Citation
Edge effects on the pollination of tropical cloud forest plants

Material Information

Title:
Edge effects on the pollination of tropical cloud forest plants
Creator:
Murcia, Carolina, 1960-
Publication Date:
Language:
English
Physical Description:
vii, 93 leaves : ill. ; 29 cm.

Subjects

Subjects / Keywords:
Ecology ( jstor )
Ecosystems ( jstor )
Edge effects ( jstor )
Flowering ( jstor )
Flowers ( jstor )
Forests ( jstor )
Pollination ( jstor )
Pollinators ( jstor )
Species ( jstor )
Vegetation ( jstor )
Dissertations, Academic -- Zoology -- UF
Zoology thesis Ph. D
Genre:
bibliography ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1993.
Bibliography:
Includes bibliographical references (leaves 84-92).
General Note:
Typescript.
General Note:
Vita.
Statement of Responsibility:
by Carolina Murcia.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright Carolina Murcia. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Resource Identifier:
030082287 ( ALEPH )
29996177 ( OCLC )

Downloads

This item has the following downloads:


Full Text








EDGE EFFECTS ON THE POLLINATION OF TROPICAL
CLOUD FOREST PLANTS









By

CAROLINA MURCIA


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


1993












ACKNOWLEDGMENTS


I would like to thank the members of my doctoral committee, Drs.
Peter Feinsinger, Frank G. Nordlie, Douglas J. Levey, Richard A. Kiltie, Francis E. Putz and Kent H. Redford, for their advice during all phases of the project. I would like to thank especially Peter Feinsinger for his support and encouragement through all my graduate student years. I would also like to extend my appreciation to the following persons: Natalia Arango for her assistance and companionship during the field work phase; Ronald Edwards for graciously walking me through the nuances of IBM computing for the Correspondence Analysis, G. Kattan, R. Edwards and Dr. Carmine Lanciani for commenting on portions of earlier drafts of this dissertation; Mr. Luis Felipe Carvajal, Mr. Eduardo Calder6n and Mr. Antonio Gonzilez Caicedo for permission to work on their properties; Mr. Luis Miguel Constantino for identifying some butterflies; Mr. Germin Parra for the identification of bees; Mr. Stinger Guala and Dr. Walter Judd for their help in determinating the plant material. Many of the ideas presented here benefitted greatly from discussions with fellow students in the Department of Zoology, to all of them I thank.
Financial support for this project was provided by the Department of Zoology, University of Florida, the Underhill Foundation, and the Fundaci6n para la Promoci6n de la Investigaci6n y la Tecnologfa, Banco de la Repiblica, Bogotd, Colombia.


ii







Finally and most importantly, I would like to acknowledge the support and encouragement of my husband, Gustavo Kattan, who put up with my ups and downs with love and a smile.


iii













TABLE OF CONTENTS


page

ACKNOWLEDGEMENTS .................................. ii

ABSTRACT .............................................. vi

CHAPTERS

1 INTRODUCTION .................................. . 1

2 BACKGROUND .................................... . 5

Landscape Ecology and Edges .......................... 5
Historical Perspectives on Edges ........................ 7
Why Is Exposure to the Edge Deletereous
for the Fragments? ............................... 10
Physical Edge Effects .......................... 10
Direct Biological Edge Effects .................... 12
Indirect Biological Edge Effects .................. . 14
Edges and Plant-Animal Interactions ................... . 15

