ofnative and non-native vegetation, dead trees, bare ground, and open sky at KNAR study site 90

Vll LIST OF FIGURES

Figure 1.1. Remnant dry forests ofthe Hawaiian Islands 3

Figure 2.1. Study area within Kanaio Natural Area Reserve 16

Figure 2.2. Mean density ofseeds dispersed by birds under six dry forest tree species... 30

Figure 2.3. Mean density ofconspecific seeds under five dry forest trees at KNAR...... 31

Figure 2.4. Mean density ofBocconia seeds versus conspecific seeds under four dry forest trees at KNAR 32

Figure 3.1. Removal rates ofseeds and fruits from the KNAR 57

Figure 3.2. Removal rates for Bocconiafrutescens and Diospyros sandwicensis seeds and fruits under trees and in exposed sites 58

Figure 3.3. Removal rates for Pleomele auwahiensis seeds and fruits under trees and in exposed sites 59

Figure 3.4. Removal rates for Pleomele auwahiensis seeds and fruits on the open ground under trees and in exposed sites 60

Figure 3.5. Removal rates for Reynoldsia sandwicensis and Santalum ellipticum seeds under trees and in exposed sites 61

Figure A.1. Tree dimensions for study species and used in sampling offleshy-fruited seeds 74

Figure B.1 Monthly percentage offleshy-fruited trees with mature fruit.. 78

Figure B.2. Monthly percentage offleshy-fruited shrubs with mature fruit.. 79

Figure B.3. Mean monthly percentage ofstems with immature and mature fruits by tree species 80

Figure B.4. Mean monthly number ofimmature and mature fruits on branches by shrub species 81

Figure C.1. Relative frequency ofbird species recorded at 10 sampling stations over a 12- month period in KNAR 84

V111 Figure C.2. Relative abundance ofbird species recorded at 10 sampling stations over a 12-month period in KNAR 85

Figure C.3. Mean number ofthree important non-native frugivores at 10 sampling stations over a 12-month period in KNAR. 86

IX CHAPTER 1. INTRODUCTION

BACKGROUND

Tropical dry forests are among the most diverse, yet imperiled, natural communities worldwide (Janzen 1988; Lerdau et al. 1991), and those ofthe Hawaiian

Islands are no exception (Loope 1998). Tropical dry forests, which are found from

Mexico to South America, Africa, Australia, Southeast Asia, and many ofthe world's tropical islands, are characterized by a mean annual rainfall of250 to 2000 mm, mean annual temperatures higher than 17°C, and one to two dry periods per year (Murphy and

Lugo 1986). Unlike other tropical dry forests, which generally contain fewer tree species than neighboring rain forest does (Murphy and Lugo 1986; Janzen 1988), Hawaiian dry forests are richer in tree diversity than comparable areas ofwet forest (Rock 1913;

Carlquist 1980; Sohmer and Gustafson 1987). Like their global counterparts, which have experienced a rapid and significant loss ofarea throughout the world (Murphy and Lugo

1986; Janzen 1988; Bullock et at. 1995), these unique Hawaiian communities have now been reduced to approximately 10% oftheir former extent (Bruegmann 1996; Mehrhoff

1998). Extensive impacts on and alterations ofthese Hawaiian ecosystems began with the agricultural and hunting practices ofthe early Polynesians, their use offire for land clearing, and the introduction ofnon-native animals such as the Polynesian rat, (Rattus exutans) (Kirch 1982; Sadler 1999; Burney et at. 2001; Athens et at. 2002). This deterioration and loss accelerated after the arrival ofEuropeans. As has occurred in dry forests throughout the world (Murphy and Lugo 1986; Janzen 1988), the introduction of ungulates such as cattle and goats, further land clearing for agriculture and development, 1 accidental and intentional anthropogenic fires, and the introduction ofaggressive weeds,

a particularly fire-carrying grasses such as fountain grass (Pennisetum setaceum ), all contributed to the rapid decline ofthe Hawaiian dry forests (Stone 1989; Cuddihy and

Stone 1990; Loope 1998).

Despite these losses, however, biologically diverse areas ofdry forest still remain on several ofthe Hawaiian Islands. Notable remnant forests can be found at Pu'u Wa'a

Wa'a on the island ofHawai'i, on the southern slopes ofMaui at Auwahi, Kanaio and

Pu'u-o-kali, on Lanai at Kanepu'u, on Molokai at Kauhako Crater, and at Mokuleia on

Oahu (Figure 1.1; Medeiros et al. 1984; Medeiros et al. 1986; Cuddihy 1989; Medeiros et al.1996; Loope 1998; Mueller-Dombois and Fosberg 1998). Nevertheless, all ofthe factors that contributed to the initial loss still threaten the integrity ofthe remaining dry forests, and other threats continue to impact the health and functioning ofthese ecosystems (Stone et al. 1992; Pratt and Gon 1998; Blackmore and Vitousek 2000;

Staples and Cowie 2001). One particularly troubling phenomenon that seriously affects the persistence as well as the potential future restoration ofthese communities is the almost complete lack ofnatural reproduction ofnative tree species (Medeiros et al. 1986;

Medeiros et al. 1993; Loope et a1. 1995; Loope 1998; Cabin et al. 2000). Janzen (1988) refers to adult trees that fail to reproduce as "the living dead" and indicates that this problem is symptomatic oftropical dry forests worldwide.

a Plant taxonomy follows Wagner et al. (1999) 2 o Kauhako Crater MOkuleia~ ~ Pu'u 0 Kali Kanepu' u 'L> Au\vahi _~f"S"':;:::"__ Kanaio

Kaupulehu

PU'U Wa'a Wa'a

100 0 100 200 Kilometers ~ EI==:::::3====:E1======31======:51 N

Figure 1.1. Remnant dry forests ofthe Hawaiian Islands

In addition to browsing and trampling ofseedlings by feral ungulates, and competition with aggressive, non-native weeds, several other factors may contribute to the lack ofseedling recruitment in tropical dry forests. These include the loss ofnative pollinators with subsequent reduction or loss ofoutcrossing, the impacts ofnon-native invertebrates and pathogens on seed and seedling survival, modifications in microclimate and microhabitats suitable for seed gennination and seedling survival, loss ofnative birds that scarified and dispersed seeds, and seed predation by introduced rodents (Janzen

1988; Medeiros et al. 1993; Aide et al. 2000; Cabin et al. 2000; Holl et al. 2000;

Zimmennan et al. 2000). The loss ofnative dispersal agents and the likely modification ofthe existing seed shadows, or seed distribution patterns around the source, as well as 3 the effects ofrodent predation on seeds, are two factors that could be influencing community composition and structure ofdry forests throughout the Hawaiian Islands.

Loss o/Native Seed Dispersers

Direct and indirect impacts ofboth Polynesians and Europeans have led to the widespread extinction ofnative birds throughout Polynesia, including the Hawaiian

Islands (Kirch 1982; James and Olson 1991; Olson and James 1982, 1991; Steadman

1995; Burney et at. 2001; Ziegler 2002). What roles these birds may have played in the scarification and dispersal ofnative tree seeds is unknown, but others have suggested that a loss ofnative dispersers and replacement by more generalist, non-native dispersal agents could affect both the shape ofthe seed shadow and the types ofseeds being dispersed (Temple 1977; Cole et at. 1995; Willson and Traveset 2001; Loiselle and Blake

2002; Meehan et at. 2002). As plants that depend on animals for seed transport are susceptible to dispersal failure when their seed vectors become rare or extinct, disruptions to the mutualism can have serious consequences for the maintenance ofplant populations

(Willson and Traveset 2001). This may be particularly true ofHawaiian dry forest areas, in which almost two-thirds ofnative tree and shrub species have fleshy fruits presumably adapted for some type ofbird dispersal (Medeiros et at. 1993). These areas once supported a diverse and conspicuous native avifauna prior to the arrival ofthe

Polynesians (James and Olson 1991; Olson and James 1991). Other than the native owl orpueo (Asio jlammeus sandwicensis), the Pacific Golden plover or kotea (ptuvialis futva) and the infrequently seen Hawaiian goose or Nene (Branta sandvicensis), recent

4 surveys within Kanaio recorded only introduced game birds and small, generalist frugivores such as the Japanese white-eye (Zosterops japonicus) and the common myna

(Acridotheres tristis) (Medeiros et al. 1993; Chimera pers. obs.). This replacement of native with non-native birds is common to the lowland dry forests ofthe Hawaiian

Islands (Pratt et al. 1987). Although little is known about how the loss or gain ofa dispersal agent alters a plant's seed shadow, potential consequences could include changes in the amounts and sizes ofseeds being dispersed (Willson and Traveset 2001).

The size and structure ofthe fruits and seeds that birds can consume or transport are constrained by the size ofthe bird (Wheelwright 1985; Stiles 2001), such that small, alien birds may be preferentially consuming only smaller fruits and seeds. Anecdotal observations ofseedling distributions suggest that birds are dispersing some small-seeded native and non-native taxa, ofwhich several ofthe latter are serious, habitat-modifying weeds. It is unknown to what degree these introduced birds act as surrogates ofthe extinct avifauna in the dispersal ofseeds, how frequently the seeds ofthese and other species, ifany, are being dispersed, and whether or not there is a size limit to seeds that may be dispersed.

Rodents andSeed Predation

The flora and fauna ofPacific island ecosystems, including the Hawaiian Islands, evolved in the absence ofnative rodents, and presumably lack defenses against them.

Introduced rodents such as the Polynesian rat (Rattus exulans), the black or roofrat (R. rattus), the Norway rat (R. norvegicus) and the house mouse (Mus musculus) have been

5 inadvertently introduced to many island chains throughout Polynesia, with devastating impacts on the insular biota (Tomich 1986). In the Hawaiian Islands, rats have negatively impacted the native avifauna (Atkinson 1977; Scott et al. 1986), the endemic snail and arthropod fauna (Hadfield et al. 1993; Cole et al. 2000) and the flora, including many rare and endangered species (Wirtz 1972; Baker and Allen 1978; Russel 1980; Scowcroft and Sakai 1984; Stone et al. 1984; Stone 1985; Sugihara 1997; Cole et al. 2000).

Although introduced rodents are notorious plant and seed predators in other insular ecosystems (Fall et al. 1971; Clark 1982; Campbell et al. 1984; Allen et al. 1994; Moles and Drake 1999; Williams et al. 2000; Campbell and Atkinson 2002; Delgado Garcia

2002; McConkey et al. 2003), and have been anecdotally implicated in the destruction of seeds in dry forest ecosystems, including the Kanaio Natural Area Reserve (Medeiros et al. 1986; Medeiros et al. 1993; Cabin et al. 2000), no studies have been published that quantitatively document the impacts ofthese introduced rodents on the seeds ofparticular dry forest taxa. As many ofthese species produce fruits with one to few larger seeds

(Wagner et al. 1999), and as rodents in general have been shown to be voracious predators ofintermediate and larger seeds (Stiles 2001), the effects that introduced rodents have on seed mortality and subsequent levels ofseedling recruitment could be profound (Crawley 2001).

Influence ofVegetation on Seed Dispersal andPredation

A modified seed dispersal regimen due to losses ofnative dispersal agents would be expected over time to influence structure and composition ofdry forest communities.

6 A scattered distribution ofdry forest trees would, in tum, play an influential role in the deposition ofseeds throughout an area. Remnant Hawaiian dry forests like that in

Kanaio, Maui, typically consist ofisolated trees and scattered groves oftrees surrounded by native shrublands, non-native grasslands or barren lava flows (Medeiros et aZ. 1993;

Mueller-Dombois and Fosberg 1998). The highest densities in a seed shadow normally occur close to the seed source (Willson and Traveset 2001). In the case offruiting trees, greatest seed densities would therefore be expected under tree crowns. In addition, directed dispersal often results in non-random seed shadows due to the preferential movements ofbirds to these feeding and perch sites (Howe and Primack 1975; Jordano and Schupp 2000; Wenny 2001; Clark et aZ. 2004). By providing suitable microsites for the growth ofseedlings, isolated trees or groves oftrees in otherwise open areas often act as recruitment foci for bird-disseminated species (Yarranton and Morrison 1974;

Debussche et aZ. 1982; McDonnell and Stiles 1983; Izhaki et aZ. 1991; Guevara and

Laborde 1993; Debussche and Isenmann 1994; Ferguson and Drake 1999). Isolated trees may therefore play an important role in the succession and regeneration offormerly disturbed or degraded areas (Uhl et aZ. 1982; Guevara et aZ. 1986; Eriksson and Ehrlen

1992; Loiselle and Blake 1993; Martinez-Ramos and Soto-Castro 1993; UhI1998;

Wijdeven and Kuzee 2000; Cordell et aZ. 2002).

Despite the potential benefits, however, this non-random deposition under trees, combined with the higher seed rain predicted by the seed shadow under parent trees, could also influence seed predation rates (Hulme 1998). Several authors have documented higher predation rates under trees and vegetation, as opposed to more open areas (Howe et aZ. 1985; Callaway 1992; Aide and Cave1ier 1994; Osunkoya 1994;

7 Chapman and Chapman 1996; Hau 1997; Hulme 1997; Wenny 2000; Ho1l2002). Others have found no difference (DeSteven and Putz 1984; Terborgh et al. 1993), or higher rates ofpredation in exposed areas or grasslands (Hay and Fuller 1981; Uhl 1998; Wijdeven and Kuzee 2000). One study reported lower predation rates under isolated pasture trees as opposed to either open pastures or intact forests (Holl and Lulow 1997).

To better predict how the interaction between the scattered distribution ofdry forest trees, seed dispersal and seed predation may shape the dry forest community composition ofthe Kanaio Natural Area Reserve, I addressed the following hypotheses as part ofmy Master's Thesis:

Hypotheses

Seed Dispersalb

1) There is no difference in the seed rain offleshy-fruited, bird-dispersed seeds

under trees versus in exposed areas.

2) Non-native birds such as the Japanese white-eye (Zosterops japonicus), the

common mynah (Acridotheres tristis), the zebra dove (Geopelia striata), the

northern cardinal (Cardinalis cardinalis) and the northern mockingbird (Mimus

polyglottos) are not dispersing the seeds offleshy-fruited native and non-native

trees.

3) There is no relationship between seed size and quantity ofseeds dispersed for

fleshy-fruited species.

Seed Predation

b The term "seed" is used broadly in this proposal to refer to the persistent part ofthe diaspore. Thus, for Reynoldsia sandwicensis, the units studied will be the multiple pyrenes within the drupe. 8 4) Rodents are not removing and destroying seeds ofdry forest species.

5) There is no difference in removal rates ofseeds under trees versus exposed sites.

6) There is no relationship between seed size and removal rates for dry forest tree

specIes.

9 CHAPTER 2. PATTERNS OF SEED DISPERSAL: VEGETATION

STRUCTURE, FRUGIVORY AND SIZE EFFECTS

INTRODUCTION

Tropical dry forests, including those ofthe Hawaiian Islands, are among the world's most diverse, yet imperiled, natural communities (Janzen 1988; Lerdau et al.

1991; Loope 1998). Unlike other tropical dry forests, with generally fewer species than neighboring rain forest (Murphy and Lugo 1986; Janzen 1988), Hawaiian dry forests have greater tree diversity than comparable areas ofwet forest (Rock 1913; Carlquist

1980; Sohmer and Gustafson 1987). Like their global counterparts, which have experienced a rapid and significant loss in area (Murphy and Lugo 1986; Janzen 1988;

Bullock et al. 1995), these unique Hawaiian communities have now been reduced to approximately 10% oftheir former cover (Bruegmann 1996; Mehrhoff 1998). Extensive alterations ofthese ecosystems began with the agricultural and hunting practices ofthe early Polynesians, and with the introduction ofnon-native animals such as the Polynesian rat, (Rattus exulans) (Kirch 1982; Sadler 1999; Burney et al. 2001; Athens et al. 2002).

This deterioration and loss accelerated after the arrival ofEuropeans. Hawaiian dry forests rapidly declined following the introduction ofungulates such as cattle (Bas taurus) and goats (Capra hircus), land clearing for agriculture and development, accidental and intentional anthropogenic fires, and the introduction ofaggressive weeds, factors commonly attributed to the loss ofdry forests throughout the world (Murphy and

Lugo 1986; Janzen 1988; Stone 1989; Cuddihy and Stone 1990; Loope 1998).

10 Despite these losses, however, biologically diverse, iffragmented, areas ofdry forest still remain on several ofthe Hawaiian Islands (Medeiros et al. 1984; Medeiros et al. 1986; Loope 1998; Mueller-Dombois and Fosberg 1998). Nevertheless, all ofthe factors that contributed to the initial loss still threaten the integrity ofthe remaining dry forests, and other threats continue to impact the health and functioning ofthese ecosystems (Stone et al. 1992; Pratt and Gon 1998; Blackmore and Vitousek 2000;

Staples and Cowie 2001). One particularly troubling phenomenon that seriously affects the persistence as well as the potential future restoration ofthese communities is the almost complete lack ofnatural reproduction ofnative tree species (Medeiros et al. 1986;

Medeiros et al. 1993; Loope et al. 1995; Loope 1998; Cabin et al. 2000). Janzen (1988) refers to adult trees that fail to reproduce as "the living dead" and indicates that this problem is symptomatic oftropical dry forests worldwide.

In addition to browsing and trampling ofseedlings by feral ungulates, and competition with non-native weeds, several other factors may contribute to the lack of seedling recruitment in tropical dry forests. These include the loss ofnative pollinators, the impacts ofnon-native invertebrates and pathogens on seed and seedling survival, modifications in microclimate and microhabitats suitable for seed germination and seedling survival, loss ofnative birds that scarified and dispersed seeds, and seed predation by introduced rodents (Janzen 1988; Medeiros et al. 1993; Aide et al. 2000;

Cabin et al. 2000; Holl et al. 2000; Zimmerman et al. 2000). The loss ofnative dispersal agents and the likely modification ofthe existing seed shadows are factors that could be influencing community composition and structure ofHawaiian dry forests.

11 Loss o/Native Seed Dispersers

Direct and indirect impacts ofboth Polynesians and Europeans have led to the widespread extinction ofnative birds throughout Polynesia, including the Hawaiian

Islands (Kirch 1982; James and Olson 1991; Olson and James 1982, 1991; Steadman

1995; Burney et al. 2001; Ziegler 2002). What roles these birds may have played in the scarification, dispersal and predation ofnative tree seeds is unknown, but others have suggested that a loss ofnative dispersers and replacement by more generalist, non-native dispersal agents could affect both the shape ofthe seed shadow and the types ofseeds being dispersed (Temple 1977; Cox et al. 1991; Cole et al. 1995; Loiselle and Blake

2002; Meehan et al. 2002; McConkey and Drake 2002). Plants that depend on animals for seed transport are susceptible to dispersal failure when their vectors become rare or extinct, and disruptions to the mutualism can have serious consequences for the maintenance ofplant populations (Willson and Traveset 2001). This may be particularly true ofKanaio and adjacent dry forest areas on Maui, in which over 59% ofnative trees and shrubs have fleshy fruits presumably adapted for some type ofbird dispersal

(Medeiros et al. 1993). These areas once supported a diverse and conspicuous native avifauna prior to the arrival ofthe Polynesians (James and Olson 1991; Olson and James

1991) but are now almost entirely dominated by introduced game birds and non-native passerines such as the Japanese white-eye (Zosteropsjaponicus) (Medeiros et al. 1993;

Chimera pers. obs.). This replacement ofnative with non-native birds is common in the lowland dry forests ofthe Hawaiian Islands (Pratt et al. 1987). Little is known about how the loss or addition ofa dispersal agent alters the seed shadow ofa plant, but potential consequences could include changes in the amounts and sizes ofseeds being dispersed

12 (Willson and Traveset 2001). The size and structute ofthe fruits and seeds that birds can consume or transport are constrained by the size ofthe bird (Wheelwright 1985; Stiles

2001), such that small, non-native birds may be preferentially consuming only smaller fruits and seeds. Yet it is unknown to what degree these introduced birds act as surrogates ofthe extinct avifauna in the dispersal ofseeds, how frequently the seeds ofthese and other species, ifany, are being dispersed, and whether or not there is a size limit to seeds that may be dispersed.

Influence ofVegetation on Seed Dispersal

A modified seed dispersal regimen due to losses ofnative dispersal agents would be expected to influence structure and composition ofdry forest communities. The spatial distribution ofdry forest trees would, in turn, play an influential role in the deposition of seeds throughout an area. Remnant Hawaiian dry forests like that in Kanaio, Maui, typically consist ofisolated trees and scattered groves oftrees surrounded by native shrublands, non-native grasslands or barren lava flows (Medeiros et al. 1993; Mueller­

Dombois and Fosberg 1998). The highest densities in a seed shadow normally occur close to the seed source (Willson and Traveset 2001). In the case offruiting trees, greatest seed densities would therefore be expected under tree crowns. In addition, directed dispersal often results in non-random seed shadows due to the preferential movements ofbirds to these feeding and perch sites (Howe and Primack 1975; Jordano and Schupp 2000; Wenny 2001; Clark et al. 2004). By providing suitable microsites for the growth ofseedlings, isolated trees or groves oftrees in otherwise open areas often act

13 as recruitment foci for bird-disseminated species (Yarranton and Morrison 1974;

Debussche et al. 1982; McDonnell and Stiles 1983; Izhaki et al. 1991; Guevara and

Laborde 1993; Debussche and Isenmann 1994; Ferguson and Drake 1999). Isolated trees may therefore play an important role in the succession and regeneration offormerly disturbed or degraded areas (UbI et al. 1982; Guevara et al. 1986; Eriksson and Ehrlen

1992; Loiselle and Blake 1993; Martinez-Ramos and Soto-Castro 1993; Uh11998;

Wijdeven and Kuzee 2000; Cordell et al. 2002).

This study attempts to predict how the interaction between the distribution ofdry forest trees and seed dispersal patterns by non-native frugivores may shape the dry forest community composition ofKNAR by addressing the following questions:

1) Is there a difference in the seed rain offleshy-fruited, bird-dispersed seeds

under trees versus in exposed areas?

2) Are non-native birds acting as surrogates ofthe native avifauna and dispersing

the seeds offleshy-fruited native trees, or are they primarily dispersing seeds

ofnon-native plants?

3) Is there a relationship between seed size and amount ofseeds being dispersed

by the non-native birds?

14 METHODS

Study Site

This study was conducted in the Kanaio Natural Area Reserve (KNAR), on the leeward side ofEast Maui, Hawai'i, at an elevation between 750-850 meters (200 36'N,

156°20'W; Figure 2.1). The climate in the reserve is hot and dry, with annual temperatures between 20-30° C and with mean annual rainfall ofapproximately 750 mm, mostly falling between the months ofOctober to April (Giambelluca et al. 1986). The substrate is estimated to be less than 10,000 years old and consists mostly of'a'a lava with some accumulation ofoverlying soils patchily distributed throughout the area

(Crandell 1983). Medeiros et al. (1986; 1993) have provided a thorough description of the community composition ofthe dry forest ofKNAR and the leeward slopes of

Haleakala, Maui. Supplemental characterization ofthe vegetation was provided using a modified point-intercept method (Bonham 1989; Appendix D). The 354-hectare reserve was established in 1990 to protect exemplary representatives ofHawaiian dry forest species and plant communities, including Dodonaea ('A 'ali'i) lowland shrublands,

Diospyros (Lama) forest and Erythrina (Wiliwili) forests (Gagne and Cuddihy 1990). The reserve also contains individuals and groves ofat least twenty-two relatively short­ statured native tree species, with typical tree heights ranging between three to ten meters.

Anecdotal observations suggest that only eight ofthese species may currently be reproducing (Medeiros et al. 1993). Certain non-native trees, however, including the widespread Bocconiafrutescens (Papaveraceae), are able to reproduce without the apparent difficulties experienced by native species (Chimera pers. obs.).

15 c=J Kanaio NAR Boundary /\ / Study Area Boundary .' .... Contour Interval (100 feet)

Figure 2.1. Study area within Kanaio Natural Area Reserve (200 36'N, l56~O'W).

Measurement ofSeed Rain Under Trees Versus in ExposedAreas

To quantify the difference in seed rain under trees and in exposed areas, one seed trap was placed under each study tree and one in an exposed site at least five meters from

the edge ofeach study tree's crown. This distance was sufficient to prevent fruits or seeds from falling offthe parent trees directly into seed traps in exposed sites, as all study trees were relatively short-statured (Appendix A). Traps under trees were placed in a random

16 compass direction approximately halfthe distance from the trunk to the crown edge. Seed traps in the exposed areas were placed in a random compass direction away from the parent tree, such that the position ofthe seed trap was not within five meters ofany other tree. These exposed sites consisted ofbarren rock, low-statured grasses, woody shrubs and annual or perennial herbaceous species generally less than 50 cm tall. Seed traps in exposed sites were not placed directly under woody shrubs, although they were placed next to shrubs when encountered. Fifteen trees ofeach offour different native dry forest species, ofone non-native species and fifteen dead, standing trees were selected from the northeastern portion ofKNAR, the area ofgreatest tree diversity and abundance

(Medeiros et al. 1993). For purposes ofcomparison, the 15 dead, standing trees were arbitrarily considered a sixth study "species". Trees were selected from those that were not directly adjacent to a conspecific individual. Species ofnative trees were also selected on the basis ofavailability ofenough individuals meeting the spacing criterion and having fruits with adaptations for bird dispersal (Jordano 2001; Stiles 2001). Native species were selected to represent a range offruit and seed sizes. Trees were selected along 12 parallel, east-west oriented transects, ranging in length from 100 to 675 meters and spaced 50 meters apart. Transects stopped within 5 m ofnon-forest vegetation on adjacent ranchlands, thus accounting for differences in transect length. All appropriate study trees were first recorded along the length ofeach transect. Study trees were randomly selected from this subset ofindividuals such that no trees were closer than 25 meters to each other. The entire study site encompassed an area ofapproximately 0.25

2 km • Native fleshy-fruited taxa used for this study include Reynoldsia sandwicensis,

(Araliaceae), Diospyros sandwicensis (Ebenaceae), Santalum ellipticum (Santalaceae),

17 and Pleomele auwahiensis (Agavaceae). For the dioecious D. sandwicensis, only female, fruit-producing trees were used. One non-native tree, Bocconiafrutescens

(Papaveraceae), was also used. This tree is common in the reserve, and its spread is likely being facilitated by non-native birds attracted to the red, pulpy aril attached to the shiny black seeds (Wagner et al. 1999). The fifteen dead standing trees were selected so that they were roughly similar in branch architecture and height. Basal diameter, height, and crown diameter were recorded for each study tree (Appendix A). Seed vouchers were collected from all fruiting trees in the vicinity during the duration ofthe study to aid in the identification ofseeds in the traps.

Seed traps were constructed ofpairs ofcircular plastic pots with the bottoms removed. One pot was placed inside the other with a cotton cloth in between, a modification ofthe design described by Drake (1998). Traps were 23 cm deep with a

25.4 cm diameter, such that the total area sampled under each ofthe six tree species (15 traps/species) and in the adjacent open area (15 traps/species) was 0.76 m2 each. Trap tops were covered with wire screens with 4 x 3 x 2.5 cm hexagonal apertures, large enough to allow fruits and seeds to fall through and small enough to deter rodents or game birds from consuming seeds.

Traps were emptied and the liners replaced every month, for a twelve-month period (April 2003 to March 2004). Numbers ofseeds in traps were counted by species and the number ofintact fruits from the parent tree was also recorded. Because fruits of

Diospyros sandwicensis, Pleomele auwahiensis and Reynoldsia sandwicensis potentially contained multiple seeds, 50 ofeach haphazardly-selected fruit were cut open and the seeds inside counted. Means were 1.54 ± 0.11 seeds per D. sandwicensis fruit, 1.3 ± 0.09

18 seeds per P. auwahiensis fruit and 9.08 ± 0.23 seeds per R. sandwicensis fruit. Any fruits from these species found in traps were therefore multiplied by the mean. In addition, seeds were investigated for any invertebrate or rodent damage. The presence ofimmature and ripe fruits on sample trees and common taxa in the area was recorded on a monthly basis from March 2003 to February 2004 to document part ofthe potential pool ofbird­ dispersed seeds that could be disseminated in the area (Appendix B).

Seed Dispersal Analysis

All seeds collected in traps were counted and categorized by one offour general dispersal vectors: bird-dispersed; gravity-dispersed; uncertain; wind-dispersed. Fleshy­ fruited seeds collected in traps under trees were quantified as "bird-dispersed" ifthey were from species other than that ofthe overhanging tree. All fleshy-fruited seeds collected in traps in exposed areas or under dead, standing trees were counted as "bird­ dispersed". Seeds from intact fruits that had fallen directly from the parent tree into seed traps were categorized as "gravity-dispersed". Seeds from one weedy annual, Bidens pi/osa (Asteraceae), were present in many traps and were categorized as "gravity­ dispersed" because they often fell from plants overtopping seed traps. Seeds lacking pulp but captured beneath a conspecific tree were counted as "uncertain". This category accounts for the possibilities that birds may have dropped seeds from the parent tree or that invertebrates consumed pulp from fruits after they fell from parent trees. Smaller, wind-borne seeds and those with adaptations for dispersal by wind currents (e.g. tufts of hairs, winged-fruits etc.) were classified as "wind-dispersed".

19 As several seeds without pilip were collected in traps under conspecifc trees, a large number ofseeds were categorized as having an "uncertain" dispersal vector. These data were combined with "bird-dispersed" data, and fleshy-fruited "gravity-dispersed" data into a single category: all fleshy-fruited seeds, to indicate that all seeds in this category possess adaptations for bird dispersal. Fleshy-fruited seed rain was then compared between trees and exposed areas for all species. Total seed rain numbers were

2 converted into seed densities (seeds m- ) and square root transformed to improve normality. Bird-dispersed and fleshy-fruited seed densities were compared between trees and associated exposed sites with paired t-tests. Species richness offleshy-fruited seeds was also compared between trees and exposed sites with paired t-tests. (Zar 1999).

Fleshy-fruited seed rain ofnative and non-native species was compared under trees of each species using paired t-tests. Fleshy-fruited seed rain ofparticiliar seed species (see

Results) was compared between different tree species using a one-way analysis of variance, with Tukey tests determining which densities were significantly different (Zar

1999).

As an indirect estimate ofseed dispersal, the presence/absence ofall species of fleshy-fruited seedlings less than 10 cm tall was noted under the crowns ofall study trees.

For presentation and discussion ofresults, dead trees are treated as a "species".

Bird Observations

To determine what birds, ifany, were consuming and potentially dispersing the seeds ofthe study trees, 30-40 hours oftimed watches per study tree species were

20 conducted to coincide with fruiting. A different haphazardly-selected fruiting tree was observed each day for a total ofseven to ten days. Only trees estimated to have mature fruit on at least 25 percent oftree branches were used. Observations took place in the morning, from sunrise until as late as 1130 a.m., and only occurred on days when climatological conditions were favorable (i. e. absence ofheavy rains or extremely strong winds). Observations were made with binoculars and a spotting scope at a distance and in a position that was unlikely to affect bird activity. Bird species, foraging visits per hour and fruit consumed per visit were recorded for each study tree. Observations were conducted on Pleomele auwahiensis for eight days in May and June 2003, on Bocconia frutescens for ten days in May and June 2003, on Diospyros sandwicensis for six days in

December 2003 and January 2004, on Reynoldsia sandwicensis for nine days in January

2004, and on Santalum ellipticum for seven days in February 2004. Anecdotal observations offrugivory on non-study species were also noted. Because observations were conducted on different numbers ofdays and under different weather conditions, bird visitation data were not compared statistically between species.

To assess the relative frequency and abundance ofbird species at the study site, additional sampling was conducted from April 2003 to March 2004 at 10 stations along a bird transect roughly bisecting the seed sampling transects. For more information on methods used, and for results, consult Appendix C.

Seed Size Measurements

21 To determine the range ofsizes ofdispersed seeds, length and width were recorded for those seeds collected in seed traps and classified as "bird-dispersed" (see

Seed Dispersal Analysis section). Spearman's rank correlation was used to test for relationships between seed length, width, and total numbers ofbird-dispersed seeds collected in all seed traps. Although no seeds ofPleomele auwahiensis, Santalum ellipticum, or Reynoldsia sandwicensis were recorded as "bird-dispersed", correlations were run both with and without size dimensions ofthese species as they are common to the area and were specific subjects ofthe seed dispersal analysis.

Characterization ofthe vegetation

Plant cover in the study site was sampled at 1000 points using the point-intercept method (Bonham 1989). For a more detailed description of methods used, and results, consult Appendix D.

RESULTS

Seed Rain

A total of 12,769 seeds (Table 2.1) from at least 32 different species (Table 2.2) were found in the 180 traps under trees and in exposed sites. Over 90% ofseeds were non-native. Ofthese, two wind-dispersed grasses, Melinis repens and M minutiflora, dominated the seed rain, comprising 55.6% and 10.6% ofthe total density respectively

(Table 2.2).

22 Table 2.1. Mean density (seeds m-2 ± 1 SEM) by dispersal vector and location (tree vs. exposed sites) at KNAR from April 2003 to March 2004. Densities under trees versus exposed sites were compared with paired t-tests on square root transformed data. Seed densities were not compared between different dispersal vectors. Dispersal Vector Tree Exposed P-value % Total Wind 595.6 (96.5) 1362.4 (283.6) 0.167 70 Gravity 321.0 (80.1) 42.5 (17.7) <0.001 13.0 Birds 237.9 (30.2) 7.7 (2.97) <0.001 8.8 Uncertain 230.5 (68.0) 1.3 (0.75) <0.001 8.3

23 2 Table 2.2. Mean density (seeds m- ) under trees and in exposed sites (Exp.), and percentage oftotal seed rain for all species collected in 180 seed traps in KNAR from April 2003 to March 2004. N = native; A = alien/non-native. Seed Category Family Status Tree Exp. % Total Melinis repens Poaceae A 373.7 1181.7 55.55 Bocconiafrutescens Papaveraceae A 473.0 6.1 17.11 Melinis minutiflora Poaceae A 162.3 133.1 10.55 Bidens pilosa Asteraceae A 78.3 41.0 4.26 Reynoldsia sandwicensis Araliaceae N 96.0 0.0 3.43 Dodonaea viscosa Sapindaceae N 33.6 26.8 2.14 Diospyros sandwicensis Ebenaceae N 51.8 0.0 1.85 Lantana camara Verbenaceae A 47.6 3.5 1.82 Santalum ellipticum Santalaceae N 16.2 0.0 0.58 Chenopodium oahuense Chenopodiaceae N 11.4 0.0 0.41 Galinsoga parviflora Asteraceae A 1.3 8.8 0.36 Pleomele auwahiensis Agavaceae N 9.2 0.0 0.33 Nothocestrum latifolium Solanaceae N 7.2 0.0 0.26 Sonchus oleraceus Asteraceae A 2.2 4.0 0.22 Osteomeles anthyllidifolia Rosaceae N 5.3 0.0 0.19 Unidentified (one spp.) NA NA 2.9 2.0 0.17 Ageratina adenophora Asteraceae A 1.5 0.9 0.09 Emiliafosbergii Asteraceae A 0.4 2.0 0.09 Wikstroemia monticola Thymelaeaceae N 2.4 0.0 0.09 Asclepias physocarpa Asce1piadaceae A 2.2 0.0 0.08 Sporobolus africanus Poaceae A 0.0 2.2 0.08 Cocculus orbiculatus Menispermaceae N 1.8 0.0 0.06 Digitaria ciliaris Poaceae A 1.8 0.0 0.06 Zinnia peruviana Asteraceae A 0.0 1.8 0.06 Petroselinum crispum Apiaceae A 0.7 0.9 0.05 Leptecophylla tameiameiae Epacridaceae N 1.1 0.0 0.04 Conyza bonariensis Asteraceae A 0.0 0.4 0.02 Schinus terebinthifolius Anacardiaceae A 0.4 0.0 0.02 Alyxia oliviformis Apocynaceae N 0.2 0.0 0.01 Hyochoeris radicata Asteraceae A 0.2 0.0 0.01 Myoporum sandwicense Myoporaceae N 0.2 0.0 0.01 Tridax procumbens Asteraceae A 0.2 0.0 0.01

24 Table 2.3. Mean species richness (± 1 SE) per trap offleshy-fruited seed rain under trees and in exposed sites for six dry forest tree species. All pairwise comparisons between trees and exposed sites are significantly different (paired t-test; P < 0.05).

Tree Species Tree Exposed Bocconiafrutescens 2.07 (0.18) 0.07 (0.07) Dead Trees 2.53 (0.24) 0.13 (0.09) Diospyros sandwicensis 3.40 (0.21) 0.27 (0.12) Pleomele auwahiensis 2.60 (0.40) 0.40 (0.16) Reynoldsia sandwicensis 3.33 (0.33) 0.07 (0.07) Santalum ellipticum 2.27 (0.36) 0.00 (0.00)

Mean species richness offleshy-fruited seeds per trap ranged from 0-6 species under trees and 0-2 species in exposed sites. In all cases, species richness was higher under trees than in exposed sites (Table 2.3). For all study tree species, there were significantly more bird-dispersed seeds collected under trees than in associated exposed areas (Table 2.4). In the fleshy-fruited seed categories, only 11 species were conclusively bird-dispersed. Ofthese, eight were native species and three were non-native (Table 2.5).

Ofthe 1085 bird-dispersed seeds in the seed rain, 96.8% were collected under trees, and

3.2% were collected in exposed sites. Although more species ofnatives were dispersed by birds, total seed rain was dominated by two non-native species, Bocconiafrutescens and Lantana camara, comprising 74.7% and 17.7% ofthe bird-dispersed seed rain respectively. These two species were also the only ones to be bird-dispersed to exposed sites. There were also significant differences in bird-dispersed seed densities among the six study trees (Figure 2.2). In many cases, it was difficult to assess whether conspecific seeds collected under parent trees were dispersed by birds from other conspecific trees, or

25 merely fell into seed traps. Numbers ofbird-dispersed seeds in the seed rain are therefore a conservative estimate.

A total of 14 species with fleshy-fruited seeds were collected in the seed rain, 11 that had been dispersed by birds and three (Pleomele auwahiensis, Reynoldsia sandwicensis and Santalum ellipticum) that had not. Ofthese fleshy-fruited species,

98.7% were collected under trees, and 1.3% were collected in exposed sites. Fleshy­ fruited seed rain was significantly higher under trees versus exposed areas in all cases

(Table 2.4). Bocconiafrutescens comprised 66.6% ofthe fleshy-fruited seed rain, whereas all native species combined comprised only 26.6%. Bocconia seeds were also the most widespread, collected under 93.3% oftrees and 5.6% ofexposed sites (Table

2.6). In contrast, all native species combined were collected under only 61.1 % oftrees and in 0% ofexposed sites (Table 2.6). The second most widespread species in the fleshy-fruited seed rain was Lantana camara, collected under 74.4% ofall trees and in

11.1% ofall exposed sites (Table 2.6). Seeds collected beneath conspecific fruiting trees likely fell directly from the parent tree into seed traps. This conspecific seed rain was highest under Bocconia trees, but did not significantly differ among the four native trees

(Figure 2.3). In comparisons ofBocconia seed rain under native trees versus conspecific native seed rain, Bocconia seed densities did not differ from conspecific seed rain under three native trees, but there were significantly more Bocconia than Santalum seeds under

Santalum ellipticum trees (Figure 2.4).

Seedlings ofonly eight fleshy-fruited species were present under study trees

(Table 2.7). Bocconia seedlings were the most widespread, occurring under 84% ofall trees and present under all six study species. Diospyros seedlings were present under

26 Diospyros trees, but were not recorded under other species. Reynoldsia seedlings were present under 13% ofReynoldsia trees, but were not found under any other species. No

Pleomele or Santalum seedlings were noted under any trees. Lantana camara seedlings were the second most widespread species, occurring under 72% ofall trees. Two native fleshy-fruited shrubs, Osteomeles anthyllidifolia and Wikstroemia monticola, were also relatively widespread, occurring under 22% and 32% ofall trees respectively (Table 2.7).

Table 2.4. Mean density (seeds m-2 ± 1 SE) ofseeds dispersed by birds and fleshy-fruited seed rain under trees and in associated exposed sites for six dry forest tree species in KNAR. All pairwise comparisons between trees and exposed sites are significantly different (paired t-test; P < 0.05).