3 EDGE EFFECTS ON THE POLLINATION OF TROPICAL CLOUD
FOREST PLANTS

Introduction ...................................... . 18
M ethods ........................................... 23
Study Site .................................... 23
Sampling Scheme .............................. 26
The Plants .................................. . 31
Statistical Analyses and Variables ................ 34
Effect of Sample Size on Standard Deviation ........ . 36 Temporal Variation of Edge Effect ................ 36
Experim ents .................................. 37
Edge Description .............................. . 38
Results ........................................... 40
Pollination .................................... 40
Effect of Sample Size on Standard Deviation ........ . 55 Temporal Variation of Edge Effect ................ 55
Experim ents .................................. 58
Edge Description .............................. . 58


iv








D iscussion ........................................ . 66
Study Design ............................... .. . 66
The Plant-Pollinator Interaction System ............ 68
Edge Effect on Physical Conditions ........ ........ 71
Time Since Edge Creation ....................... . 72
Conclusions .................................. . 74
4 TOWARDS A UNIFIED THEORY OF EDGES
Contrast as a Determinant of Edge effects........... .75
Future Directions .............................. . 82

LITERATURE CITED .................................... . 84

BIOGRAPHICAL SKETCH ................................. 93


V












Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

EDGE EFFECTS ON THE POLLINATION OF TROPICAL CLOUD FOREST PLANTS

By

Carolina Murcia

May 1993

Chairperson: Dr. Frank Nordlie
Major Department: Zoology

Exposure to edge is one consequence of habitat fragmentation that can result in detrimental effects on the fauna and flora of fragmented habitats. In this study, edge is defined as a sharp transition between natural and anthropogenic habitats, such as between forest and pastures. I assessed the effect of proximity to the edge on the pollination levels of 13 cloud forest plant species in Colombia. I collected the styles of plants located in three zones of each of three forest fragments: edge (0-10 m), transition zone (10-50 m), and interior (50-200 m).
Pollination levels, measured as the median number of pollen tubes that reached the base of the styles, were not consistently affected in their pollination by the proximity to the edge. Intra-individual coefficient of variation, i.e. variation in pollination levels among flowers in indidividuals was also not affected by the proximity to the edge. Few plant species were significantly affected by the edge, and those that were affected differed in the direction of the edge effects. To measure the edge effect on the potential for


vi







limitation of seed set and its penetration on successive levels of the plantpollinator interaction, I measured the proportion of flowers in each individual with a number of pollen tubes equal to or larger than half the number of ovules. In one of two cases where there was an edge effect on pollination levels, the potential for seed set was also affected. Edge effects, when present, were inconsistent between flowering seasons.
Field experiments with two introduced plant species showed no
differences in pollination levels between the plants placed at the edge and in the forest interior.
There are several possible explanations for these results. First, aboveground environmental conditions do not differ between the edge and the interior, thus making it unlikely that the pollinators are affected by the edge. Second, below-ground conditions are affected only for the first 10 m into the forest, and this affects the densities of the focal plants in only one of the fragments. Finally, plant-pollinator interactions can be robust to some perturbations because generally, these interactions have little speciesspecificity.


vii













CHAPTER ONE
INTRODUCTION
When a natural ecosystem is modified by humans the result is often a mosaic of isolated remnants of native vegetation interspersed with large areas of modified habitat. This process of habitat fragmentation has deleterious consequences for most of the native biota. First and most obviously the total area of the original vegetation is reduced. This reduction in area often results in species extinction. Because the process of extinction in fragments is analogous to that occurring on islands, the theory of island biogeography provided the initial conceptual framework for studying this process in terrestrial ecosystems (Diamond 1975, Simberloff and Gotelli 1984, Simberloff and Abele 1976, Terborgh 1976, Wilcox 1980), but see (Margules, et al. 1982).
In addition to a reduction in total area, fragmentation can have an
impact on the organisms that remain in the fragment through what has been termed "edge effects" (Lovejoy, et al. 1986, Saunders, et al. 1991, Soul6 1986, Wilcove, et al. 1986). Edge effects result from the interaction between two adjacent habitats. In fragmented habitats, edge effects are exacerbated as a consequence of bidimensional geometry. With the reduction in fragment area, the area/perimeter ratio decreases. As a result, a larger proportion of the habitat island is exposed to the edge, or line of contact with the other habitat.
The intensity of edge effects is strongest in the immediate vicinity of the edge, and declines away from the edge (Blanchard 1992, Gates and Gysel 1978, Hester and Hobbs 1992, Kapos 1989, Ranney, et al. 1981, Wales 1972,


1





2


Williams-Linera 1990a, Williams-Linera 1990b). In addition to distance from the edge, several factors seem to mitigate the strength of the effect: edge age (Blanchard 1992, Williams-Linera 1990b), aspect (Gysel 1951, Ranney et al. 1981), orientation (Kapos 1989, Palik and Murphy 1990, Ranney et al. 1981, Wales 1972), and also the type of vegetation on both sides of the edge (Chasko and Gates 1982).
Attention to edge effects has focused mainly on the fragmentation of forests and their replacement with simple agricultural or pastoral ecosystems. The general pattern of changes for forest-field edges that can be extracted from the literature indicates that forest edges tend to have higher solar incidence than the interior of the fragments (Blanchard 1992, Kapos 1989, Oosting and Kramer 1946). Consequently, air temperatures near the edge are usually higher and more variable (Kapos 1989, Williams-Linera 1990a), and soils tend to be dryer (Kapos 1989, Oosting and Kramer 1946) than in the forest interior. With higher light availability, plant growth is enhanced. Therefore edges tend to have higher foliar densities (Blanchard 1992, Malcolm 1991), and higher plant densities (Chen, et al. 1992, Palik and Murphy 1990, Ranney et al. 1981, Sork 1983, Wales 1972, Williams-Linera 1990a), in spite of higher tree mortality caused by exposure and wind throw (Chen et al. 1992, Lovejoy et al. 1986). This abundance in cover and plant density attracts herbivorous insects (Blanchard 1992), and nesting birds (Chasko and Gates 1982), which in turn attract nest predators (Gates and Gysel 1978), and brood parasites (Brittingham and Temple 1983). Although organisms from outside the fragments may be attracted to the edge (Brittingham and Temple 1983, Chasko and Gates 1982, Hester and Hobbs 1992, Laurance and Yensen 1991), the change in the physical environment associated with edges may also cause some plants and animals to avoid the





3


edges (Chen et al. 1992, Malcolm 1991, Ranney et al. 1981). Thus, changes associated with edges can have effects at many levels. At the individual level edges can affect the reproductive potential of plants (Blanchard 1992) and animals (Angelstam 1986, Brittingham and Temple 1983, Chasko and Gates 1982, Gibbs 1991, Wilcove et al. 1986); at the population level, edges can affect the presence of species near the edge (Blanchard 1992, Lovejoy et al. 1986, Quintela 1986, Ranney et al. 1981); and at the community level changes in the interaction among species can be affected as well. Little is known on the effects of edge at the community level because, with the exception of avian nest predation and brood parasitism, edge effects on processes that involve species interactions have rarely been studied. Of particular interest are plant-animal interactions (such as pollination, frugivory, herbivory, or seed dispersal), due to their importance in the dynamics of many forests.
Plant-animal interactions can be very sensitive to disturbance because of the dual effect on both the plants and the animals, and because of the speciesspecific nature of many of these interactions (Kevan 1975, McClanahan 1986). In this community-level study, I address how interactions among plants and their pollinators are affected by their proximity to a forest-pasture edge. Because researchers perceive and define the concepts of edge and their effects differently, I begin (chapter 2) by reviewing the evolution of the concepts, and their current definitions. I confine this review to edges between forests and field/pastures, because these edges are the most relevant to the conservation of forest fragments. I also draw general patterns on edge effects at all levels of organization (individuals, populations or communities), and divide edge effects into three classes: effects on the physical environment, direct biological effects, and indirect biological effects. This context allows me