Bird- dispersed Fleshy-fruited Tree Species Tree Exposed Tree Exposed Bocconiafrutescens 105.3(35.8) 6.58(6.58) 1886.7(494.3) 6.58(6.58) Dead 202.6(48.7) 1.32(1.32) 203.9(48.4) 2.63(1.79) Diospyros sandwicensis 442.1(131.6) 21.05(12.25) 752.6(177.2) 22.37(12.16) Pleomele auwahiensis 165.8(55.9) 15.79(10.58) 227.6(61.9) 23.68(11.59) Reynoldsia sandwicensis 263.1(50.0) 1.32(1.32) 840.7(313.2) 1.32(1.32) Santalum ellipticum 248.7(59.8) 0.00(0.00) 347.3(76.5) 1.32(1.32)

27 z Table 2.5. Mean density (seeds m- ) ofseeds dispersed by birds for 11 species under six dry forest tree species in KNAR. Boc fro: Bocconiafrutescens; Dio san: Diospyros sandwicensis; Pie auw: Pleomele auwahiensis; Rey san: Reynoldsia sandwicensis; San ell: Santalum ellipticum. Non-native species are marked with an asterisk (*). Seed species Boc fro Dead Dio san Pie auw Rey san San ell Alyxia oliviformis 0.0 0.0 1.3 0.0 0.0 0.0 Bocconiafrutescens* 28.9 114.5 388.1 127.6 177.6 227.6 Cocculus orbiculatus 0.0 0.0 4.0 5.3 0.0 1.3 Diospyros sandwicensis 0.0 0.0 0.0 0.0 2.6 0.0 Lantana camara* 50.0 48.7 35.5 23.7 75.0 17.1 Leptecophylla tameiameiae 0.0 0.0 5.3 0.0 1.3 0.0 Myoporum sandwicense 0.0 1.3 0.0 0.0 0.0 0.0 Nothocestrum latifolium 21.1 19.7 0.0 0.0 2.6 0.0 Osteomeles anthyllidifolia 4.0 11.8 5.3 6.6 4.0 0.0 Schinus terebinthifolius* 0.0 0.0 0.0 0.0 0.0 2.6 Wikstroemia monticola 1.3 6.6 2.6 2.6 0.0 0.0

28 Table 2.6. Relative frequency offleshy-fruited species in the seed rain under six dry forest tree species and all exposed (Exp.) seed traps at KNAR. Boc fru = Bocconia frutescens; Dio san = Diospyros sandwicensis; PIe auw = Pleomele auwahiensis; Rey san = Reynoldsia sandwicensis; San ell = Santalum ellipticum. Non-native species are marked with an asterisk (*). Boc Dio PIe Rey San All Dead Exp. Seed Species fru san auw san ell Trees Non-natives 100.0 100.0 100.0 93.3 100.0 86.7 15.6 96.7 Natives 33.3 53.3 100.0 53.3 80.0 46.7 0.0 61.1

Alyxia oliviformis 0.0 0.0 6.7 0.0 0.0 0.0 0.0 1.1 Bocconiafrutescens* 100.0 93.3 100.0 86.7 93.3 86.7 5.6 93.3 Cocculus orbiculatus 0.0 0.0 13.3 6.7 0.0 6.7 0.0 4.4 Diospyros sandwicensis 0.0 0.0 100.0 0.0 13.3 0.0 0.0 18.9 Lantana camara* 73.3 86.7 66.7 73.3 93.3 53.3 11.1 74.4 Leptecophylla tameiameiae 0.0 0.0 20.0 0.0 6.7 0.0 0.0 4.4 Myoporum sandwicense 0.0 6.7 0.0 0.0 0.0 0.0 0.0 1.1 Nothocestrum latifolium 6.7 13.3 0.0 0.0 6.7 0.0 0.0 4.4 Os~omeksanffiyllidifolia 13.3 33.3 20.0 20.0 13.3 0.0 0.0 16.7 Pleomele auwahiensis 0.0 0.0 0.0 46.7 0.0 0.0 0.0 7.8 Reynoldsia sandwicensis 0.0 0.0 0.0 0.0 80.0 0.0 0.0 13.3 Santalum ellipticum 0.0 0.0 0.0 0.0 0.0 46.7 0.0 7.8 Schinus terebinthifolius* 0.0 0.0 0.0 0.0 0.0 13.3 0.0 2.2 Wikstroemia monticola 13.3 20.0 6.7 6.7 0.0 0.0 0.0 7.8

29 700

---. 600 2: w (f) 500 + c m 400 ---E 1: 300 (/) ~ 200 (J) (/) 100 o Boc fru Dead Dio san Pie auw Rey san San ell

Figure 2.2. Mean density ofseeds dispersed by birds under six dry forest tree species at KNAR. Boc fro = Bocconiafrutescens; Dio san = Diospyros sandwicensis; PIe auw = Pleomele auwahiensis; Rey san = Reynoldsia sandwicensis; San ell = Santalum ellipticum. Letters at the base ofbars show the results ofa one-way ANOVA and Tukey test. Trees sharing the same letter are not significantly different (P > 0.05).

30 2500

~ 2000 w (J) + c ro 1500 Q) E N -- 'E 1000 en "'C (J) ~ 500

o Soc fru Dio san Pie auw Rey san San ell

Figure 2.3. Mean density of conspecific seeds under five dry forest tree species at KNAR. Boc fro = Bocconiafrutescens; Dio san = Diospyros sandwicensis; PIe auw = Pleomele auwahiensis; Rey san = Reynoldsia sandwicensis; San ell = Santalum ellipticum. Letters at the base ofbars show the results ofa one-way ANOVA and Tukey test. Trees sharing the same letter are not significantly different (P > 0.05).

31 1000 [ill Bocconia o Conspecific 2 800 w (J) + c 600 «l

A o Diospyros P/eome/e Reyno/dsia Santa/um sandwicensis auwahiensis sandwicensis ellipticum

Figure 2.4. Mean density ofBocconia seeds versus conspecific seeds under four dry forest tree species at KNAR. Letters at the base ofbars show the results ofpaired t-tests for Bocconia seeds and conspecific seeds under each tree species. Densities were square root transformed for analyses. Bars sharing the same letter are not significantly different (P > 0.05). Most conspecific seeds likely fell directly from the parent tree into seed traps.

32 Table 2.7. Relative frequency ofseedlings offleshy-fruited species under six dry forest tree species at KNAR. Boc fru = Bocconiafrutescens; Dio san = Diospyros sandwicensis; PIe auw = Pleomele auwahiensis; Rey san = Reynoldsia sandwicensis; San ell = Santalum ellipticum. Non-native species are marked with an asterisk (*). (n = 15 for individual species; n = 90 for all trees). Seedling Species Bocfru Dead Dio san PIe auw Rey san San ell All Trees Alyxia oliviformis 0.0 13.3 0.0 0.0 0.0 6.7 3.3 Bocconiafrutescens * 93.3 93.3 93.3 53.3 80.0 93.3 84.4 Diospyros sandwicensis 0.0 0.0 20.0 0.0 0.0 0.0 3.3 Lantana camara* 80.0 86.7 86.7 73.3 46.7 60.0 72.2 Leptecophylla tameiameiae 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Myoporum sandwicense 0.0 0.0 0.0 6.7 0.0 0.0 1.1 Nothocestrum latifolium 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Osteomeles anthyllidifolia 13.3 26.7 53.3 20.0 0.0 20.0 22.2 Pleomele auwahiensis 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Psydrax odorata 0.0 0.0 13.3 0.0 0.0 0.0 2.2 Reynoldsia sandwicensis 0.0 0.0 0.0 0.0 13.3 0.0 2.2 Santalum ellipticum 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Schinus terebinthifolius* 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Wikstroemia monticola 6.7 20.0 73.3 33.3 20.0 40.0 32.2

Bird Observations

A total of 162 hours were spent observing bird visits to the five species offleshy- fruited study trees. Birds were observed feeding on the fruits or seeds offour ofthe five species, but no birds were observed feeding on any Pleomele auwahiensis fruits or seeds

(Table 2.8). Bocconiafrutescens had the greatest number ofbird foraging visits per hour and the highest number ofseeds consumed per visit (Table 2.8).

A total ofseven species ofnon-native birds visited the five study tree species during observation hours, but only three were observed feeding on fruits or seeds (Table

2.8). Japanese white-eyes (Zosterops japonicus) were the most frequent visitor to all five tree species, but only fed on the fruits ofthree (Table 2.8). These birds generally 33 swallowed entire Bocconia seeds and arils, but on several occasions were observed passing seeds from one to another. In 35 of39 (89.7%) foraging visits to Reynoldsia fruits, white-eyes only pecked at pulp and dislodged berries from panicles, apparently without swallowing seeds. White-eyes were also only observed pecking at the pulp of

Diospyros fruits, although in one instance a bird carried the fruit a short distance to a nearby tree before consuming the fruit pulp. Northern cardinals (Cardinalis cardinalis) visited all five tree species and swallowed whole fruits and seeds oftwo (Bocconia frutescens and Reynoldsia sandwicensis). On four occasions, cardinals were observed clipping the unripe fruits ofSantalum ellipticum in halfand consuming the seed inside.

They were never observed consuming only pulp ofripe fruits. Northern mockingbirds

(Mimus polyglottos) visited all five tree species and swallowed whole fruits and seeds of two (B. frutescens and R. sandwicensis).

34 Table 2.8. Mean visits per hour and mean fruits consumed (± SE) per foraging visit by seven non-native bird species to five dry forest tree species at KNAR. Boc fru: Bocconia frutescens; Rey san: Reynoldsia sandwicensis; San ell: Santalum ellipticum; Dio san: Diospyros sandwicensis; PIe auw: Pleomele auwahiensis. Bocfru Rey san San ell Dio san PIe auw Observation days 10 9 7 6 8 Observation hours 40 30 30 30 32

Visits per hour by bird spp. Acridotheres tristis 0.00 0.10 0.07 0.00 0.00 Cardinalis cardinalis 0.23 0.17 0.17 0.03 0.03 Carpodacus mexicanus 0.00 0.00 0.03 0.03 0.63 Lonchura punctulata 0.03 0.00 0.07 0.00 0.13 Mimus polyglottos 0.08 0.07 0.03 0.03 0.03 Streptopelia chinensis 0.03 0.00 0.03 0.00 0.00 Zosterops japonicus 5.75 3.86 1.97 1.37 0.97 Total visits perlhour 6.10 (0.83) 4.19 (0.94) 2.37 (0.37) 1.46 (0.40) 1.78 (0.40) Total foraging visitslhour 3.13 (0.46) 1.40 (0.40) 0.13 (0.06) 0.10 (0.06) 0.00 (0.00)

Fruits consumed per visit Acridotheres tristis 0.00 0.00 0.00 0.00 0.00 Cardinalis cardinalis 6.60 (3.39) 1.00 (0.00) 1.50 (0.29)b 0.00 0.00 Carpodacus mexicanus 0.00 0.00 0.00 0.00 0.00 Lonchura punctulata 0.00 0.00 0.00 0.00 0.00 Mimus polyglottos 5.00 (0.00) 3.00 (1.00) 0.00 0.00 0.00 Streptopelia chinensis 0.00 0.00 0.00 0.00 0.00 Zosterops japonicus 2.03 (0.12) 1.18 (O.13)a 0.00 0.67 (0.33)C 0.00 Total consumed/foraging visit 2.26 (0.19) 1.26 (0.14) 1.50 (0.29) 0.67 (0.33) 0.00 (0.00) amostly pecked at pulp bseeds depredated Conly pulp consumed

Seed Size andBirdDispersal

Ofthe 11 bird-dispersed species collected in the seed rain at KNAR (Table 2.9), there was a significant negative relationship between numbers ofbird-dispersed seeds

35 and seed width (rs = -0.699, P = 0.015) but not seed length (rs = -0.374, P = 0.245). When the three additional fleshy-fruited species present in the seed rain were included in the analysis, there were significant negative relationships between seed length (rs = -0.586, P

= 0.026), seed width (rs = -0.686, P = 0.006) and numbers ofseeds dispersed by birds.

Table 2.9. Mean (± 1 SE) seed length and width of 14 seeds collected in the seed rain at KNAR. Sample sizes are for seed measurements. Greatest length Greatest width # bird- Seed species without pulp without pulp n dispersed (rnrn) (rnrn) Bocconia frutescens 4.08 (0.03) 2.71 (0.02) 100 809 Lantana camara 4.27 (0.06) 2.64 (0.04) 59 190 Nothocestrum latifolium 2.79 (0.04) 1.45 (0.13) 33 33 Osteomeles anthyllidifolia 5.28 (0.12) 2.92 (0.10) 16 24 Wikstroemia monticola 5.48 (0.32) 2.73 (0.10) 9 10 Cocculus orbiculatus 5.01 (0.28) 3.21 (0.23) 7 8 Leptecophylla tameiameiae 2.88 (0.05) 2.40 (0.11) 6 5 Schinus terebinthifolius 3.90 (0.00) 3.40 (0.00) 1 2 Diospyros sandwicensis 13.20 (1.90) 6.90 (0.09) 100 2 Myoporum sandwicense 4.90 (0.00) 3.20 (0.00) I 1 Alyxia oliviformis 6.80 (0.00) 6.20 (0.00) 1 1 Reynoldsia sandwicensis 5.34 (0.05) 2.92 (0.03) 100 0 Pleomele auwahiensis 7.62 (0.06) 6.15 (0.06) 100 0 Santalum ellipticum 7.88 (0.05) 6.96 (0.04) 100 0

DISCUSSION

The results ofthis study demonstrate the importance oftrees as a factor influencing the distribution ofbird-dispersed seeds. The much higher densities and greater species richness ofbird-dispersed seeds under all trees contrasts with the relative absence ofthese seeds in exposed sites. The lack ofsuitable perches appears to limit

36 deposition ofbird-dispersed seeds into exposed areas, as birds tend to concentrate seed rain offleshy-fruited species under trees or artificial perches ofsimilar structure

(McDonnell and Stiles 1983; Guevara and Laborde 1993; Ferguson and Drake 1999;

Jordano and Schupp 2000; Shiels and Walker 2003). The significantly higher seed rain and species richness ofbird-dispersed seeds under dead standing trees versus exposed sites further emphasizes this fact (Table 2.4). Data collection under fleshy-fruited species in this study could also increase seed deposition under trees, as other studies have documented higher bird-dispersed seed rain under fruiting versus non-fruiting trees

(Slocum and Horvitz 2000; Clark et al. 2004). Although this may be true in some cases, the bird-dispersed seed rain under dead trees did not differ from any ofthe fleshy-fruited tree species (Figure 2.2). Nevertheless, comparisons ofbird-dispersed seed rain during the fruiting seasons offleshy-fruited trees may detect differences between fruiting and non-fruiting trees that were not apparent in the 12-month analysis.

The higher densities offleshy-fruited seeds under trees versus exposed sites, whether bird-dispersed or not, conform to expectations ofa leptokurtic seed shadow, with greatest densities occurring close to the seed source (Willson and Traveset 2001).

Disperser limitation may exaggerate this pattern, as three ofthe species used in this study,

(Pleomele auwahiensis, Reynoldsia sandwicensis, and Santalum ellipticum), were not recorded in the bird-dispersed seed rain and another, (Diospyros sandwicensis), was relatively scarce (Table 2.5). In contrast, densities ofwind-dispersed seeds did not differ significantly between trees and exposed areas (Table 2.1). This could be partly due to the relative abundance ofwind-adapted grasses at the study site (Table D.1, Appendix D), but the lack ofdependence on animal vectors for dissemination must also play an

37 influential role. Ofthe 11 bird-dispersed species recorded in this study, two fleshy-fruited non-natives, Bocconiafrutescens and Lantana camara, were the most abundant and widespread in the bird-dispersed seed rain (Table 2.5-2.6). Several factors could account for this dominance. Lantana camara is the most common shrub in the study site (Table

D.l, Appendix D), and B.frutescens is the most common non-native tree (Table D.2,

Appendix D). Nevertheless, all native fleshy-fruited trees combined are more abundant than Bocconia alone (Tables D.2-D.3, Appendix D), so additional factors must account for the latter's higher seed rain. Lantana and Bocconia may produce larger fruit crops of relatively small-sized seeds that, by sheer number, have a greater probability ofbeing deposited in seed traps, or that are removed at higher rates by frugivores attracted to the abundant food source (Howe and Estabrook 1977; Izhaki 2002). Although fruit crop sizes were not estimated for the different trees in this study, comparisons ofconspecific seed rain under trees indicate that Bocconia seeds fall at significantly higher densities under parent trees than do any ofthe native species (Figure 2.3). In addition, similar or greater densities ofBocconia seeds versus conspecific seeds (i. e. seeds ofsame species as overhanging tree) were collected under the four native fleshy-fruited trees (Figure 2.4).

These results suggest that fruit crops ofBocconia are much larger than those ofthe four native study trees, despite the fact that Bocconia trees are ofrelatively equal or smaller stature than the natives (Figure A.l, Appendix A).

Fruiting phenology oftrees may also account for the disparities in the bird­ dispersed seed rain (Carlo et al. 2003). Frugivores have been shown to track changes in availability ofimportant food supplies (Loiselle and Blake 1991; Price 2004), and may focus on particular species that produce fruit when most others are sterile (Howe and

38 Primack 1975). The large fruit crop ofBocconia trees, in particular, which peaked during the summer months ofJune through August, may have been one ofthe few foods available to frugivores at a time ofyear when other fruit resources were relatively scarce

(Table 8.1, Figures 8.1 and 8.3, Appendix B). This could be a result ofannual variation, however, so longer-term comparisons ofbird-dispersed seed rain, fruiting phenology and frugivore abundance would be useful in determining to what degree fruiting seasonality influences amounts and rates offruit removal and dispersal ofparticular fleshy-fruited speCIes.

Because many ofthe extinctions ofthe avifauna in the Hawaiian Islands and throughout Polynesia occurred during the early periods ofhuman colonization (Olson and

James 1982, 1991; Steadman 1995; Burney et al. 2001; Ziegler 2002), it is impossible to determine conclusively all ofthe community level effects these extinctions have had on plant demographics. Nevertheless, the abundance offleshy-fruited taxa in the Hawaiian

Islands and similar insular ecosystems such as New Zealand suggests that some ofthese extinct birds would have been attracted to and likely played an important role in the dispersal and potential scarification offleshy-fruited species (Clout and Hay 1989;

Wagner et al. 1999). Because isolated islands tend to have more depauperate assemblages ofdispersers than continental areas do, islands may be even more dependent on the limited dispersal pool for the preservation offloral biodiversity (Bleher and Bohning­

Gaese 2001; Cox et al. 1991; McConkey and Drake 2002). When native species do go extinct, dispersal failure could result, or the original native dispersers may be replaced by other natives or newly introduced species that serve as ecological substitutes for processes such as seed dispersal (Loiselle and Blake 2002; Hampe 2003).

39 In the Hawaiian Islands, the almost complete replacement ofthe native avifauna with a suite ofnon-native generalist frugivores and game birds at lower elevations provides an opportunity to explore the role that non-natives do playas potential surrogates in seed dispersal and other ecological processes (Pratt 1997). Bird observations on the four fleshy-fruited native study tree species and one non-native species, although admittedly limited in scope, do provide insights into whether or not ecological substitutions are occurring. Ofthe 16 non-native bird species documented in

KNAR (Medeiros et al. 1993), 12 were observed at some point during the 12-month duration ofthis study (Table C.1, Appendix C), but only nine consume fruit pulp as a regular part oftheir diet (Scott et al. 1986). Ofthese nine, only three were observed visiting and consuming the pulp and/or seeds ofsome ofthe study trees. The most abundant species at the study site was the Japanese white-eye, a now ubiquitous bird first introduced to the islands in 1929 (Appendix C; Scott et al. 1986). Although this bird was regularly observed swallowing the seeds and attached arils ofBocconiafrutescens, it predominantly pecked at the pulp ofthe two native species it also visited (Table 2.8).

Many ofthe berries ofReynoldsia sandwicensis were dislodged from the fruiting panicle as white-eyes pecked at the pulp, causing them to fall directly under the parent canopy.