4


to discuss the importance of plant-animal interactions, and in particular plant-pollinator interactions, in the dynamics of tropical forests, and their potential susceptibility to edges.
In chapter 3, I present a study on the effect of proximity to the edge on the pollination of 13 tropical cloud-forest plant species. I found no overall effects on pollination at the community level. At the species level a few species showed edge effects, yet these effects were inconsistent between flowering seasons. This lack of edge effects was also confirmed by field experiments using two introduced plant species. Finally, edge effects on the physical environment, when present, extended only 10 m into the forest. These results do not accord well with the predictions made in the literature. This lack of agreement is perhaps due to two factors acting in concert: a) the intensity of edge effects varies depending on the types of ecosystems on each side of the boundary, and age, orientation or aspect of the edge; and b) there is little agreement among researchers on what is an edge and how to measure its effects.
In the final chapter, I propose a model that predicts the intensity of edge effects based on the degree of contrast between the ecosystems separated by the edge. The concept of contrast is an old one, and has been mentioned as one of the factors that may determine the intensity of the edge effect (Santos and Telleria 1992, Wiens, et al. 1985), yet it has not been formally defined and its role on modulating edge effects has been overlooked. I finish by suggesting future directions for research on edges. In this final section, I also point out common problems in the design of the studies and interpretation of the results that, in my concept, may obstruct the advancement of our knowledge on edges.













CHAPTER TWO
BACKGROUND
Landscape Ecology and Edges
A landscape is an area composed of interacting ecosystems (Forman and Godron 1986, p. 11). In natural landscapes, a variety of ecosystems may result from heterogeneity in the spatial distribution of critical resources and disturbances. Because resources tend to vary spatially following gradients, ecosystems do not have sharp boundaries but merge gradually (McCoy, et al. 1986). The area of contact between ecosystems is an ecotone.

When boundaries between ecosystems are abrupt, the ecotone is almost unidimensional and it is called an edge. Natural edges are rare and occur mostly as a result of abrupt changes in the soil as occurs, for example, in rocky outcrops, or serpentine soils. In contrast, edges are more prevalent in human-made ecosystems or when those ecosystems are created and maintained next to natural vegetation.
Ecotones and edges separate ecosystems, but are not impermeable to
fluxes. Edges may function as semi-permeable membranes that filter flows of energy, nutrients and species between ecosystems (Forman and Godron 1986, Wiens et al. 1985). Flows across these membranes cause transitory or permanent changes in the composition and abundance of species in the area of contact. These changes are called edge effects. The intensity of an edge effect has been measured as the distance (d) into an ecosystem that these changes penetrate (Laurance and Yensen 1991). Changes due to edge effects


5





6


modify the ecosystem beyond any intrinsic natural variation of the ecosystem. In other words, the conditions at the edge occur only when an edge is present.

The area near the edge that is modified by edge effects has the potential of becoming an ecotone if no further human intervention occurs (Gysel 1951). In that case, species from both ecosystems would gradually invade the other, "smoothing" the transition between the two. If, on the other hand, the invasion of at least one ecosystem by the other is artificially suppressed, the result is the maintenance of an abrupt edge. The points of edge creation and edge maintenance have been used by Ranney et al. (1981) to define three different types of edges: cantilevered, canopy dripline, and advancing. Cantilevered edges are maintained at their point of creation. These edges are characterized by the overhanging canopy of the branches that grow towards the open space, and a thick understory among large tree trunks. Canopy dripline edges are those maintained at the outer tips of the horizontal branches of the canopy trees. These edges have a dense understory, shaded by the branches of the outermost canopy trees, but have no large tree trunks exposed to the edge. Canopy dripline edges represent an abrupt change in height and plant density between fields and forests. Advancing edges are maintained several meters away from their point of creation, or not maintained at all. They are characterized by a dense vegetation that gradually declines in height between the point of edge creation and the point of edge maintenance. Advancing edges are not abrupt as are cantilevered or canopy dripline edges, and consequently I categorize them as ecotones rather than edges. The effects of ecosystems on one another across an ecotone are likely to differ from the effects across an edge. This difference is perhaps a result of the longer distances separating ecosystems and the presence of a set of plants and animals that are be more tolerant to the conditions of both





7


ecosystems. For that reason, I will restrict the remainder of the dissertation to abrupt ecotones, i.e., cantilevered and dripline edges.
The effects of ecosystems on one another across the edge may be asymmetrical depending on the relative sizes of the ecosystems. In a landscape, the ecosystem with the largest total area and highest connectivity among patches is called the matrix, and it dominates the dynamics of the landscape (Forman and Godron 1986, p. 159). Elements in the landscape that are less connected than the matrix are called patches or fragments, and are considered to be "immersed" in the matrix. Because flows of energy, nutrients and species are dominated by the matrix, edge effects caused by the contact between the fragments and the matrix are likely to affect the fragments more strongly than the matrix. This asymmetry in the effects has historically directed the interest of ecologists towards the ecosystem that experiences the strongest effects.
Historical Perspective on Edge Effects
In North America, wildlife managers concerned themselves with edges when they noticed that most game species thrived where forest, brush, pastures, and crops were in contact. The increase in the populations of game species at these edges was termed the "edge effect" by Aldo Leopold (Leopold 1933)(p. 131). In the interest of increasing game populations, Leopold (1933, p. 131) formulated his Law of Interspersion. It stated that "...{it is highly desirable to grant game} simultaneous access to more than one environmental type". Leopold's recommended management technique required interspersion of habitats maximizing the amount of edge for any ecosystem area.
To maximize the area of contact between ecosystems or habitats some management manuals recommend the interspersion of fields and croplands





8


with sedges and vegetation that would make an otherwise depauperate area more attractive to wildlife (Burger 1973, Yoakum, et al. 1980, Yoakum and Dasmann 1971). Although this management technique could potentially increase the are available for forest habitat, the recommendations were not always interpreted to the benefit of the forests. In some instances, for example, large expanses of forest were considered depauperate (Dasmann 1964), and consequently manuals recommended the interspersion of forest with other ecosystems at the expense of the forest (Dasmann 1964, Thomas, et al. 1979). Interestingly, European wildlife managers seem unconcerned with edges as a wildlife management technique (Dagg 1976), perhaps because the degree of fragmentation present in Europe in the last centuries has made any further creation of edges unnecessary.
Foresters and silviculturalists, on the other hand, have perceived edges as detrimental, although the relative importance given to the edge effects varies between the two schools of silviculture (Bradshaw 1992). Since the Middle Ages, the school that advocates clearcutting, and the school that advocates selective felling have debated the relative importance of edges in maintaining harvestable forests. Those in favor of clearcutting claim that the technique minimizes the amount of forest that could be exposed to the edge, in spite of a costly reduction in the rate of recovery of the logged area (Bradshaw 1992). Advocates of selective felling, on the other hand, are aware that this technique increases the area exposed to edge and the problems associated with the creation of edge, yet favor the technique because the rates of recovery of the cleared area are higher that with clearcutting (Bradshaw 1992).

As forested areas shrink faster than they can recover, the ongoing
controversy between silvicultural schools is of utmost interest. As the total





9


forest area decreases, the rates of recovery of the cleared land decrease due to the shift in the dynamics of the landscape. From a forest dominated landscape, where clearcuts were the fragments and were readily affected by the recolonization of the forest, the landscape has changed to one dominated by clearcuts. In the latter scenario, fluxes of energy and species are dominated by the clearcut matrix to the detriment of the forest fragments. As a result of this shift, it is likely that the school of forestry that advocates clearcutting may loose support, and the controversy may be resolved. In Australia and USA, for example, there is increasing pressure for eliminating clearcutting (Bradshaw 1992).
With the increase in the rates of forest fragmentation, the concepts of edges and edge effects have also changed among wildlife managers. In the past two decades, the concern for game species has given way to concern about the edge effects on forest wildlife and its preservation. Consequently, the focus of wildlife managers has also switched towards non-game wildlife (Harris 1988, Yahner 1988), and their perception of edges as management tools has changed.
Efforts have been made to identify the factors that determine the
potential for species conservation in habitat remnants. Under the influence of island biogeography theory, the initial emphasis was on "area" as the major factor determining the conservation potential of a site (Diamond 1975). Many heated debates, however, convinced ecologists that area per se is not the only, or best ,predictor of the number of species that can be supported by a particular reserve (Margules et al. 1982, Simberloff and Gotelli 1984, Simberloff and Abele 1976). Other factors associated with fragmentation but partially independent from area, have been identified as contributing to the deleterious. effect of fragmentation on species conservation (Blouin and