White-eyes are apparently dispersing Bocconia seeds and in limited instances may disperse the seeds ofnative trees, but in general, species that do not provide the benefit of seed dispersal when feeding on fleshy fruits have been referred to as "fruit predators"

(Herrera 1995). Ofthe other two non-native species observed feeding on fruits, both the northern cardinal and the northern mockingbird swallowed whole fruits and seeds ofboth

B. frutescens and R. sandwicensis (Table 2.8). Either could be considered legitimate

40 dispersers ifthey pass viable seeds ofthese or other trees (Jordano 2001). Although both species were among the more abundant recorded in KNAR, both are relatively uncommon compared to the Japanese white-eye, and their roles as dispersal agents may be limited (Table C1 and C2, Appendix C). In addition, rather than dispersing seeds of

Santalum ellipticum, northern cardinals act as pre-dispersal seed predators by clipping the immature fruits in halfto feed on seeds. This behavior, combined with the impacts of post-dispersal seed predation by introduced rodents (Chapter 3), could severely limit or entirely prevent the recruitment ofthis endemic species.

Ofthe other plant species recorded in the bird-dispersed seed rain (Table 2.5),

Japanese white-eyes were anecdotally observed feeding on the fruits offour (Lantana camara, Nothocestrum latifolium, Osteomeles anthyllidifolia and Wikstroemia monticola). Dispersal ofthese and other fleshy-fruited taxa, in addition to being related to fruiting phenology, plant abundance and fruit crop size, may also be influenced by seed size. Both fruit and seed size are considered to be important factors affecting the type and number offrugivores capable ofconsuming fruits and potentially dispersing seeds (Snow

1981; Wheelwright 1985; Levey 1987). Smaller frugivores, limited by body size and gape width, may in turn influence the evolution offruit and seed size by selectively feeding on relatively smaller fruits or seeds (Lord 2004). In KNAR, the most abundant generalist frugivore, the Japanese white-eye, is a fairly small bird ofbetween 11-12 cm body length, 9.75-12.75 g weight, and gape width ofbetween 5-8 mm (Corlett 1986; Van

Riper 2000). In Australia and New Zealand, the silvereye (20 lateralis), a related species ofsimilar size and gape, does consume the pulp offruits exceeding its gape width

(Williams and Karl 1996), and prefers relatively larger over smaller fruits, up to a certain 41 point (Stanley et al. 2002; Stansbury and Vivian-Smith 2003). Neither ofthese studies indicates whether silvereyes are capable ofdispersing the seeds ofthese larger fruit, but in general, most birds have a size limit over which the handling costs outweigh the benefits offruit and seed consumption (Martin 1985; Levey 1987). In this study, there was a negative relationship between seed width and number ofseeds dispersed, and three ofthe larger-seeded species were rarely, or never, consumed by birds or collected in the bird-dispersed seed rain (Table 2.8-2.9). Most ofthe bird-dispersed seed sizes fell within or below the recorded gape width ofthe Japanese white-eye, and presumably also the much larger northern cardinal and mockingbird (Table 2.9). Therefore, although further study is necessary to determine whether seed size limits the dispersal oflarger-seeded species, the relatively smaller sizes ofthe abundantly dispersed Bocconia and Lantana seeds certainly did not hinder their consumption and dissemination.

Because bird-dispersed seeds in forested areas have heterogeneous spatial distributions (Dalling et al. 1998; Ferguson and Drake 1999), it is possible that several species that birds had dispersed were missed by chance and were therefore not represented in the documented seed rain. Nevertheless, it is clear that a large proportion ofthe fleshy-fruited seeds from the study species are not being dispersed and are instead falling directly under parent trees. Several negative consequences could result from this lack ofdispersal. Seeds occurring at higher densities under parent trees may be subjected to greater levels ofpredation (Howe et al. 1985; Chapman and Chapman 1996; Hulme

1997; Wenny 2000; Ho1l2002). The presence offruit pulp on seeds that are not dispersed may even aid seed predators in finding and consuming seeds (Moles and Drake

1999, Nystrand and Granstrom 1997, Chapter 3). For those seeds that escape predation, 42 germination may be reduced due to the presence ofpulp on seeds or lack ofseed scarification by birds or other dispersal agents (Fukui 1995, Yagihashi et al. 1998;

Traveset and Verdu 2002). Seedlings that germinate under parent trees may experience higher rates ofmortality due to density-dependent seedling predation, seedling competition or greater exposure to seedling pathogens (Augspurger 1984; Chapman and

Chapman 1994; Kitamura et al. 2004). Finally, restoration ofdegraded areas may be limited by lack ofdispersal ofseeds into these sites (Holl et al. 2000; Cordeiro and Howe

2001; Howe and Miriti 2004; Makana and Thomas 2004). Many ofthese factors may be contributing to the relatively low frequency ofseedlings ofcertain native fleshy-fruited species under parent trees (Table 2.8), and, combined with other factors such as lack of outcrossing and browsing and trampling by feral ungulates, may be hastening the continued decline ofdry forests at KNAR and throughout the Hawaiian Islands.

Plants that benefit from directed dispersal (e.g. seed dispersal by birds under trees) tend to increase in abundance (Wenny 2001). For those species benefiting from the effects ofdirected dispersal by the non-native frugivores at KNAR, seedling recruitment under trees does not appear to be a problem (Table 2.8). Although the effects ofdispersal may prove beneficial to the recruitment and recovery ofnative species, directed dispersal may also facilitate the spread ofinvasive, habitat modifying weeds (Richardson et al.

2000; Cordeiro et al. 2004). This appears to be the case for Bocconiafrutescens and

Lantana camara, the two most common fleshy-fruited non-native plants in KNAR.

Lantana camara, first introduced to the Hawaiian Islands in 1858, is now widespread throughout the low-elevation dry and mesic habitats ofall the main islands (Wagner et al.

1999). Bocconiafrutescens, a more recent invader, was first collected in Kanaio, Maui in 43 1920, but has rapidly spread throughout KNAR and the leeward slopes ofthe eastern half ofthe island (Medeiros et al. 1993; Wagner et al. 1999; Appendix D). Habitat shaping, the result ofdisperser and plant interactions, occurs when seed dispersal contributes to plant composition and abundance over an area (Herrera 1995). This seed dispersal by frugivorous birds can also lead to the homogenization ofplant distributions over an entire region (Debussche et al. 1982). The relative abundance ofBocconia and Lantana plants in the study site, the frequency and abundance oftheir seeds in the seed rain under all study species, the frequency ofseedlings under and away from parent trees, and the frequent interaction with Japanese white-eyes, all suggest that just such a homogenization is occurring at KNAR and adjacent dry forest areas, to the likely detriment ofthe remaining native species.

Conclusion

The much higher seed densities offleshy-fruited and bird-dispersed seeds under trees versus exposed sites emphasizes the importance oftrees as dispersal and potential recruitment foci for seeds and seedlings. In degraded ecosystems such as those ofthe remaining Hawaiian dry forests, these sites may facilitate restoration and recovery processes when desirable species are present in the seed rain (Otero-Arnaiz et al. 1999;

Holl et al. 2000; Guevara et al. 2004). Despite the loss ofnative dispersal agents, non­ native species will often serve as ecological substitutes and fill the roles left vacant by extinct dispersers (Loiselle and Blake 2002). Although not investigated in this study, non­ native game birds may be scarifying and secondarily dispersing both native and non­ native seeds, as they have done in other Hawaiian ecosystems (Cole et al. 1995). Aside

44 from a few small-seeded natives, however, the majority offleshy-fruited species being dispersed by non-native frugivores at KNAR are relatively small-seeded non-native trees and shrubs. Whether or not seed size and gape width limitations are contributing to this

disparity, it is clear that bird perches and other structures that enhance seed dispersal into degraded habitats will only facilitate further invasion by these non-native species. New introductions oflarger frugivores could increase the number oflarge-seeded natives being dispersed, but until the seed sources ofinvasive species are eliminated, only human intervention will ensure the selective dispersal and perpetuation ofdesirable native species over invasive non-native weeds.

45 CHAPTER 3: POST-DISPERSAL SEED PREDATION: LOCATION AND

EFFECT ON SPECIES

INTRODUCTION

The flora and fauna ofPacific island ecosystems, including the Hawaiian Islands, evolved in the absence ofnative rodents, and presumably have fewer defenses against them than species that evolved with rodents. Introduced rodents such as the Polynesian rat (Rattus exulans), the black or roofrat (R. rattus), the Norway rat (R. norvegicus) and the house mouse (Mus musculus) have been introduced to many islands throughout the world, with devastating impacts on the insular biota (Tomich 1986). In the Hawaiian

Islands, rats have negatively impacted the native avifauna (Atkinson 1977; Scott et al.

1986), the endemic snail and arthropod fauna (Hadfield et al. 1993; Cole et al. 2000) and the flora, including many rare and endangered species (Wirtz 1972; Baker and Allen

1978; Russel 1980; Scowcroft and Sakai 1984; Stone et al. 1984; Stone 1985; Sugihara

1997; Cole et al. 2000). Although introduced rodents are notorious plant and seed predators in other insular ecosystems (Fall et al. 1971; Clark 1982; Campbell et al. 1984;

Allen et al. 1994; Moles and Drake 1999; Williams et al. 2000; Campbell and Atkinson

2002; Delgado Garcia 2002; McConkey et al. 2003), and have been anecdotally implicated in the destruction ofseeds in Hawaiian dry forests, (Medeiros et al. 1986;

Medeiros et al. 1993; Cabin et al. 2000), no studies have been published that quantitatively document the impacts ofthese introduced rodents on the seeds ofparticular dry forest taxa. As many ofthese species produce fruits with one to few larger seeds

(Wagner et al. 1999), and as rodents in general have been shown to be voracious seed

46 predators ofintennediate and larger seeds (Stiles 2001), the effects that introduced rodents have on seed mortality and subsequent levels ofseedling recruitment could be profound (Crawley 2001).

A troubling phenomenon that seriously affects the persistence as well as the potential future restoration ofdry forest communities is the almost complete lack of natural reproduction ofnative tree species (Medeiros et al. 1986; Loope et al. 1995;

Loope 1998; Cabin et al. 2000). Janzen (1988) refers to adult trees that fail to reproduce as "the living dead" and indicates that this problem is symptomatic oftropical dry forests worldwide. Several factors may contribute to the lack ofseedling recruitment in tropical dry forests. These include browsing and trampling by ungulates, competition with weeds, the loss ofnative pollinators, the impacts ofnon-native invertebrates and pathogens on seed and seedling survival, modifications in microclimate and microhabitats suitable for seed gennination and seedling survival, loss ofnative birds that scarified and dispersed seeds, and seed predation by introduced rodents (Janzen 1988; Aide et al. 2000; Cabin et al. 2000; Holl et al. 2000; Zimmennan et al. 2000). The effect ofrodent predation on seeds is one factor that could be influencing community composition and structure of

Hawaiian dry forests.

The scattered spatial distribution ofdry forest trees could also influence the deposition ofseeds throughout an area. At the Kanaio Natural Area Reserve and other remnant Hawaiian dry forests, isolated trees and groves are surrounded by lower-statured native shrublands, non-native grasslands and largely barren lava flows (Medeiros et al.

1993). Previous studies suggest that seed deposition and predation rates can differ markedly between open habitats and under trees. Isolated trees or groves oftrees in

47 otherwise open areas may serve as recruitment foci for bird-disseminated species

(McDonnell and Stiles 1983; Ferguson and Drake 1999). This dispersal results in a non­ random seed rain due to the preferential movements ofbirds to these feeding and perch sites (Jordano and Schupp 2000; Clark et ai. 2004). These recruitment foci also sometimes serve as suitable microsites for the growth ofseedlings and may play important roles in the succession and regeneration ofdisturbed or degraded areas (Uhl

1998; Wijdeven and Kuzee 2000).

Despite the potential benefits, however, this non-random deposition under trees, combined with the higher seed rain predicted by the seed shadow under parent trees, could also influence seed predation rates (Hulme 1998). Several authors have documented higher seed predation rates under trees and vegetation as opposed to more open areas (Howe et ai. 1985; Callaway 1992; Aide and Cavelier 1994; Osunkoya 1994;

Chapman and Chapman 1996; Hau 1997; Hulme 1997; Wenny 2000; HoIl2002). Others have found no difference (DeSteven and Putz 1984; Terborgh et al. 1993), or higher rates ofpredation in exposed areas or grasslands (Hay and Fuller 1981; Uhl1998; Wijdeven and Kuzee 2000). One study reported lower predation rates under isolated pasture trees as opposed to either open pastures or intact forests (Holl and Lulow 1997). Seed predators may avoid exposed sites to minimize risks from aerial predators such as owls (Munia and

Gonzalez 1982; Sanchez-Cordeiro and Martinez-Gallardo 1998), a factor that could influence seed removal rates in Hawaiian dry forests. Lower seed densities in exposed sites may also contribute to lower predation rates, as some seed predators may concentrate their activity in optimal foraging areas ofhigher seed densities closer to the seed source (Janzen 1971; Hulme 1998).

48 Therefore, to better understand the interaction between the distribution ofdry forest trees and seed predation in different microhabitats ofthe Kanaio Natural Area

Reserve, the following questions are addressed in this paper:

1) Are rodents removing and destroying seeds ofdry forest species?

2) Is there a difference in removal rates ofseeds under trees and in exposed

sites?

3) Does the presence offruit pulp or fleshy arils affect rates ofremoval?

4) Is there a relationship between seed size and removal rates for dry forest tree

species?

METHODS

Study Site

This study was conducted in the Kanaio Natural Area Reserve (KNAR), on the leeward side ofEast Maui, Hawai'i, at an elevation between 750-850 meters (200 36'N,

156°20'W; Figure 2.1). The climate in the reserve is hot and dry, with annual temperatures between 20-30° C and with mean annual rainfall ofapproximately 750 mm, mostly falling between the months ofOctober to April (Giambelluca et al. 1986). The substrate is estimated to be less than 10,000 years old and consists mostly of'a'a lava with some accumulation ofoverlying soils in some areas (Crandell 1983). Medeiros et al.

(1986; 1993) have provided a thorough description ofthe community composition ofthe dry forest ofKNAR and the leeward slopes ofHaleakala, Maui. The 354-hectare reserve was established in 1990 to protect exemplary representatives ofHawaiian dry forest species and plant communities, including Dodonaea ('A 'ali'i) lowland shrublands,

49 Diospyros (Lama) forest and Erythrina (Wiliwili) forests (Gagne and Cuddihy 1990). In addition to these communities, the reserve also contains individuals and groves ofat least twenty-two native tree species. Anecdotal observations suggest that only eight ofthese trees may be reproducing (Medeiros et al. 1993). Certain non-native trees, however, including the widespread Bocconiafrutescens (Papaveraceae), are able to reproduce without the apparent difficulties experienced by native species (Chimera pers. obs.). At least two species ofnon-native rodents, the black or roofrat (Rattus rattus), and the house mouse (Mus musculus) currently occur in KNAR (Medeiros et al. 1993).

Measuring Seed Predation Levels

To examine to what degree rodents remove tree seeds in KNAR, seeds from four native and one non-native tree species were used in a series ofpredation trials. Native species were selected to represent a range ofseed sizes. Native fleshy-fruited taxa used for this study include Reynoldsia sandwicensis (Araliaceae), Diospyros sandwicensis

(Ebenaceae), Santalum ellipticum (Santalaceae), and Pleomele auwahiensis (Agavaceae), as all are locally abundant at the site. One non-native tree with arillate seeds, Bocconia frutescens (Papaveraceae), was also used. Mean tree heights ranged between approximately three to six meters (Appendix A). Fifteen trees ofeach species were selecteq from the northeasWm portion ofKNAR for the pl

50 Study trees were randomly selected from this subset ofindividuals such that no selected trees were closer than 25 meters from each other.

Treatments

Four treatments were used, based on designs ofMoles and Drake (1999): open ground, a depression (11 cm x 11 cm x 1 cm deep) in the ground with a rim so that seeds did not wash away; open pot, square black plastic flower pots (110 mm x 110 mm wide;

150 mm tall) with one side cut away and filled with soil such that the inside and outside soil levels are the same; pot with rodent access, open pots covered with 12-mm steel mesh with an opening (35 mm x 35 mm) large enough to allow access to rodents but small enough to deter feeding by larger game birds; pot with rodent-proofmesh, open pots with 12-mm mesh designed to prevent entry by rodents but not invertebrates.

Sets offour treatments were placed under the canopy ofeach tree species such that a distance ofat least one meter separated each treatment. These treatments were generally shaded from direct sunlight for a good portion ofdaylight hours. Sets offour treatments were also placed in exposed sites (i.e. not under trees), on a random compass direction and at a minimum distance offive meters from the nearest tree crown. These exposed sites consisted ofbarren rock, low-stature grasses, woody shrubs and annual or perennial herbaceous species less than 50 cm tall. Treatments in exposed sites were not placed under vegetation and were therefore exposed to direct sunlight. Experiments were arranged in a 2x4 factorial block design for three species and a 2x5 factorial block design for two species (Bocconia and Diospyros) with an additional treatment (see below). Each

51 tree constituted an individual block with four or five treatments (factor 1) and two locations (under trees and in exposed sites, factor 2), per block (Zar 1999).

For each species, a pile offive seeds was placed in each ofthe four treatments under each ofthe 15 conspecific fruiting trees and in each ofthe 15 exposed sites, for a total of600 seeds per species (300 under trees, 300 in exposed sites). Seeds were collected from no fewer than five haphazardly-selected trees ofeach species as they became available. Length and width were recorded for 100 haphazardly-selected seeds of each ofthe five species used in the predation trials. Fruit pulp was removed from seeds unless otherwise stated. Pulp was removed using water and paper towels, and seeds were allowed to dry before being used. For two species, Diospyros sandwicensis and Bocconia frutescens, additional piles offive intact berries for Diospyros (1-3 seedslberry), or seeds with arils still attached ("arillate seeds" for Bocconia) were also placed in open ground treatments under trees and in exposed areas. For these species, removal rates were compared between seeds and whole fruits to determine the effect offruit pulp on predation. For one species, Pleomele auwahiensis, a pile offive fresh berries (1-3 seeds/berry) with pulp intact was placed in each ofthe four treatments under a different set of 15 trees and exposed sites, for a total of600 fruits from this species (300 under trees, 300 in exposed sites).

Removal ofseeds and fruits was recorded over five 15-day periods in 2003, commencing on 2 April for P. auwahiensis seeds and fruits, 4 April for S. ellipticum, 15

April for R. sandwicensis, 29 April for D. sandwicensis, and 4 June for B.frutescens. In all trials, seeds were used within a few days after collection to mimic the removal of seeds when they would normally be available to predators. Seeds were categorized as

52 removed ifthey were taken from the treatment area or ifthey had been destroyed at the treatment site. It is possible that some ofthe removed seeds may have actually been dispersed by rodents or game birds, and were not necessarily consumed. Measures of seed removal may therefore overestimate levels ofpredation.

Analysis

To determine whether predation levels under parent trees differ from those in exposed sites away from the parent tree, numbers ofseeds remaining in the sets of treatments under trees were compared with those in exposed sites. This analysis investigated differences between removal levels under trees ofeach species versus associated exposed areas, but did not compare levels between different species.

Differences between combinations oftreatment and location (tree or exposed) were compared using a general linear model, with mean number ofseeds remaining as the dependent variable, individual tree as the blocking or random factor, and treatment and location as explanatory variables with one interaction term: treatment x location. Tukey tests were used to determine which treatments were significantly different (Zar 1999;

Ryan and Joiner 2001). Because removal trials for Pleomele auwahiensis seeds and fruits were conducted at the same time, mean numbers ofseeds and fruits remaining in open ground treatments were further compared between locations using a two-way analysis of variance, with a Tukey test used to identify differences among treatments.

53 Spearman's rank correlation was used to test for relationships between seed length, width and the mean number ofcleaned seeds remaining in open treatments after

15 days, both under trees and in exposed sites.

RESULTS

Seed removal rates after 15 days varied greatly per species and by location (Table

3.1; Figure 3.1), ranging from 100% ofSantalurn seeds under trees to 0% ofReynoldsia seeds under trees or in exposed sites. Rodent tooth marks, droppings and seed husks suggest that rodents are responsible for at least some ofthe seed removal recorded.

Bocconia and Pleornele seed husks remained in many ofthe treatments. Rodent bite marks and partially consumed seeds were present in several ofthe Diospyros treatments.

Seed husks and fragments were also present in about halfofthe accessible Santalurn treatments in which seeds were removed.

There were significant differences in removal rates between seeds and fruits set out in different treatments and in different locations (Table 3.2; Figures 3.2-3.5). For

Bocconiafrutescens, there were significant differences in removal rates between treatments, but not between locations (Table 3.2; Figure 3.2). Significantly more arillate seeds were removed than cleaned seeds (Figure 3.2). Some ofthe remaining arillate seeds had arils removed with no apparent damage to seeds. For Diospyros sandwicensis, significant differences in removal rates among treatments were dependent on location, and overall removal rates were higher under trees (Table 3.2; Figure 3.2). Significantly more fruits than seeds were removed from trees and exposed sites (Figure 3.2). For

Pleornele auwahiensis, significantly more seeds were removed from trees than exposed

54 sites, but removal rates for fruits were not significantly different between locations (Table

3.2; Figure 3.3). Significantly more Pleornele fruits were removed than seeds, but location was not significant (Table 3.3; Figure 3.4). For Reynoldsia sandwicensis, no seeds were removed from any ofthe treatments during the IS-day duration ofthe trial

(Figure 3.5). For Santalurn ellipticum, significantly more seeds were removed from trees than exposed sites (Table 3.2; Figure 3.5). No seeds or fruits ofany species were removed from rodent-proofpots at any point during the trials, and no signs of invertebrate damage were noted during the IS-day study duration.

No significant relationships were found between the number ofcleaned seeds remaining in open ground treatments after 15 days under trees or exposed sites and seed length (P= 1.0), or seed width (P=0.517)

Table 3.1. Means (± 1 SE) ofseed length and width and mean removal rate under trees and in exposed sites for species used in the seed predation trials. Percentages represent open ground treatments only. n = 100 for seed measurements. % ofcleaned Greatest Greatest % ofcleaned seeds length width seeds Species removed after without pulp without removed after 15 days (mm) pulp (mm) 15 days (tree) (exposed) Bocconiafrutescens 4.08 (0.03) 2.71 (0.02) 77.33 64.00 Diospyros sandwicensis 12.44 (0.15) 4.98 (0.09) 49.33 2.67 Pleornele auwahiensis 7.62 (0.06) 6.15 (0.06) 66.67 56.00 Reynoldsia sandwicensis 5.34 (0.05) 2.92 (0.03) 0.00 0.00 Santalurn ellipticurn 7.88 (0.05) 6.96 (0.04) 100.00 81.33

55 Table 3.2. General linear models for number ofseeds and/or fruits remaining versus block, treatment, location (tree or exposed), and the interaction oftreatment (Trt) and location (Loc) after 15 days for four dry forest trees in Kanaio Natural Area Reserve. Tests were not performed on Reynoldsia, as no seeds were removed from any treatments.

Source of SS Species d.f. MS F P variation (model III) Bocconiafrutescens Block 14 47.76 3.41 2.58 0.003 Treatment 4 481.76 120.44 90.92 0.000 Location 1 3.23 3.23 2.44 0.121 Trt x Loc 4 4.107 1.03 0.78 0.543 Error 126 166.91 1.33 Total 149 703.76 Diospyros sandwicensis Block 14 22.17 1.58 1.03 0.427 Treatment 4 380.17 95.04 61.91 0.000 Location 1 42.67 42.67 27.79 0.000 Trt x Loc 4 25.13 6.28 4.09 0.004 Error 126 193.43 1.54 Total 149 663.57 Pleomele auwahiensis Seeds Block 14 116.12 8.29 3.30 0.000 Treatment 3 144.03 48.01 19.11 0.000 Location 1 28.033 28.03 11.16 0.001 Trt x Loc 3 16.03 5.34 2.13 0.102 Error 98 246.15 2.51 Total 119 550.37 Pleomele auwahiensis Fruits Block 14 110.05 7.86 3.34 0.000 Treatment 3 336.17 112.06 47.65 0.000 Location 1 4.03 4.03 1.71 0.193 Trt x Loc 3 8.57 2.86 1.21 0.309 Error 98 230.48 2.35 Total 119 689.30 Santalum ellipticum Block 14 74.97 5.36 4.33 0.000 Treatment 3 440.57 146.86 118.78 0.000 Location 1 14.70 14.70 11.89 0.001 Trt x Loc 3 6.57 2.19 1.77 0.158 Error 98 121.17 1.24 Total 119 657.97

56 100% \, Boc fru .-- --I. 80% -, Dio san 60% :i--_-!~-{--) 40% -;~:~::::::~"1-.__ -1--L-_. i \ "'\t=~~ \, I ------+-- -t- --. -__~ 20% - t,J -_ i ! 0% ------~T-==----'i-=--=--=-f=-:::::----T -----·'·11~--~---·-"t~J-----I,}---:;J

0) c Pie auw Seeds Pie auw Fruits c ro \,\\ E Q) \:~--~~-~~-~-+- t ---i lo- t --'I~J--I--! (f) +------t t ""0 Q) Q) (f) 100% San ell 80% Rey san 60% ­ '.",'.\ 40% ­ \ \ "\'.----- _ f r r r 20% - --+--. ... • 1 I 1 < J______0% --- , I '! I o 3 6 9 12 15 0 3 6 9 12 15

Number of days since start of trial

Figure 3.1. Removal rates ofseeds and fruits from the Kanaio Natural Area Reserve. Percentage ofseeds or fruits remaining ± 1 SEM for open ground treatments under trees and in exposed sites (n = 15 trees * 5 seeds per treatment).0 = cleaned seeds under trees; ~= cleaned seeds in exposed sites; 0= seeds under trees (with pulp); • = seeds in exposed sites (with pulp). Boc fru = Bocconiafrutescens; Dio san = Diospyros sandwicensis; PIe auw = Pleomele auwahiensis; Rey san = Reynoldsia sandwicensis; San ell = Santalum ellipticum.

57 100% - Bec fru T % - 1 1 80 T

tn ~ 60% - ~ "C an 40 % - r

""""~ (1) I ¢:: 20% - ~ J,. C) BC AA A B AA A (t s:::: % II - I .- 0 I .-s:::: % 100 - T T - E Die san T ~ tn - (1) 80% ~ 0 r C. % - tn 60 .-~ l c 40% - r 20% - CD BC AB A E AAAA DE 0% G 0 RA RP F G 0 RA RP F Tree Exposed Treatment

Figure 3.2. Removal rates for BocconiaJrutescens (Doc fru) and Diospyros sandwicensis (Dio san) seeds and fruits under trees and in exposed sites. Treatments include G: open ground; 0: open pots; RA: pots with rodent access; RP: rodent-proof pots; F: fruits on open ground (n = 15 trees * 5 seeds/fruits per treatment). Bars represent the mean + 1 SEM. Letters at the base ofbars show the results ofthe Tukey test. Treatments sharing the same letter are not significantly different (p> 0.05). Removal rates were not compared between Bocconia and Diospyros.

58 - 100°A» Pie auw seeds r r 80°A» - 1 60% - 1 ~ ca 1 "C 40% - r ....It) ~ - CD 20% =ca e e Be A e AB AB A en O°A» I .-c c 100°A» - .-ca Pie auw fruits E l! 80% - In CD ~ a.0 60% - Inca .-c 40°A» - 1 1 1 1 1 20% - B B B A BBB A O°A» G 0 RA RP G 0 RA RP

Tree Exposed Treabnent

Figure 3.3. Removal rates for Pleomele lIuwllhiensis (pIe auw) seeds and fruits under trees and in exposed sites. Treatments include G: open ground; 0: open pots; RA: pots with rodent access; RP: rodent-proofpots (n = 15 trees· 5 seedslfruits per treatment). Bars represent the mean + 1 SEM. Letters at the base ofbars show the results ofthe Tukey test. Treatments sharing the same letter are not significantly different (p > 0.05).

59 (J) >- ns "C LO 100% - T'"".... Q) iI= ns 75% - c0> "c 50% - ns I E Q) T .... - (J) 25% ~ ::J.... AB ~ A AB ~ (J) 0% "C Q) Q) Seed Fruit Seed Fruit (J) Tree Exposed Treatment

Figure 3.4. Removal rates for Pleomele auwahiensis seeds and fruits on the open ground under trees and in exposed sites. (n = 15 trees * 5 seeds/fruits per treatment). Bars represent the mean + 1 SEM. Letters at the base ofbars show the results ofthe Tukey test. Treatments sharing the same letter are not significantly different (p > 0.05).

Table 3.3. Two-way ANDVA results for the effects oftreatment (seed or fruit) and location (tree or exposed) on removal rates for Pleomele auwahiensis Source ofvariation d.f. SS MS F p Treatment (seed/fruit) 1 22.82 22.82 5.56 0.022 Location (tree/exposed) 1 12.15 12.15 2.96 0.091 Interaction 1 2.02 2.02 0.49 0.486 Error 56 230.00 4.11 Total 59 266.98

60 100°A> - . - Rey an 80°A> -

60°A> -

- t/) 40°A> ~ ~ 'C - 10 20°A> ~ L0- A AA AAA AA ....S O°A> I ~ C) 100°A> - .-c: San ell .-c: ~ 80% - E CD Lo- t/) 60°A> - 'C CD CD (J) 40% - 1 20°A> - r r f")h C C r'~ 1 A BC B BC A 0% G 0 RA RP G 0 RA RP

Tree Exposed Treatment

Figure 3.5. Removal rates for Reynoldsia sandwkensis (Rey san) and Santalum elliptkum (San ell) seeds under trees and in exposed sites. Treatments include G: open ground; 0: open pots; RA: pots with rodent access; RP: rodent-proofpots (n = 15 trees * 5 seeds per treatment). Bars represent the mean + 1 SEM. Letters at the base ofbars show the results ofthe Tukey test. Treatments sharing the same letter are not significantly different (p > 0.05). Removal rates were not compared between Reynoldsia and Santalurn.

61 DISCUSSION

The highly variable seed removal rates for the five tree species used in this study are not surprising, as differences between species are frequently documented in the seed predation literature (Chapman and Chapman 1996; Hulme 1998; Figueroa et al. 2002;

Jones et al. 2003). The average levels ofpredation on cleaned seeds for all species combined, whether under trees (59%) or in exposed sites (41 %), are comparable to relatively high levels documented in forests throughout the world. Rates ofseed predation by rodents typically exceed 50% for many species in a number ofdifferent ecosystems (Hulme 1998). In a forest site in Spain, about 70% offlex aquifolium seeds were removed after two 2-week periods (Ramon Obeso and Fernandez-Calvo 2002). In the northern territories ofHong Kong, between 73% and 86% ofseeds were removed for four species in shrublands, and between 33% and 83% were removed for six species in grasslands after 60 days (Hau 1997). For a tropical dry forest shrub in Mexico, 64% of seeds were removed (Gryj and Dominguez 1996). Excluding Reynoldsia sandwicensis, which had no seeds removed during the 15-day duration ofthis study (Table 3.1; Figure

3.5), average levels ofseed removal were 73% under trees and 51 % in exposed sites.

Relatively high levels ofseed predation might be expected for Hawaiian species that evolved in the absence ofrodent seed predators (Ziegler 2002). These rates are much higher than the 10% removal rate reported by Moles and Drake (1999) for 11 species in

New Zealand, in an island biota similarly lacking native rodents. For this study, however, evolutionary history may be inconsequential, as the non-native Neotropical tree Bocconia

62 frutescens had the second highest percentage ofseeds removed both from under trees and in exposed sites (Table 3.1). Other native Hawaiian dry forest species, including Nestegis sandwicensis (Oleaceae), Alectryon macrococcus (Sapindaceae), Nesoluma polynesicum

(Sapotaceae) and Pouteria sandwicensis (Sapotaceae) all suffer from high levels ofseed predation, as evidenced by the large piles ofseed husks observed under almost every fruiting tree crown (Chimera pers. obs.). Different, and potentially much higher, removal rates would be expected for every unique species combination used in similar predation trials.

Location ofseeds and fruits, whether under trees or in exposed sites, also resulted in variable removal rates. In three trials, more seeds were removed from under trees than exposed sites, but location was not significant for Bocconiafrutescens seeds and fruits or

Pleomele auwahiensis fruits (Table 3.2; Figures 3.1-3.5). Several authors have found microhabitat to be an important factor influencing levels ofseed predation. Many studies found lower levels ofpredation by small mammals in open areas (Aide and Cavelier

1994; Bustamante and Simonetti 2000; Hulme 1994; Holl 2002). Others report that higher levels occur in open sites versus under trees (UhI1998; Wijdeven and Kuzee

2000). Because seed densities have been shown to influence seed predation (Hulme

1998), different removal rates were expected under trees and exposed sites due to the much higher seed densities found under trees at KNAR (Chapter 2).

Some authors suggest that aerial predators such as owls may deter rodents from foraging in exposed areas, reducing seed removal in these sites (Murua and Gonzalez

1982; Sanchez-Cordeiro and Martinez-Gallardo 1998). Two owl species occur in KNAR, the native Hawaiian short-eared owl (Asio j/ammeus), and the common bam-owl (Tyto

63 alba), a species introduced from North America in 1958 for rodent control in sugar cane fields (Tomich 1962). Ifseed densities or aerial predators do influence removal rates in different locations, these effects were negligible in two ofthe trials (Table 3.2). It is possible that owls occur at densities too low to noticeably affect seed predation in exposed sites, and those differences that were significant may be attributed to some other factor.

For the trials that did not differ, the location ofthe exposed site treatments, within five meters ofthe tree crown, may be too close to account for a truly separate microhabitat, despite the differences in vegetative cover, exposure, amounts ofleaflitter, understory humidity or seed densities. In addition, placing five seeds in each treatment results in much higher seed densities in exposed sites than would naturally occur away from the tree crown. This increased density, intended to represent either a hypothetical dispersal event by a frugivore or a supplementary seed-scattering technique for restoration purposes, could contribute to unnaturally high predation levels, as rodents have demonstrated the ability to more easily locate bigger seed piles (Hammond 1995).

None ofthe preceding explanations satisfactorily address why removal rates differed in locations for three trials, but not for the other two, however, so yet another factor may be responsible for the variability. It may simply be that the short duration ofthe study accounted for differences detected by location (Figure 3.1), and keeping seeds out longer may result in equivalent final predation levels for all trials.

Seed size is another factor thought to affect amounts and rates ofseed predation

(Hulme 1998). Although no relationship was found between seed length and width and rates ofremoval in either location, the range ofsizes used in this study may be too small

64 to detect any size effects on predation levels. As the seeds used in this study were relatively large (Table 3.1), future experiments should utilize a broader range ofseed sizes, along with different densities, to more accurately determine ifany real relationships exist between seed size or mass and removal rates under trees and exposed sites.

Higher removal rates for fruits than seeds were reported in New Zealand (Moles and Drake 1999) and Spain (Hulme 1997). Presence ofpulp on Diospyros seeds and the fleshy aril on Bocconia seeds also resulted in higher rates ofremoval than those recorded for most ofthe cleaned seed treatments (Figure 3.2). More Pleomele auwahiensis fruits than seeds were removed, but their location was not significant (Table 3.3; Figure 3.4).

For these three species, it is unknown whether fruit removal results in destruction or dispersal ofthe seed or seeds. As intact Reynoldsia seeds have previously been found in rat droppings (Medeiros et al. 1986), the use ofcleaned seeds in the various treatments may explain the lack ofremoval ofany Reynoldsia seeds (Figure 3.5). For this species, rodents may be more attracted to fruits and could potentially benefit the tree by legitimately dispersing seeds. In general, fruit pulp appears to contribute to higher overall removal rates, but further tests with a wider range ofHawaiian dry forest species is necessary before more definitive conclusions can be made about the effects ofpulp on seed removal or predation.

As very few fleshy-fruited dry forest species are dispersed to exposed sites

(Chapter 2), differences in removal rates under trees and in exposed sites may be unimportant to the fitness and survival ofthese plants. Even ifdispersal to exposed sites did allow seeds to escape predation, the harsher microsite conditions would likely limit seedling recruitment and survival (Kitajima and Fenner 2001). In addition, the species of

65 rodent removing seeds is probably also unimportant, as the non-native black rat (Rattus rattus) and mouse (Mus musculus) recorded in KNAR (Medeiros et al. 1993), and possibly the Polynesian rat (Rattus exulans), are all known to be important, generalist seed predators on a range ofseed sizes. No trapping was conducted during this study, so it is unknown which rodent species is responsible for the seed removal levels reported here. Differences in removal rates from some accessible treatments may be due to the reluctance ofrodents to enter these artificial environments, as other studies have also reported (Inglis et al. 1996; Moles and Drake 1999). The high rates ofremoval recorded for all rodent-accessible treatments ofSantalum ellipticum (Figure 3.5) suggest that rodents will enter artificial environments to attain highly desirable seeds ofcertain species. Some ofthe removal ofseeds and fruits from all trials may be due to the foraging ofnon-native game birds, including the black (Francolinus francolinus) and gray (F pondicerianus) francolin, the ring-necked pheasant (Phasianus colchicus) and the common peafowl (Pavo cristatus), and removal ofseeds or fruits could therefore result in both dispersal and predation. Nevertheless, presence ofteeth marks on seeds and gnawed husks suggest that rodents were also responsible for at least some ofthe seed and fruit removal.

Seed predation ofparticular Hawaiian dry forest tree species appears to be one of the many important factors contributing to lack ofseedling recruitment ofthese species.

Seed predators such as rodents sometimes benefit seeds by moving them to sites favorable for seedling establishment (Vander Wall and Longland 2004). Although introduced rodents may have limited beneficial effects ifremoval results in dispersal rather than seed destruction, for many Hawaiian species, the negative impacts probably

66 outweigh any benefits. Howe and Brown (1999) report that seed predators selectively remove individuals ofsome species more than others, thereby giving a competitive edge to less preferred species. For Hawaiian plants that evolved in the absence ofrodent predation, such as the heavily depredated Santalum ellipticum, this is particularly relevant, but for other species experiencing moderate levels ofpredation, the effects may be negligible. For invasive plants such as Bocconiafrutescens that evolved with rodent predators, seed removal and predation do not appear to be preventing the continued spread ofthis species into native forest habitat. As seed predation can be a major constraint for tree regeneration in neotropical forests (Guariguata and Pinard 1998), it would be valuable to know to what degree predation limits recruitment ofeach species.

Future conservation and restoration efforts in degraded forest ecosystems ofthe Hawaiian

Islands and elsewhere could therefore begin to address and attempt to alleviate the impacts ofseed predation on those species identified as particularly vulnerable.

67 CHAPTER 4: CONCLUSION

Tropical dry forests throughout the world have been severely reduced in extent and continue to be lost at an alarming rate. Several factors have contributed to this decline, mostly related to anthropogenic alteration ofecosystems. Although the direct and indirect impacts ofhumans have greatly modified the vegetation ofHawaiian dry forests, native tree diversity and abundance remain relatively high. Ungulate exclusion and other management strategies may address some ofthe problems associated with lack ofnative tree reproduction, but other factors will continue to play an important role in the future composition ofthese areas. Native dry forest trees are extremely important seed sources and microsites for both natural recruitment and propagation programs, and non-native trees are centers ofcontinued invasion and habitat modification. Understanding the dynamic interaction oftrees with potential dispersers and seed predators not only allows for comparisons with and evaluations ofmore general theories ofseed ecology, but also helps to guide management decisions aimed at preserving and restoring the remaining dry forest ecosystems.

In this study, I attempted to address how seed dispersal patterns and seed predation by rodents interact with tree distribution in a Hawaiian dry forest. As the native avifauna has been entirely extirpated and replaced by an assemblage ofgeneralist frugivores, granivores and game birds, I expected that the distribution ofseeds throughout the area and the types ofseeds being dispersed would be influenced by the types ofbirds currently found at the site. To gain insights into the processes ofseed dispersal, I addressed the following hypotheses:

68 1) There is no difference in the seed rain offleshy-fruited, bird-dispersed seeds

under trees versus in exposed areas.

2) Non-native birds such as the Japanese white-eye (Zosteropsjaponicus), the

common mynah (Acridotheres tristis), the zebra dove (Geopelia striata) the

northern cardinal (Cardinalis cardinalis) and the northern mockingbird (Mimus

polyglottos) are not dispersing the seeds offleshy-fruited native and non-native

trees.

3) There is no relationship between seed size and amount ofseeds dispersed for

fleshy-fruited species.