10


Connor 1985, Lovejoy et al. 1986, Saunders et al. 1991, Sould 1986). Among these factors is the increased exposure of the fragment species to the edge (Lovejoy et al. 1986, Saunders et al. 1991, Soul6 1986, Williamson 1975).
Edge effects have been recognized qualitatively for many decades, yet
quantification of these effects (with a few exceptions) has only taken place on the last 12 years (Gates and Gysel 1978, Gysel 1951, Oosting and Kramer 1946, Wales 1972). This quantification has focused on measuring the distance that edge effects penetrate the fragment with the ultimate purpose of determining what proportion of the fragment still contains typical forest conditions, and thus its potential for conservation.
Why Is Exposure to the Edge Considered Deleterious for the Fragments?
Forest fragmentation shifts the fluxes of energy, nutrients and species in the landscape from being forest-dominated to being dominated by the new matrix. These shifts are bound to cause edge effects on the forest. I divide the types of edge effects in three categories. 1) Changes in the environmental conditions that result from the proximity to the matrix are physical edge effects. 2) Changes in the abundance and distribution of species caused directly by the physical conditions near the edge, e. g. desiccation, windthrow, plant growth, are direct biological edge effects. These are determined by the physiological tolerances of species to the conditions on and near the edge. 3) Changes in species interactions, e.g. predation, brood parasitism, competition, herbivory, biotic pollination, zoochorous seed dispersal, are the result of indirect biological edge effects. Physical Edge Effects
Compared to a structurally complex vegetation such as a forest, structurally simple vegetation, like crops and pastures, allow a higher amount of solar radiation to reach the ground during the day, and higher





11


reradiation to the atmosphere at night (Geiger 1965). Consequently, diurnal temperatures tend to be higher, and daily temperatures fluctuate more widely (Fetcher et al. 1985). High temperatures cause a reduction in air moisture (Fetcher et al. 1985, Geiger 1965). Under taller and more complex vegetation, on the other hand, there are lower air temperatures and narrower temperature fluctuations due to the larger biomass that can absorb solar radiation (Etherington 1982, Fetcher et al. 1985) as well as lower light levels (Chazdon 1986). These differences between the vegetation on each side of the edge are likely to create a gradient perpendicular to the edge.
Changes in the physical conditions associated with the edge show clear patterns. Measurements of incident light, air temperature, air moisture, soil temperature, and soil moisture at the edge (0 m) differ from the interior, displaying intermediate values between the interior of the fragment and the matrix. Differences in these factors between the edge and the interior usually disappear over a distance of 5 to 20 m in a variety of tropical and temperate zone forests (Blanchard 1992, Kapos 1989, MacDougall and Kellman in press, Oosting and Kramer 1946). The less the exposure to solar radiation, the weaker are the physical edge effects. Thus, for north or northeastern facing edges in the Brazilian Amazon and New Jersey, for example, differences between the edge and the interior were smaller than for edges facing other directions, and edge effects penetrated less into the forest (Kapos 1989, Wales 1972).
Physical edge effects can also result from the movement of chemical compounds across the edge, that can alter environmental conditions. Chemical fertilizers from adjacent croplands can penetrate up to 5 m into the wheatbelt shrubland and up to 50 m into the wheatbelt woodland of Australia (Hester and Hobbs 1992). Also, nitrates, sulfates and herbicides from





12


adjacent croplands are known to penetrate into riparian forests in Maryland (Correll 1991). In both cases, values at the edge were highest, and declined with distance into the forest fragment. Direct biological edge effects
Changes in the physical environment caused by edges have different effects on the fauna and flora of the fragments depending on the organisms' physiological tolerances. In the case of plants, edge effects could ultimately determine their densities near the edge. A variety of responses have been observed: some plant species are never found near edges, while others have higher densities near the edge than in the interior, and yet others show no changes in densities as a function of distance to the edge (MacDougall and Kellman in press, Ranney et al. 1981). In the case of forest animals, density and activity have showed diverse responses to the edge, from avoidance to preference (Chasko and Gates 1982, Quintela 1986). In addition, species from the matrix may react to the edge as well. A favorable environment (Hester and Hobbs 1992), or dispersal by abiotic vectors (Romano 1990, Willson and Crome 1989) may cause species from the matrix to converge on the edge, and even penetrate some distance into the fragment.
In general, individual plants show a positive response to the increase in incident light availability near the edge, in spite of the potential dehydration that can reach 5-10 into the forest (Kapos 1989). Leaf density (Blanchard 1992, Malcolm 1991), stem densities and basal areas tend to be higher within 20 m of the edge in a variety of tropical and temperate zone forests (Palik and Murphy 1990, Ranney et al. 1981, Wales 1972, Williams-Linera 1990bX but see )(Chen et al. 1992). In some ecosystems such as undisturbed sugar maplebeach forests in Michigan, however, the increased plant biomass response is stronger among canopy and subcanopy trees (Palik and Murphy 1990), while





13


in second growth forests of sugar maple-beach forests in Michigan, and in the lowland rainforests of the Brazilian Amazon the response is stronger among understory plants (Malcolm 1991). Plant growth is also spurred by the conditions near the edge in a variety of forests in the United States and Panama (Chen et al. 1992, Sork 1983, Wales 1972, Williams-Linera 1990a).
In contrast to the greening caused by increased growth and leaf
production, increased mortality can also result from proximity to edge. Tree mortality can result from windthrow (Chen et al. 1992), and possibly as a result of fire following the creation of the edge (Lovejoy et al. 1986).
Changes in the physical environment also result in shifts in species
composition. Some plants show lower densities or are absent near the edge (Chen et al. 1992, MacDougall and Kellman in press, Ranney et al. 1981, Sork 1983, Wales 1972), while others show higher densities (Chen et al. 1992, MacDougall and Kellman in press, Ranney et al. 1981, Wales 1972), or no changes in density in association with distance to the edge (Blanchard 1992, MacDougall and Kellman in press, Ranney et al. 1981). As a result of these different responses, species composition may differ between the edge and the interior. Tree species composition, for example, differed in one study between the first 5-45 m and the interior in undisturbed sugar maple/beach forest fragments in Michigan (Palik and Murphy 1990). Species composition in insects (Malcolm 1991), and birds (Quintela 1986) also have shown differences between the edge (0- 50 m) and the interior of lowland rain forests in Brazil. In other cases, however, differences in species composition do not occur. Studies on tree and seedling species composition have found no differences in species composition as a result of proximity to the edge in second growth sugar/beach forests in Michigan and in undisturbed lowland