In all three cases involving seed dispersal, I rejected the null hypotheses. For hypothesis

1, seed rain ofbird-dispersed seeds was significantly higher under all tree species versus exposed sites. In particular, by foraging, perching and nesting in trees, birds are directing the dispersal ofcertain species under their crowns. In exposed sites, bird-dispersed seeds were almost entirely absent, and only seeds not dependent on animal vectors for dispersal, such as wind-adapted grasses, were reaching these areas. For hypothesis 2, indirect evidence from the seed rain and direct observations offoraging on fruiting trees indicated that at least three species ofnon-native birds were dispersing the seeds ofsome native and non-native plants. Among these bird species, the Japanese white-eye was by far the most common at the site. Northern cardinals and northern mockingbirds were also frequently observed, but at much lower numbers. The quality offrugivory differs among the bird species, with white-eyes swallowing entire fruits and seeds ofsome tree species, and pecking at the pulp ofothers. They are therefore acting as both legitimate dispersers

69 and as pulp predators. Mockingbirds and cardinals were legitimate dispersers ofcertain species, but cardinals were also pre-dispersal seed predators ofSantalum ellipticum and possibly other trees. Furthermore, although a larger number of native species compared to non-natives were collected in the bird-dispersed seed rain, non-natives far outnumbered natives in total seed numbers and densities. In particular, Bocconia frutescens and Lantana camara were the most abundant and widespread in all seed traps under trees. For hypothesis 3, there was a significant negative relationship between seed width and numbers ofseeds dispersed by birds. This may be due to limitations on body and gape size ofthe three most common frugivores observed at the study site.

The Hawaiian Islands, which lacked native rodents, now have four species that are widely distributed in both modified and native ecosystems. As rodents are well­ documented seed predators, I expected that rodent seed predation at KNAR would be an important factor influencing seedling recruitment and future tree distribution. To gain insights into the processes and potential effects ofseed predation, I addressed the following null hypotheses:

4) Rodents are not removing and destroying seeds ofdry forest species.

5) There is no difference in removal rates ofseeds under trees versus exposed

sites.

6) There is no relationship between seed size and removal rates for dry forest tree

species.

I rejected null hypothesis 4 for certain dry forest species, as rodents were removing and destroying the seeds offour study tree species. In particular, direct removal trials and presence ofseed husks indicated that rodents were removing and destroying seeds of

70 Bocconiafrutescens, Diospyros sandwicensis, Pleomele auwahiensis, and Santalum ellipticum. Ofthese tree species, S. ellipticum seeds appeared to suffer the highest rates of removal and destruction. I did not reject the null hypothesis for Reynoldsia sandwicensis seeds, as rodents removed no seeds during the entire duration ofthe study. I rejected null hypothesis 5 for D. sandwicensis, P. auwahiensis seeds and S. ellipticum, as removal rates for these species were significantly higher under trees versus exposed sites. I did not reject the null hypothesis for Bocconiafrutescens, P. auwahiensis fruits and R. sandwicensis, as there were no significant differences in removal rates under trees versus exposed sites. I did not reject null hypothesis 6, as no significant relationship was found between seed size and removal rates for the dry forest species used in this study. As only five species were used for the size analysis, however, small sample size may account for the lack ofany relationship. Further studies using more species with a broader range of sizes will better detect whether or not a relationship between seed size and predation does exist for dry forest trees.

In conclusion, current dispersal patterns indicate that a few readily disseminated non-native species are being spread throughout KNAR. For fleshy-fruited species, non­ native frugivores are responsible for the dispersal ofpredominantly non-native invasive species under standing trees, while the majority ofnative seeds are falling directly onto the ground without the benefits ofscarification or dispersal. This pattern ofseed rain, the distribution ofseedlings under trees, and the high levels ofpredation on native seeds suggest that, without management intervention, the diverse, native dry forest community that currently exists will eventually be replaced by a homogenous landscape ofa few prolific invaders. For any semblance ofa functioning native Hawaiian dry forest

71 ecosystem to persist into the future, immediate threats must be addressed by removing non-native species dispersed by birds and by eliminating rodents depredating native seeds. The critical issue will then be whether extinct birds can be replaced by truly effective surrogates capable ofscarifying and disseminating the wide range ofnative species still extant in the remaining dry forests, or whether the conservation ofthese species has become entirely dependent on human intervention.

72 APPENDIX A. TREE DIMENSIONS (BASAL DIAMETER, HEIGHT, CROWN

DIAMETER) OF STUDY SPECIES AND DEAD TREES

Table A. 1. Tree dimensions ofstudy species used in sampling offleshy-fruited seeds and seed removal trials. Values are means (± 1 SE) for 15 ofeach species and 15 dead trees. Basal diameter Crown diameter Tree species Height (m) (cm) (m) Santalum ellipticum 17.62 (1.34) 2.76 (0.19) 3.70 (0.23)

Diospyros sandwicensis 22.02 (1.38) 4.72 (0.39) 4.41 (0.22)

Dead Trees 22.80 (2.48) 3.77 (0.29) 3.00 (0.23)

Bocconiafrutescens 27.30 (1.94) 3.21 (0.12) 3.27 (0.12)

Pleomele auwahiensis 33.32 (1.63) 5.30 (0.22) 3.23 (0.17)

Reynoldsia sandwicensis 38.28 (2.43) 5.77 (0.32) 5.73 (0.34)

73 .-36 !~ . ,...-- 25 . ,...-- 1"20 )" 15 10 5 A AB AB BC D o CDI 7

6 5

----!-

2

1 A BC AB AC C o 7 6

--..l..- -=--

1 o ,AB 81 A A AC san ell DIo lin Dead Boc fru PIe auw Rey san

Figure A.I. Tree diIIlensions for study speeies and used in sampUng ofOeshy-fruited seeds. Bars represent means +1 SE (n = 15 per tree). Tree dimensions were compared with a one-way ANOVA. Letters above the bars show the results ofthe Tukey test. Treatments sharing the same letter are not significantly different (p > 0.05).

74 APPENDIX B. PHENOLOGY OF STUDY TREES AND COMMON FLESHY­

FRUITED SHRUBS OF THE KANAIO NATURAL AREA RESERVE

METHODS

The presence ofimmature and ripe fruits on sample trees and other taxa in the area was recorded on a monthly basis from March 2003 to February 2004 to document the potential pool ofbird-dispersed seeds that could be disseminated in the area. For the five fleshy-fruited tree species used in the seed dispersal study, percentages ofstems with immature and ripe fruit were estimated by categories (0 = none; 1 = 1-5 percent; 2 = 6-25 percent; 3 = 26-50 percent; 4 = 51-75 percent; 5 = 76-100 percent) on 15 to 30 trees per species occurring along seed sampling transects. Means for trees were calculated using the following values for each phenology category: 0 = 0; I = 3 percent; 2 = 15 percent; 3

= 37.5 percent; 4 = 62.5 percent; 5 = 87.5 percent (Tables BI-B3; Figures B.1. and B.3.).

For three widespread, fleshy-fruited shrubs, the non-native Lantana camara, and the natives Osteomeles anthyllidifolia and Wikstroemia monticola, 25 plants ofeach species were selected along the seed-sampling transects such that no conspecific plants were within 25 meter ofeach other. For each plant, a 50-cm branch segment was tagged and numbers ofimmature and ripe fruits counted along that length for each month (Table BA;

Figures B.2. and BA.). Phenology for these plants was recorded as they are common throughout the reserve and possess relatively small, fleshy fruits with characteristics attractive to either obligate or opportunistic frugivores (Howe 1986).

75 Table B.1. MontWy percentage offleshy-fruited trees and shrubs with mature fruit. Boc fro: Bocconiafrutescens (n = 28); Dio san: Diospyros sandwicensis (n = 25); PIe auw: Pleomele auwahiensis (n = 30); Rey san: Reynoldsia sandwicensis (n = 15); San ell: Santalum ellipticum (n = 28); Lan cam: Lantana camara (n = 25); Ost ant: Osteomeles anthyllidifolia (n = 25); Wik mon: Wikstroemia monticola (n = 25).

Date Boc fro Dio san PIe auw Rey san San ell Lan cam Ost ant Wik mon Mar-03 0.00 72.00 63.33 40.00 21.43 20.00 0.00 0.00 Apr-03 0.00 72.00 60.00 0.00 28.57 84.00 4.00 4.00 May-03 96.43 64.00 43.33 0.00 17.86 92.00 0.00 24.00 Jun-03 100.00 36.00 13.33 0.00 3.57 80.00 4.00 0.00 Jul-03 100.00 16.00 0.00 0.00 7.14 12.00 8.00 0.00 Aug-03 53.57 52.00 6.67 0.00 3.57 48.00 20.00 8.00 Sep-03 0.00 64.00 6.67 0.00 3.57 4.00 52.00 0.00 Oct-03 0.00 96.00 6.67 0.00 3.57 0.00 64.00 0.00 Nov-03 0.00 100.00 3.33 0.00 0.00 0.00 60.00 0.00 Dec-03 0.00 92.00 13.33 46.67 10.71 0.00 28.00 0.00 Jan-04 0.00 80.00 13.33 80.00 0.00 12.00 0.00 0.00 Feb-04 0.00 80.00 13.33 20.00 42.86 84.00 0.00 32.00

Table B.2. Mean (± 1 SE) monthly percentage ofstems with immature fruits by tree species. Means were calculated by assigning values to phenology categories 1-5 (0 = none; 1 = 2.5 percent; 2 = 15.5 percent; 3 = 37.5 percent; 4 = 62.5 percent; 5 = 87.5 percent). Boc fro: Bocconiafrutescens (n = 28); Dio san: Diospyros sandwicensis (n = 25); PIe auw: Pleomele auwahiensis (n = 30); Rey san: Reynoldsia sandwicensis (n = 15); San ell: Santalum ellipticum (n = 28).

Date Boc fro Dio san Pleauw Reysan San ell Mar-03 53.95 (4.50) 26.82 (6.12) 0.00 (0.00) 0.20 (0.20) 2.73 (1.41) Apr-03 71.88 (4.20) 12.58 (3.32) 0.00 (0.00) 0.00 (0.00) 1.71 (0.57) May-03 77.68 (2.68) 11.48 (3.68) 0.00 (0.00) 0.00 (0.00) 2.30 (1.33) Jun-03 59.02 (2.42) 25.28 (4.55) 0.10 (0.10) 0.00 (0.00) 2.36 (0.88) Jul-03 24.30 (2.58) 47.00 (4.83) 2.08 (2.08) 0.00 (0.00) 4.39 (1.92) Aug-03 2.04 (0.56) 55.90 (4.75) 0.00 (0.00) 0.00 (0.00) 1.18 (0.56) Sep-03 0.00 (0.00) 67.60 (4.28) 0.00 (0.00) 0.00 (0.00) 1.29 (0.57) Oct-03 0.00 (0.00) 67.60 (4.28) 0.00 (0.00) 3.70 (2.61) 1.39 (0.57) Nov-03 0.00 (0.00) 62.80 (4.88) 0.00 (0.00) 17.60 (6.97) 1.93 (0.89) Dec-03 0.00 (0.00) 52.52 (4.80) 0.00 (0.00) 22.80 (6.41) 7.41 (3.20) Jan-04 0.00 (0.00) 36.18 (5.07) 0.00 (0.00) 15.93 (5.89) 10.95 (4.71) Feb-04 0.00 (0.00) 20.68 (3.53) 0.00 (0.00) 2.20 (1.36) 15.54 (5.13)

76 Table B.3. Mean (± 1 SE) monthly percentage ofstems with mature fruits by tree species. Means were calculated by assigning values to phenology categories 1-5 (0 = none; 1 = 2.5 percent; 2 = 15.5 percent; 3 = 37.5 percent; 4 = 62.5 percent; 5 = 87.5 percent). Boc fru: Bocconiafrutescens (n = 28); Dio san: Diospyros sandwicensis (n = 25); PIe auw: Pleomele auwahiensis (n = 30); Rey san: Reynoldsia sandwicensis (n = 15); San ell: Santalum ellipticum (n = 28).

Date Boc fru Dio san PIe auw Rey san San ell Mar-03 0.00 (0.00) 4.02 (1.51) 24.57 (5.83) 1.20 (0.39) 0.64 (0.24) Apr-03 0.00 (0.00) 3.60 (0.90) 16.50 (4.57) 0.00 (0.00) 0.86 (0.26) May-03 8.36 (1.85) 1.92 (0.29) 4.88 (2.19) 0.00 (0.00) 0.54 (0.22) Joo-03 25.34 (2.62) 1.56 (0.63) 0.40 (0.19) 0.00 (0.00) 0.11 (0.11) Ju1-03 11.89 (1.75) 0.48 (0.22) 0.00 (0.00) 0.00 (0.00) 0.21 (0.15) Aug-03 1.61 (0.29) 2.52 (0.81) 2.18 (2.08) 0.00 (0.00) 0.11 (0.11) Sep-03 0.00 (0.00) 1.92 (0.29) 0.20 (0.14) 0.00 (0.00) 0.11 (0.11) Oct-03 0.00 (0.00) 3.36 (0.50) 0.20 (0.14) 0.00 (0.00) 0.11 (0.11) Nov-03 0.00 (0.00) 11.88 (2.52) 1.25 (1.25) 0.00 (0.00) 0.00 (0.00) Dec-03 0.00 (0.00) 18.94 (3.14) 4.77 (2.91) 5.30 (2.65) 0.75 (0.55) Jan-04 0.00 (0.00) 8.04 (1.27) 2.78 (2.12) 7.10 (2.58) 0.00 (0.00) Feb-04 0.00 (0.00) 7.14 (1.74) 2.70 (1.73) 0.60 (0.32) 1.71 (0.57)

Table B.4. Mean (± 1 SE) monthly number ofimmature fruits (IF) and mature fruits (MF) on branches by shrub species. LC: Lantana camara (n = 25); OA: Osteomeles anthyllidifolia (n = 25); WM: Wikstroemia monticola (n = 25). LC LC OA OA WM WM Date IF MF IF MF IF MF Mar-03 4.32(0.71) 0.28(0.12) 3.64(0.90) 0.00(0.00) 1.80 (0.92) 0.00 (0.00) Apr-03 1.40 (0.31) 2.04 (0.39) 7.36 (1.39) 0.20 (0.20) 1.24 (0.68) 0.40 (0.40) May-03 1.68(0.56) 3.80(0.67) 8.88(1.68) 0.00(0.00) 0.16 (0.16) 0.48 (0.22) Joo-03 0.52 (0.25) 3.36 (0.68) 8.96(1.74) 0.04(0.04) 0.00 (0.00) 0.00 (0.00) Ju1-03 0.80 (0.32) 0.24 (0.15) 8.24 (1.56) 0.08 (0.06) 0.00 (0.00) 0.00 (0.00) Aug-03 0.04 (0.04) 0.88 (0.22) 6.48(1.11) 0.40(0.19) 0.44 (0.28) 0.08 (0.06) Sep-03 0.00(0.00) 0.04(0.04) 5.32(0.97) 1.24(0.35) 0.00 (0.00) 0.00 (0.00) Oct-03 0.00(0.00) 0.00(0.00) 2.08 (0.86) 1.76(0.68) 0.00 (0.00) 0.00 (0.00) Nov-03 0.00(0.00) 0.00(0.00) 0.20(0.16) 2.68(0.73) 0.00 (0.00) 0.00 (0.00) Dec-03 0.76(0.30) 0.00(0.00) 0.00 (0.00) 0.80 (0.35) 1.44 (0.91) 0.00 (0.00) Jan-04 4.40(0.75) 0.16(0.09) 0.00 (0.00) 0.00 (0.00) 4.52 (1.63) 0.00 (0.00) Feb-04 2.00(0.53) 6.20(0.98) 0.00(0.00) 0.00(0.00) 1.08 (0.36) 3.16 (1.23)

77 100% )J""...G"" I ~ '\. 80% I !t"-{] c: .2 o--q.., I I' -Soc fru J \ -ns 60% • - / I --0- Dio san ::J -. c. \ \ 0 \ J -.- Pie auw c. ~. . \- , \ -0- Rey san ..... 40% '\ , I . 0 , q, I ~ , • .{ ...... San ell 0 \. \ I I 20% rtI", •..... ', \1 \ ~ ~

0% M A M JJ AS 0 N D J F Month

Figure B.I Monthly percentage offleshy-fruited trees with mature fruit. Phenology recorded from March 2003 to February 2004. Boc fro: Bocconiafrutescens (n = 28); Dio san: Diospyros sandwicensis (n = 25); PIe auw: Pleornele auwahiensis (n = 30); Rey san: Reynoldsia sandwicensis (n = 15); San ell: Santalurn ellipticurn (n = 28).

78 100%

&:: 80% 0 :.;:; J!! 60% -Lan cam :::J / -"\ a. I 0 \ - -Ost ant I \ ~40% - - -Wik mon 0 I \

:::R0 ~ \ 20% I , \ I ...... 0% M A M JJ A S 0 N 0 J F Month

Figure B.2. Monthly percentage offleshy-fruited shrubs with mature fruit. Phenology recorded from March 2003 to February 2004. Lan cam: Lantana camara (n = 25); 08t ant: Osteomeles anthyllidifolia (n = 25); Wik mon: Wikstroemia monticola (n = 25).

79 100 A 100 B 75 - 75 - 50 r 50 2~ ~~:'-r-'~~'_~. 25 -1--,-, .. ".,..,-.,...,-.,.....,-'-,-',--",_ O~:;::::=:r::=r-=r-""'F"':;==;::::::="'--.---'---:'" MAMJJASONDJF MAMJJASONDJF

50 50 C D P 40 40 p 30 30 '- - 1 ~ 20 20 - ;:!2. 10 10 - ,/:I~ 0 I ~+~ "'f J ~, 0 0 ""r , , i I I M AM J J ASONDJ F MAMJ J ASONDJ F

50 E 40 30 20 10 .;···f··· ~ ~ ~. o +"=F!'i¥-,.....,..,::' ". • .. .;...... ~,"':.....,...:.,...... w,:..;'-'.,,:.i.:T,~,0). .... "" - .. --,.,• --T'=-' MAMJJASONDJF

Month

Figure B.3. Mean (± 1 SE) monthly percentage ofstems with immature and mature fruits by tree species. Means were calculated by assigning values to phenology categories 1-5 (0 = none; 1 = 2.5 percent; 2 = 15.5 percent; 3 = 37.5 percent; 4 = 62.5 percent; 5 = 87.5 percent). Dotted lines represent immature fruits and solid lines represent mature fruits. A: Bocconiafrutescens (n = 28); B: Diospyros sandwicensis (n = 25); C: Pleomele auwahiensis (n = 30); D: Reynoldsia sandwicensis (n = 15); E: Santalum ellipticum (n = 28). Note that Bocconia and Diospyros (A & B) are scaled 0-100, and Pleomele, Reynoldsia and Santalum (C-E) are scaled 0-50. Phenology recorded from March 2003 to February 2004.

80 10 10 B...... 1 8 - .... .", 8 ;' " . "', [ 6 6 . " l'"j [.. 4 4 2 2 •• r~ (j) 0 :!::::: MAM JJ A SON D J F MAM J J A SON D J F ::J lo- ;10 Ie 8 .., I o'I' •

.,>. .,-1'" II MAM J J A SON D J F Month

Figure B.4. Mean (± 1 SE) monthly number ofimmature and mature fruits on branches by shrub species. Dotted lines represent immature fruits and solid lines represent mature fruits. A: Lantana camara (n = 25); B: Osteomeles anthyllidifolia (n = 25); C: Wikstroemia monticola (n = 25). Phenology recorded from March 2003 to February 2004.

81 APPENDIX C. RELATIVE FREQUENCY AND ABUNDANCE OF NON-NATIVE

BIRDS IN KNAR STUDY SITE

METHODS

To assess the relative frequency and abundance ofbird species at the study site in

KNAR, sampling was conducted at 10 stations along a north-south oriented transect rougWy bisecting the seed sampling transects and extending from approximately 650-830 meters elevation. Sampling stations were placed at 134-m intervals, following protocols described by Scott et al. (1986). Each station was surveyed montWy on days ofgood weather (i.e. no heavy rain or wind) from April 2003 to March 2004. Surveys began shortly after sunrise and continued until all 10 stations were sampled. Estimates ofbird numbers at each station were made using the variable circular-plot method (Reynolds et al. 1980). During an eight-minute count period at each station, each bird species both heard and seen was recorded, and horizontal distances to each bird were estimated.

Because these counts were intended to give rough estimates ofbird abundance, and were not meant to derive estimates ofdensities for all birds in the dry forest habitat, data for each species from all 10 stations per month were combined, and mean numbers ofbirds ofeach species per month were calculated for the 12-month duration ofthe sampling.

The frequency ofeach bird species was calculated by recording the presence ofa species at each ofthe ten stations for the 12-month duration ofthe sampling, and dividing this number by 120. The relative frequency and abundance ofeach species was then calculated from these totals (Table C.1; Figures C.1-Co2). For three important non-native

82 frugivores, the Japanese white-eye (Zosterops japonicus), the northern cardinal

(Cardinalis cardinalis), and the northern mockingbird (Mimus polyglottos), mean number ofbirds at the 10 sampling stations was also calculated for each sampling month, as these species have been observed consuming fruits ofsome ofthe study species (Figure C.3).

Table C.l. Relative frequency and abundance ofbird species recorded at 10 sampling stations over a 12-month period (April 2003 - March 2004) in KNAR. ReI. ReI. Species Common Name Family Freq Abund. Zosteropsjaponicus Japanese white-eye Zosteropidae 27.0 57.2 Cardinalis cardinalis northern cardinal Emberizidae 16.7 9.2 Mimus polyglottos northern mockingbird Mimidae 12.4 7.1 Carpodacus mexicanus housefinch Fringillidae 11.3 8.3 Geopelia striata zebra dove Columbidae 8.1 4.0 Francolinusfrancolinus black francolin Phasiandidae 6.8 3.4 Streptopelia chinensis spotted dove Columbidae 5.6 2.8 Francolinus pondicerianus gray francolin Phasiandidae 5.2 2.4 Acridotheres tristis common myna Sturnidae 3.8 2.4 Lonchurapunctulata nutmeg mannikin Estrildidae 1.8 2.7 Phasianus colchicus ring-necked pheasant Phasiandidae 0.9 0.4 Cettia diphone Japanese bush-warbler Muscicapidae 0.5 0.2 Total 100 100

83 30%

25% >- c0 Q) 20% ~ tT ~ 15% Q) -.~ 10 10% ~ 5%

0%

Figure C.l. Relative frequency' ofbird speeies reeorded at 10 sampling stations over a 12-month period (April 2003 - Mareh 2004) in KNAR.

84 70% 60% (1) c:0 m 50% 'Uc: :::J 40% .Qm (1) 30% .~ 10 20% ~ 10%

0%

Figure Co2. Relative abundance ofbird species recorded at 10 sampling stations over a 12-month period (April 2003 - March 2004) in KNAR.

85 Figure C.3. Mean (± SE) number ofthree important non-native frugivores at 10 sampling stations over a 12-month period (April 2003 - March 2004) in KNAR.

86 APPENDIX D. CHARACTERIZATION OF THE VEGETATION OF KNAR

STUDY SITE

METHODS

To supplement the vegetation descriptions provided by Medeiros et at. (1986;

1993), plant cover in the study site was sampled at 1000 points using the point-intercept method (Bonham 1989). Samples were taken at a random point in each 2-m segment of ten 200-m transects placed midway between and parallel to the seed dispersal transects.

Points recorded under two meters height were classified as ground cover, and points over two meters height were classified as canopy cover. For each species, percent cover for ground and canopy was calculated separately by counting the number ofpoints at which a vertically-projected line intercepted living plant material and dividing this value by 1000.

More than one different species could be sampled per point, but no single species was counted more than once per point.

87 Table D.l. Abundance (absolute, % cover and relative cover) ofground vegetation « 2- m height) at KNAR study site. Data presented in order ofdecreasing abundance. Absolute ReI. Category Form Status % Cover abundance Cover ROCK NA NA 358 35.8 30.4 Lantana camara Shrub Non-native 263 26.3 22.3 Melinis repens Graminoid Non-native 128 12.8 10.9 Bidens pilosa Forb Non-native 115 11.5 9.8 Dodonaea viscosa Shrub Native 61 6.1 5.2 Ipomoea indica Liana Native 54 5.4 4.6 Osteomeles anthyllidifolia Shrub Native 47 4.7 4.0 Melinis minutiflora Graminoid Non-native 44 4.4 3.7 LITTER NA NA 15 1.5 1.3 Wikstroemia monticola Shrub Native 11 1.1 0.9 Chamaecrista nictitans Forb Non-native 11 1.1 0.9 Santalum ellipticum Tree Native 6 0.6 0.5 Petroselinum crispum Forb Non-native 6 0.6 0.5 Emiliafosbergii Forb Non-native 5 0.5 0.4 Pennisetum clandestinum Graminoid Non-native 5 0.5 0.4 Plectranthus parviflorus Forb Native 4 0.4 0.3 Ageratina adenophora Forb Non-native 4 0.4 0.3 Neonotonia wightii Vine Non-native 3 0.3 0.3 Galinsoga parviflora Forb Non-native 3 0.3 0.3 Bocconiafrutescens Tree Non-native 2 0.2 0.2 Cocculus orbiculatus Vine Native 2 0.2 0.2 Leptecophylla tameiameiae Shrub Native 2 0.2 0.2 Carex wahuensis Graminoid Native 2 0.2 0.2 Passiflora subpeltata Vine Non-native 2 0.2 0.2 Doryopteris decipiens Fern Native 2 0.2 0.2 Sonchus oleraceus Forb Non-native 2 0.2 0.2 Leucaena leucocephala Shrub/Tree Non-native 2 0.2 0.2 Salvia coccinea Forb Non-native 2 0.2 0.2 Cyperus hillebrandii Graminoid Native 1 0.1 0.1 Senecio madagascariensis Forb Non-native 1 0.1 0.1 Ageratina riparia Forb Non-native 1 0.1 0.1 Pellaea ternifolia Fern Native 1 0.1 0.1 Anagallis arvensis Forb Non-native 1 0.1 0.1 Nephrolepis multiflora Fern Non-native 1 0.1 0.1 Alyxia oliviformis Vine Native 1 0.1 0.1 Schinus terebinthifolius Tree Non-native 1 0.1 0.1 Lepisorus thunbergianus Fern Native 1 0.1 0.1 Chenopodium oahuense Shrub Native 1 0.1 0.1 Metrosideros polymorpha Tree Native 1 0.1 0.1 88 Table D.1 (Continued) Abundance (absolute, % cover and relative cover) ofground vegetation « 2-rn height) at KNAR study site. Absolute iY< C Category Form Status o over ReI. Cover abundance Solanum americanum Forb Non-native 1 0.1 0.1 Pityogramma austroamericanaFern Non-native 1 0.1 0.1 Peperomia blanda Forb Native 1 0.1 0.1 Indigofera suffruticosa Vine Non-native 1 0.1 0.1 Portulaca pilosa Forb Non-native 1 0.1 0.1 TOTAL 1177 117.7 100.0

Table D.2. Abundance (absolute, and relative cover) ofcanopy vegetation (> 2-rn height) at KNAR study site. Data presented in order ofdecreasing abundance.

Category Form Status Absolute ReI.Cover

OPEN NA NA 844 84.4 Pleomele auwahiensis Tree Native 48 4.8 Bocconiafrutescens Tree Non-native 28 2.8 Diospyros sandwicensis Tree Native 26 2.6 Dodonaea viscosa Shrub Native 10 1.0 Santalum ellipticum Tree Native 10 1.0 Nothocestrum latifolium Tree Native 9 0.9 Wikstroemia monticola Shrub Native 6 0.6 Xylosma hawaiiense Tree Native 6 0.6 DEAD TREES Tree NA 4 0.4 Reynoldsia sandwicensis Tree Native 4 0.4 Myoporum sandwicense Tree Native 2 0.2 Opuntiaficus-indica Shrub Non-native 1 0.1 Myrsine lanaiensis Tree Native 1 0.1 Nestegis sandwicensis Tree Native 1 0.1 Total 1000 100.0

89 Table D.3. Abundance (absolute and relative cover) ofnative and non-native vegetation, dead trees, bare ground, and open sky at KNAR study site. ReI. ReI. Canopy Category Ground Ground Canopy Cover Cover Native Plants 198 16.8 123 12.3 Non-native Plants 606 51.5 29 2.9 Dead trees o 0.0 4 0.4 Bare Ground 373 31.7 NA NA Open NA NA 844 84.4 Total 1177 100.0 1000 100.0

90 LITERATURE CITED

Aide, T.M. and J. Cavelier. 1994. Barriers to lowland tropical forest restoration in the

Sierra Nevada de Santa Marta, Colombia. Restoration Ecology 2: 219-229.

Aide, T.M., J.K. Zimmerman, J.B. Pascarella, L. Rivera, and H. Marcano-Vega. 2000.

Forest regeneration in a chronosequence oftropical abandoned pastures:

implications for restoration ecology. Restoration Ecology 8: 328-338.

Allen, R.B., W.G. Lee, and B.D. Rance. 1994. Regeneration in indigenous forest after

eradication ofNorway rats, Breaksea Island, New Zealand. New Zealand Journal

ofBotany 32: 429-439.

Athens, J.S., H.D. Tuggle, J.V. Ward and D.J. Welch. 2002. Avifaunal extinctions,

vegetation change, and Polynesian impacts in prehistoric Hawai'i. Archaeology in

Oceania 37: 57-78.

Atkinson, LA.E. 1977. A reassessment offactors, particularly Rattus L., that influenced

the decline ofendemic forest birds in the Hawaiian Islands. Pacific Science 31:

109-133.

Augspurger, C.K. 1984. Seedling survival oftropical tree species: interactions of

dispersal distance, light-gaps, and pathogens. Ecology 65: 1705-1712.

91 Baker, J.L. and S. Allen. 1978. Roofrat depredations on Hibiscadelphus (Malvaceae)

trees. Pages 2-5 in C.W. Smith (ed.). Proceedings, second conference in natural

sciences, Hawai'i Volcanoes National Park. Cooperative National Park Resources

Studies Unit, University ofHawai'i, Honolulu.

Blackmore, M. and P.M. Vitousek. 2000. Cattle grazing, forest loss, and fuel loading in a

dry forest ecosystem at Pu'u Wa'aWa'a Ranch, Hawai'i. Biotropica 32: 625-632.

Bleher, B. and K. Bohning-Gaese. 2001. Consequences offrugivore diversity for seed

dispersal, seedling establishment and the spatial pattern ofseedlings and trees.

Oecologia 129: 385-394.

Bonham, C. D. 1989. Measurements for terrestrial vegetation, Wiley & Sons, NY.

Burney, D.A., H.F. James, L.P. Burney, S.L. Olson, W. Kikuchi, W.L. Wagner, M.

Burney, D. McCloskey, D. Kikuchi, F.V. Grady, R. Gage II, and R. Nishek. 200l.

Fossil evidence for a diverse biota from Kauai and its transformation since human

arrival. Ecological Monographs 71: 615-641.

Bruegmann, M.M. 1996. Hawai'i's dry forests. Endangered Species Bulletin 11: 26-27.

92 Bullock, S., H.A. Mooney, and E. Medina. 1995. Seasonally Dry Tropical Forests.

Cambridge University Press, Cambridge, MA. 448 pp.

Bustamante, RO and J.A. Simonetti. 2000. Seed predation and seedling recruitment in

plants: the effect ofthe distance between parents. Plant Ecology 147: 173-183.

Cabin, RJ., S.G. Weller, D.H. Lorence, T.W. Flynn, A.K. Sakai, D. Sandquist, and L.

Hadway. 2000. Effects oflong-term ungulate exclusion and recent Non-native

species control on the preservation and restoration ofa Hawaiian tropical dry

forest. Conservation Biology 14: 439-453.

Callaway, RM. 1992. Effect ofshrubs on recruitment ofQuercus douglasii and Quercus

lobata in California. Ecology 73: 2118-2128.

Campbell, D.J. and I.A.E. Atkinson. 2002. Depression oftree recruitment by the Pacific

rat (Rattus exulans Peale) on New Zealand's northern offshore islands. Biological

Conservation 107: 19-35.

Campbell, D.J., H. Moller, G.W. Ramsay, and J.C. Watt. 1984. Observations on foods of

kiore (Rattus exulans) found in husking stations on northern offshore islands of

New Zealand. New Zealand Journal o/Ecology 7: 131-138.

93 Carlo, T.A., J.A. Collazo, and M.J. Groom. 2003. Avian fruit preference across a Puerto

Rican forested landscape: pattern consistency and implications for seed removal.

Oecologia 134: 119-131.

Carlquist, S. 1980. Hawai'i: a natural history. National Tropical Botanical Garden,

Lawai, Hawai'i. 468 pp.

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 ofdispersed and non­

dispersed seeds ofsix African tree species. Journal ojTropical Ecology 12: 491­

504.

Clark, C.J., J.R. Paulsen, E.F. Connor and V.T. Parker. 2004. Fruiting trees as dispersal

foci in a semi-deciduous tropical forest. Oecologia 139: 66-75.

Clark, D.A. 1982. Foraging behavior ofa vertebrate omnivore (Rattus rattus): meal

structure, sampling and diet breadth. Ecology 63: 763-772.

Cole, F.R., L.L. Loope, A.C. Medeiros, C.E. Howe, and L.J. Anderson. 2000. Food habits

ofintroduced rodents in high elevation shrubland ofHaleakala National Park,

Maui, Hawai'i. Pacific Science 54: 313-329.

94 Cole, F.R., L.L. Loope, A.C. Medeiros, J.A. Raikes, and C.S. Wood. 1995. Conservation

implications ofintroduced game birds in high-elevation Hawaiian shrubland.

Conservation Biology 9: 306-313.

Cordeiro, N.J. and H.F. Howe. 2001. Low recruitment f trees dispersed by animals in

African forest fragments. Conservation Biology 15: 1733-1741.

Cordeiro, N.J., D.A.G. Patrick, B. Munisi and V. Gupta. 2004. Role ofdispersal in the

invasion ofan exotic tree in an East African submontane forest. Journal of

Tropical Ecology 20: 449-457.

Cordell, S., R.J. Cabin and L.J. Hadway. 2002. Physiological ecology ofnative and alien

dry forest shrubs in Hawai'i. Biological Invasions 4: 387-396.

Corlett, R.T. 1992. Seed dispersal by birds in Hong Kong shrubland. Memoirs ofthe

Hong Kong Natural History Society 19: 129-130.

Cox, P.A., T. Elmqvist, E.D. Pierson and W.E. Rainey. 1991. Flying foxes as strong

interactors in South Pacific island ecosystems: a conservation hypothesis.

Conservation Biology 5: 448-454.

95 Crandell, D.R. 1983. Potential hazards from future volcanic eruptions on the island of

Maui, Hawai'i. USGS Misc Investigation Series Map 1-1442.

Crawley, M.J. 2001. Seed predators and plant population dynamics. Pages 167-182 in M.

Fenner (ed). Seeds: the ecology ofregeneration in plant communities. CABI,

Wallingford, UK. 410 pp.

Cuddihy, L.W. 1989. Vegetation zones ofthe Hawaiian Islands. Pages 27-37 in C.P.

Stone and D.B. Stone (eds.). Conservation biology in Hawai'i. Cooperative

National Park Resources Studies Unit, University ofHawai'i, Honolulu. 252 pp.

Cuddihy, L.W. and C.P. Stone. 1990. Alteration ofnative Hawaiian vegetation.

Cooperative National Park Resources Studies Unit, University ofHawai'i,

Honolulu.

Dalling, J.W., M. D. Swaine, and N.C. Garwood. 1998. Dispersal patterns and seed bank

dynamics ofpioneer trees in moist tropical forest. Ecology 79: 564-578.

Debussche, M. and P. Isenmann. 1994. Bird-dispersed seed rain and seedling

establishment in patchy Mediterranean vegetation. Orkos 69: 414-426.

Debussche, M., J. Escarre and J. Lepart. 1982. Ornithochory and plant succession in

mediterranean abandoned orchards. Vegetatio 48: 255-266.

96 Delgado Garcia, J.D. 2002. Interaction between introduced rats and frugivore bird-plant

system in a relict island forest. Journal ofNatural History 36: 1247-1258.

DeSteven, D. and F.E. Putz. 1984. Impact ofmammals on early recruitment ofa tropical

canopy tree, Dipteryx panamensis, in Panama. Oikos 43: 207-216.

Drake, D.R. 1998. Relationships among the seed rain, seed bank and vegetation ofa

Hawaiian forest. Journal ofVegetation Science 9: 103-112.

Eriksson, O. and J. Ehrlen. 1992. Seed and microsite limitation ofrecruitment in plant

populations. Oecologia 91: 360-364.

Fall, M.W., A.B. Medina, and W.B. Jackson. 1971. Feeding patterns ofRattus rattus and

Rattus exulans on Eniwetok Atoll, Marshall Islands. Journal ofMammalogy 52:

69-76.

Ferguson, R.N. and D.R. Drake. 1999. Influence ofvegetation structure on spatial

patterns ofseed deposition by birds. New Zealand Journal ofBotany 37: 671-677.

Figueroa, J.A., A.A. Muiioz, J.E. Mella and M.T.K. Arroyo. 2003. Pre- and post-dispersal

seed predation in a Mediterranean-type climate montane sclerophyllous forest in

central Chile. Australian Journal ofBotany 50: 183-195.

97 Fukui, AW. 1995. The role ofthe brown-eared bulbul Hypsypetes amaurotis as a seed

dispersal agent. Researches on Population Ecology 37: 211-218.

Gagne, W.e. and L.W. Cuddihy. 1990. Vegetation. Pp. 45-114 in W.L. Wagner, D.R.

Herbst and S.H. Sohmer (eds.). Manual ofthe flowering plants ofHawai'i. RP.

Bishop Museum and University ofHawai'i Press, Honolulu. 1853 pp.

Giambelluca, T.W., M.A. Nullet, and T.A Schroder. 1986. Rainfall atlas ofHawai'i.

State ofHawai'i, Department ofLand and Natural Resources, Division ofWater

and Land Development. Honolulu. Report R76: 267 pp.

Gryj, E.O. and C.A Dominguez. 1996. Fruit removal and postdispersal survivorship in

the tropical dry forest shrub Erythroxylum havanense - ecological and

evolutionary implications. Oecologia 108: 368-374.

Guariguata, M.R. and M.A Pinard. 1998. Ecological knowledge ofregeneration from

seed in neotropical forest trees: implications for natural forest management.

Forest Ecology andManagement 112: 87-99.

Guevara, S. and J. Laborde. 1993. Monitoring seed dispersal at isolated standing trees in

tropical pastures: consequences for local species availability. Vegetatio 107/1.08:

319-338.

98 Guevara, S., J. Laborde and G. Sanchez-Rios. 2004. Rain forest regeneration beneath the

canopy offig trees isolated in pastures ofLos Tuxtlas, Mexico. Biotropica 36: 99­

108.

Guevara, S., S.E. Purata, and E. Van der Maarel. 1986. The role ofremnant forest trees in

tropical secondary succession. Vegetatio 66: 77-84.

Hadfield, M.G., S.E. Miller, and A.H. Carwile. 1993. The decimation ofendemic

Hawaiian tree snails by Non-native predators. American Zoologist 33: 610-622.

Hammond, D.S. 1995. Post-dispersal seed and seedling mortality oftropical dry forest

trees after shifting agriculture, Chiapas, Mexico. Journal ojTropical Ecology 11:

295-313.

Hampe, A. 2003. Frugivory in European laurel: how extinct seed dispersers have been

substituted. Bird Study 50: 280-284.

Hau, C.H. 1997. Tree seed predation on degraded hillsides in Hong Kong. Forest

Ecology and Management 99: 215-221.

Hay, M.E. and P.J. Fuller. 1981. Seed escape from heteromyid rodents: the importance of

microhabitat and seed preference. Ecology 62: 1395-1399.

99 Herrera, C.M. 2995. Plant-vertebrate seed dispersal systems in the Mediterranean:

ecological, evolutionary, and historical determinants. Annual Review ofEcology

and Systematics 26: 705-727.

Holl, K.D. 2002. Effect ofshrubs on tree seedling establishment in an abandoned tropical

pasture. Journal ofEcology 90: 179-187.

Holl, K.D., M.E. Loik, E.H.V. Lin and LA. Samuels. 2000. Tropical montane forest

restoration in Costa Rica: overcoming barriers to dispersal and establishment.

Restoration Ecology 8: 339-349.

Holl, K.D. and M.E. Lulow. 1997. Effects ofspecies, habitat, and distance from edge on

post-dispersal seed predation in a tropical rainforest. Biotropica 29: 459-468.

Howe, H.F. 1986. Seed dispersal by fruit-eating birds and mammals. Pages 123-190 in

D.R. Murray (ed.). Seed dispersal. Academic Press, Sydney, Australia. 322 pp.

Howe, H.F. and J.S. Brown. 1999. Effects ofbirds and rodents on synthetic tallgrass

communities. Ecology 80: 1776-1781.

Howe, H.F. and G.F. Estabrook. 1977. On intraspecific competition for avian dispersers

in tropical trees. American Naturalist 111: 817-832.

100 Howe, H.F. and M.N. Miriti. 2004. When seed dispersal matters. BioScience 54: 651­

660.

Howe, H.F. and R.B. Primack. 1975. Differential seed dispersal by birds ofthe tree

Casearia nitida (Flacourtiaceae). Biotropica 7: 278-283.

Howe, H.F., E.W. Schupp and L.C. Westley. 1985. Early consequences ofseed dispersal

for a neotropical tree (Virola surinamensis). Ecology 66: 781-791.

Hulme, P.E. 1994. Rodent post-dispersal seed predation in grassland: magnitude and

sources ofvariation. Journal o/Ecology 82: 645-652.

Hulme, P.E. 1997. Post-dispersal seed predation and the establishment ofvertebrate

dispersed plants in Mediterranean scrublands. Oecologia Ill: 91-98.

Hulme, P.E. 1998. Post-dispersal seed predation: consequences for plant demography and

evolution. Perspectives in Plant Ecology, Evolution and Systematics 1: 32-46.

Inglis, LR., D.S. Shepherd, P. Smith, P.l. Haynes, D.S. Bull, D.P. Cowan and D.

Whitehead. 1996. Foraging behavior ofwild rats (Rattus norvegicus) towards new

foods and bait containers. AppliedAnimal Behavior Science 47: 175-190.

101 Izhaki, I., P.B. Walton and U.K. Safriel. 1991. Seed shadows generated by frugivorous

birds in an eastern Mediterranean scrub. Journal ofEcology 79: 575-590.

Izhaki, I. 2002. The role offruit traits in determining fruit removal in east Mediterranean

ecosystems. Pages 161-175 in D.J. Levey, W.R. Silva and M. Galetti (eds.). Seed

dispersal and frugivory: ecology, evolution and conservation. CABI Publishing,

New York. 511 pp.

James, H.F. and S.L. Olson. 1991. Descriptions ofthirty-two new species ofbirds from

the Hawaiian Islands: Part II. Passeriformes. Ornithological Monographs 46: 7-8,

78-88.

Janzen, D.H. 1971. Seed predation by animals. Annual Review ofEcology and

Systematics 26: 465-492.

Janzen, D.H. 1988. Tropical dry forests, the most endangered major tropical ecosystem.

Pages 130-144 in E.O. Wilson (ed.). Biodiversity. National Academy Press.

Washington, D.C. 521 pp.

Jones, F.A., C.J. Peterson and B.L. Haines. 2003. Seed predation in neotropical pre­

montane pastures: site, distance and species effects. Biotropica 35: 219-225.

102 Jordano, P. 2001. Fruits and frugivory. Pages125-165 in M. Fenner (ed). Seeds: the

ecology ofregeneration in plant communities. CABI, Wallingford, UK. 410 pp.

Jordano, P. and E.W. Schupp. 2000. Seed disperser effectiveness: the quantity component

and patterns ofseed rain for Prunus mahaleb. Ecological Monographs 70: 591­

615.

Kirch, P.V. 1982. The impact ofthe prehistoric Polynesians on the Hawaiian ecosystem.

Pacific Science 36: 1-14.

Kitajima, K. and M. Fenner. 2001. Ecology ofseedling regeneration. Pages 331-359 in

M. Fenner (ed). Seeds: the ecology ofregeneration in plant communities. CABI,

Wallingford, UK. 410 pp.

Kitamura, S., T. Yumoto, P. Poonswad, N. Noma, P. Chuailua, K. Plongmai, T.

Maruhashi, and C. Suckasam. 2004. Pattern and impact ofhornbill seed dispersal

at nest trees in a moist evergreen forest in Thailand. Journal o/Tropical Ecology

20: 545-553.

Lerdau, M., J. Whitbeck, and N.M. Holbrook. 1991. Tropical deciduous forests: death of

a biome. Trends in Ecology andEvolution 6: 201-202.

103 Levey, D.J. 1987. Seed size and fruit-handling techniques ofavian frugivores. The

American Naturalist 129: 471-485.

Loiselle, B.A. and J.G. Blake. 1991. Temporal variation in birds and fruits along an

elevational gradient in Costa Rica. Ecology 72: 180-193.

Loiselle, RA. and J.G. Blake. 1993. Spatial distribution ofunderstory fruit-eating birds

and fruiting plants in a neotropicallowland wet forest. Vegetatio 107/108: 177­

189.

Loiselle, RA. and J.G. Blake. 2002. Potential consequences ofextinction offrugivorous

birds for shrubs ofa tropical wet forest. Pages 397-406 in D.J. Levey, W.R. Silva

and M. Galetti (eds.). Seed dispersal and frugivory: ecology, evolution and

conservation. CABI Publishing, New York. 511 pp.

Loope, L.L. 1998. Hawai'i and the Pacific Islands. Pages 747-774 in M.J. Mac, P.A.

Opler, C.E. Puckett Haecker and P.D. Doran (eds.). Status and trends ofthe

nation's biological resources. 2 vols. U.S. Department ofthe Interior, U.S.

Geological Survey, Reston, Va. 964 pp.

Loope, L.L., A.C. Medeiros and C.G. Chimera. 1995. Natural resources management

plan; Kanaio National Guard Training Area, Island ofMaui, Hawai'i. Available

from Hawai'i National Guard Environmental Office, Honolulu, HI. 86 pp.

104 Lord, J.M. 2004. Frugivore gape size and the evolution offruit size and shape in southern

hemisphere floras. Austral Ecology 29: 430-436.

Makana, J-R. and S.C. Thomas. 2004. Dispersal limits natural recruitment ofAfrican

mahoganies. Oikos 106: 67-72.

Martin, T.E. 1985. Resource selection by tropical frugivorous birds: integrating multiple

interactions. Oecologia 66: 563-573.

Martinez-Ramos, M. and A. Soto-Castro. 1993. Seed rain and advanced regeneration in a

tropical rain forest. Vegetatio 107/108: 299-318.

McConkey, K.R., and D.R. Drake. 2002. Extinct pigeons and declining bat populations:

are large seeds still being dispersed in the tropical Pacific. Pages 381-395 in DJ.

Levey, W.R. Silva and M. Galetti (eds.). Seed dispersal and frugivory: ecology,

evolution and conservation. CABI Publishing, New York. 511 pp.

McConkey, K.R., D.R. Drake, H.J. Meehan, and N. Parsons. 2003. Husking stations

provide evidence ofseed predation by introduced rodents in Tongan rain forests.

Biological Conservation 109: 221-225.

105 McDonnell, M.J. and E.W. Stiles. 1983. The structural complexity ofold field vegetation

and the recruitment ofbird-dispersed plant species. Oecologia 56: 109-116.

Medeiros, A.C., L.L. Loope and C.G. Chimera. 1993. Kanaio Natural Area Reserve

biological inventory and management recommendations. Natural Area Reserve

System, Hawai'i. 90 pp.

Medeiros, A.C., L.L. Loope and R. Hobdy. 1984. Vegetation ofa lowland leeward site

near Kihei, Maui, p. 78-82 in Proc. Fifth Conf. in Natural Sciences, Hawai'i

Volcanoes National Park. Cooperative National Park Studies Unit, Department of

Botany, University ofHawai'i.

Medeiros, A.C., L.L. Loope, and R.A. Holt. 1986. Status ofnative flowering plant

species on the south slope ofHaleakala, East Maui, Hawai'i. Cooperative

National Park Resources Studies Unit, University ofHawai'i, Honolulu.

Technical Report 59: 230 pp.

Medeiros, A.C., e.G. Chimera, and L.L. Loope. 1996. Ka'uhako Crater botanical

resource and threat monitoring, Kalaupapa National Historical Park, Island of

Moloka'i, Hawai'i. Cooperative National Park Resources Studies Unit, University

ofHawai'i, Honolulu. Technical Report 110: 71 pp.

106 Meehan, H.J., K.R. McConkey, and D.R. Drake. 2002. Potential disruptions to seed

dispersal mutualisms in Tonga, Western Polynesia. Journal ofBiogeography 29:

695-712.

Mehrhoff, L.A. 1998. Endangered and threatened species. Pages 150-153 in S.P. Juvik

and J.O. Juvik (eds.). Atlas ofHawai'i, Third Edition. University ofHawai'i

Press, Honolulu. 333 pp.

Moles, A.T. and D.R. Drake. 1999. Post-dispersal seed predation on eleven large-seeded

species from the New Zealand flora: a preliminary study in secondary forest. New

Zealand Journal ofBotany 37: 679-685.

Mueller-Dombois, D. and F.R. Fosberg. 1998. Vegetation ofthe tropical Pacific islands.

Springer-Verlag, New York. 733 pp.

Murphy, P.G. and A.E. Lugo. 1986. Ecology oftropical dry forest. Annual Review of

Ecology and Systematics 17: 67-88.

Mun'ia, R. and L. Gonzalez. 1982. Microhabitat selection by two Chilean cricetid rodents.

Oecologia 52: 12-15.

107 Nystrand, o. and A. Granstrom. 1997. Post-dispersal predation on Pinus sylvestris seeds

by Fringilla spp: ground substrate affects selection for seed color. Oecologia 110:

353-359.

Olson, S.L. and H.F. James. 1982. Prodromus ofthe fossil avifauna ofthe Hawaiian

Islands. Smithsonian Contributions to Zoology 365: 1-59.

Olson, S.L. and H.F. James. 1991. Descriptions ofthirty-two new species ofbirds from

the Hawaiian Islands. Part I. Non-passeriformes. Ornithological Monographs 45.

The American Ornithological Union, Washington, D.C. 88 pp.

Otero-Amaiz, A. S. Castillo, J. Meave, and G. Ibarra-Manriquez. 1999. Isolated pasture

trees and the vegetation under their canopies in the Chiapas coastal plain, Mexico.

Biotropica 31: 243-254.

Osunkoya, 0.0. 1994. Postdispersal survivorship ofnorth Queensland rainforest seeds

and fruits: effects offorest, habitat and species. Australian Journal ofEcology 19:

52-64.

Pratt, H.D., P.L. Bruner, and D.G. Berrett. 1987. The birds ofHawai'i and the tropical

Pacific. Princeton University Press, Princeton, New Jersey. 409 pp. with 45

plates.

108 Pratt, L.W. and S.M. Gon III. 1998. Terrestrial ecosystems. Pages 121-129 in S.P. Juvik

and J.O. Juvik (eds.). Atlas ofHawai'i, Third Edition. University ofHawai'i

Press, Honolulu. 333 pp.

Price, O. F. 2004. Indirect evidence that frugivorous birds track fluctuating fruit resources

among rainforest patches in the Northern Territory, Australia. Austral Ecology 29:

136-144.

Ramon Obeso, J. and I.C. Ferncindez-Calvo. 2002. Fruit removal, pyrene dispersal, post­

dispersal predation and seedling establishment ofa bird-dispersed species. Plant

Ecology 165: 223-233.

Reynolds, R.T. and J.M. Scott. 1980. A variable circular-plot method for estimating bird

numbers. Condor 82: 309-313.

Richardson, D.M., N. Allsopp, C.M. D'Antonio, S.J. Milton and M. Rejmanek. 2000.

Plant Invasions - the role ofmutualisms. Biological Reviews 75: 65-93.

Rock, J.F. 1913. The indigenous trees ofthe Hawaiian Islands. Reprinted in 1974 by

Pacific Tropical Botanical Garden, Lawai, Kauai, Hawai'i and Charles F. Tuttle,

Rutland, Vermont. 548 pp.

109 Russel, C.A. 1980. Food habits ofthe roofrat (Rattus rattus) in two areas ofHawai'i

Volcanoes National Park. Pages 269-272 in C.W. Smith (ed.). Proceedings, third

conference in natural sciences, Hawai'i Volcanoes National Park. Cooperative

National Park Resources Studies Unit, University ofHawai'i, Honolulu.

Ryan, R and RL. Joiner. Minitab handbook. Fourth Edition. Duxbury press, Pacific

Grove, CA. 464 pp.

Sadler, J.P. Biodiversity and oceanic islands: a palaeoecological assessment. Journal of

Biogeography 26: 75-87.

Sanchez-Cordeiro, V. and R. Martinez-Gallardo. 1998. Postdispersal fruit and seed

removal by forest-dwelling rodents in a lowland rainforest in Mexico. Journal of

Tropical Ecology 14: 139-151.

Scott, J.M., S. Mountainspring, F.L. Ramsey, and C.R Kepler. 1986. Forest bird

communities ofthe Hawaiian Island: their dynamics, ecology, and conservation.

Studies in Avian Biology 9. 431 pp.

Scowcroft, P.G. and H.F. Sakai. 1984. Stripping ofAcacia koa bark by rats on Hawai'i

and Maui. Pacific Science 38: 80-86.

110 Shiels, A.B. and L.R Walker. 2003. Bird perches increase forest seeds on Puerto Rican

landslides. Restoration Ecology 11: 457-465.

Slocum, M.G. and C.C. Horvitz. 2000. Seed arrival under different genera oftrees in a

neotropical pasture. Plant Ecology 149: 51-62.

Snow, D.W. 1981. Tropical frugivorous birds and their fruit plants: a world survey.

Biotropica 3: 1-14.

Sohmer, S.H. and R Gustafson. 1987. Plants and flower ofHawai'i. University of

Hawai'i Press, Honolulu. 160 pp.

Stanley, M.C., E. Smallwood, and A. Lill. 2002. The response ofcaptive silvereyes

(Zosterops lateralis) to the colour and size offruit. Australian Journal o/Zoology

50: 205-213.

Stansbury, C.D. and G. Vivian-Smith. 2003. Interactions between frugivorous birds and

weeds in Queensland as determined from a survey ofbirders. Plant Protection

Quarterly 18: 157-165.

Staples, G.W. and RH. Cowie. 2001. Hawai'i's invasive species. Mutual Publishing and

Bishop Museum Press, Honolulu, Hawai'i. 116 pp.

111 Steadman, D.W. 1995. Prehistoric extinctions ofPacific Island birds: biodiversity meets

zooarcheology. Science 267: 1123-1131.

Stiles, E.W. 2001. Animals as seed dispersers. Pp. 111-124 in M. Fenner (ed). Seeds: the

ecology ofregeneration in plant communities. CABI, Wallingford, UK. 410 pp.

Stone, c.P. 1985. Non-native animals in Hawai'i's native ecosystems: toward controlling

the adverse effects ofintroduced vertebrates. Pages 251-297 in C.P. Stone and M.

Scott (eds.). Hawai'i's terrestrial ecosystems: preservation and management.

Cooperative National Park Resources Studies Unit, University ofHawai'i,

Honolulu.

Stone, C.P. 1989. Non-native land vertebrates. Pages 88-95 in C.P. Stone and D.H. Stone

(eds.). Conservation biology in Hawai'i. Cooperative National Park Resources

Studies Unit, University ofHawai'i, Honolulu. 252 pp.

Stone, c.P., P.C. Banko, P.K. Higashino, and F.G. Howarth. 1984. Interrelationships of

Non-native and native plants and animals in Kipahulu Valley, Haleakala National

Park: a preliminary report. Pages 91-105 in C.W. Smith (ed.). Proceedings, fifth

conference in natural sciences, Hawai'i Volcanoes National Park. Cooperative

National Park Resources Studies Unit, University ofHawai'i, Honolulu.

112 Stone, C.P., C.W. Smith, and J.T. Tunison. 1992. Non-native plant invasions in native

ecosystems ofHawai'i. Cooperative National Park Resources Studies Unit,

University ofHawai'i, Honolulu. 887 pp.

Sugihara, R.T. 1997. Abundance and diets ofrats in two native Hawaiian forests. Pacific

Science 51: 189-198.

Temple, S.A. 1977. Plant-animal mutualism: coevolution with Dodo leads to near

extinction ofplant. Science 197: 885-886.

Terborgh, J., E. Loso, M.P. Riley and M.B. Filey. 1993. Predation by vertebrates and

invertebrates on the seeds offive canopy tree species ofan Amazonian forest.

Pages 375-386 in T.H. Fleming and A. Estrada (eds.). Frugivory and seed

dispersal: ecological and evolutionary aspects. Kluwer Academic Publishers,

Belgium.

Tomich, P.Q. 1962. Notes on the barn owl in Hawai'i. 'Elepaio 23: 16-17.

Tomich, P.Q. 1986. Mammals in Hawai'i. Bishop Museum Press, Honolulu. 375 pp.

113 Traveset, A. and M. Verdu 2002. A meta-analysis ofthe effect ofgut treatment on seed

germination. Pages 339-350 in D.J. Levey, W.R. Silva and M. Galetti (eds.). Seed

dispersal and frugivory: ecology, evolution and conservation. CABI Publishing,

New York. 511 pp.

Uhl, C. 1998. Restoration ofdegraded lands in the Amazon basin. Pages 326-332 in E.O.

Wilson (ed.). Biodiversity. National Academy Press. Washington, D.C. 521 pp.

Uhl, c., H. Clark, K. Clark and P. Maquirino. 1982. Successional patterns associated with

slash-and-bum agriculture in the upper Rio Negro region ofthe Amazon basin.

Biotropica 14: 249-254.

Vander Wall, S.B. and W.S. Longland. 2004. Diplochory: are two seed dispersers better

than one? Trends in Ecology andEvolution 19: 155-161.

Van Riper, S.G. 2000. Japanese white-eye (Zosterops japonicus). The Birds ofNorth

America, no. 487. The Birds ofNorth America, Inc., Philadelphia, PA.

Wagner, W.L., D.R. Herbst, and S.H. Sohmer. 1999. Manual ofthe flowering plants of

Hawai'i, revised addition. B.P. Bishop Museum and University ofHawai'i Press,

Honolulu. 1919 pp.

114 Wenny, D.G. 2000. Seed dispersal, seed predation, and seedling recruitment ofa

neotropical montane tree. Ecological Monographs 70: 331-351.

Wenny, D.G. 2001. Advantages ofseed dispersal: a re-evaluation ofdirected dispersal.

Evolutionary Ecology Research 3: 51-74.

Wheelwright, N.T. 1985. Fruit size, gape width, and the diets offruit-eating birds.

Ecology 66: 808-818.

Wijdeven, S.M.J. and M.E. Kuzee. 2000. Seed availability as a limiting factor in forest

recovery processes in Costa Rica. Restoration Ecology 8: 414-424.

Williams, P.A. and B.J. Karl. 1996. Fleshy fruits ofindigenous and adventive plants in

the diet ofbirds in forest remnants, Nelson, New Zealand. New Zealand Journal

o/Ecology 20: 127-145.

Williams, P.A., B.J. Karl, P. Bannister, and W.G. Lee. 2000. Small mammals as potential

seed dispersers inNew Zealand. Austral Ecology 25: 523-532.

Willson, M.F. and A. Traveset. 2001. The ecology ofseed dispersal. Pages 85-100 in M.

Fenner (ed). Seeds: the ecology ofregeneration in plant communities. CABI,

Wallingford, UK. 410 pp.

115 Wirtz, W.O. 1972. Population ecology ofthe Polynesian rat, Rattus exulans, on Kure

Atoll, Hawai'i. Pacific Science 26: 433-465,

Yagihashi, T., M. Hayshed and T. Miyamoto. 1998. Effects ofbird ingestion on seed

germination ofSorbus committal. Oecologia 114: 209-212.

Yarranton, G.A. and RG. Morrison. 1974. Spatial dynamics ofa primary succession:

nucleation. Journal ofEcology 62: 417-428.

Zar, J.H. Biostatistical analysis. 4th Edition. Prentice Hall, Saddle River, New Jersey. 663

pp.

Ziegler, A.C. 2002. Hawaiian natural history, ecology, and evolution. University of

Hawai'i Press, Honolulu. 477 pp.

Zimmerman, J.K., J.B Pascarella and T.M. Aide. 2000. Barriers to forest regeneration in

an abandoned pasture in Puerto Rico. Restoration Ecology 8: 350-360.

116