14


rain forests in Panama (Palik and Murphy 1990, Williams-Linera 1990a, Williams-Linera 1990b).
Although some general patterns in edge effects can be extracted from the literature, direct biological effects are not as clear as those observed in the physical environment. Perhaps the most consistent responses are those spurred by the increase in light availability, i.e., leaf and stem growth. Any other effects are less clear, especially when idiosyncratic responses of particular species are involved. This variability in responses may result from a combination of direct and indirect biological edge effects. In other words, different responses observed could be due to some species responding to direct biological edge effects, while others are responding to indirect biological edge effects, or to both.
Indirect Biological Edge Effects
Changes in the distribution and abundance of species near the edge may alter the dynamics of species' interactions near the edge. For example, a leaf flush that results from increased light incidence may attract herbivorous insects. These, in turn, may attract nesting birds, which in turn could attract nest predators and brood parasites. Thus, the indirect effect of light availability on herbivorous insects may initiate a series of cascading effects, that can spread across the fabric of the ecosystem. What this implies, is that the density of one species, could be determined by its response to the physical conditions near the edge, but it could also result from its response to other species. Most studies only describe changes in densities and species compositions near the edge, but few have explored the causes for these changes.
Studies that concentrate on the interactions among species may shed some light on the importance of indirect effects on the species near the edge.





15


The species interactions that have received most attention are nest predation and brood parasitism in birds. The results have been inconsistent, however. Some studies have found higher rates of nest predation on or near the edge in an oak-hickory forest in Michigan, a lowland rain forest in Costa Rica, and in Tennessee (Gates and Gysel 1978, Gibbs 1991, Wilcove et al. 1986). In such forests, increased rates of nest predation can occur up to 20 m into the fragments in Costa Rica (Gibbs 1991), 45 m in Michigan (Gates and Gysel 1978), and 300-600 m in Tennessee (Wilcove et al. 1986) as opposed to the forest interior. Other studies have found inconsistent or no significant effects in a variety of temperate zone forests in North America and Europe (Chasko and Gates 1982, Gibbs 1991, Moller 1989, Ratti and Reese 1988, Santos and Tellerfa 1992, Yahner and Wright 1985). Avian brood parasitism in the temperate zone seems to follow the same patterns as nest predation, with higher parasitism near the edge and declining away from it. In one study in a Wisconsin deciduous forest, cowbirds (Molothrus ate) parasitized nests up to 300 m into the forest fragments (Brittingham and Temple 1983). In continuous forests, however, cowbirds did not normally parasitize forest interior birds.
Studies on other types of species interactions are scarce. These studies have found lower post-dispersal seed predation (Sork 1983), higher herbivory (Sork 1983) and zoochorous dispersal of seeds from the matrix (Willson and Crome 1989) to a distance up to 80 m into the forest.
Edges and Plant-Animal Interactions
Plant-animal interactions are crucial to the dynamics of tropical forests. Animals are involved in the pollination, seed dispersal, seed predation and herbivory of a large percentage of tropical species. In the highlands of Costa Rica, for example, animals disperse 70-75 % of forest plants (Stiles 1985).





16


Plant-animal interactions can be very sensitive processes because disturbances can affect the interaction by affecting the plants, the animals, or both (Kevan 1975, McClanahan 1986).
Plants and their associated animals normally encounter a certain degree of environmental variation caused by small scale perturbations in the forest, such as treefall gaps. The effects of canopy gaps on the dynamics of plantanimal interactions have been explored for several systems, such as: plantfrugivorous birds (Blake and Hoppes 1986, Hoppes 1988, Levey 1988, Murray 1988), and plants and their pollinating hummingbirds (Feinsinger, et al. 1987, Feinsinger, et al. 1988a).
Large scale perturbations such as those caused by forest fragmentation and exposure to the edge, are apt to affect plant-animal interactions. Extrapolation of potential edge effects from the literature on gaps, however, may not be realistic due to the different spatial and temporal scales involved. At the spatial scale, gaps represent very small clearings compared to those created by forest fragmentation. Therefore climatic conditions may be substantially different. A study in a lowland tropical rainforest, at La Selva, Costa Rica, for example, revealed that microclimatic conditions in a 400 m2 gap are closer to those of the forest interior than to those of a 5,000 m2 clearing (0.5 ha) (Fetcher et al. 1985). Even a 0.5 ha clearing is small when compared to the open areas that typically surround forest fragments. Climatic differences between canopy gaps and the clearings that result from forest fragmentation are likely to be pronounced.
At the temporal scale, edges that result from forest fragmentation tend to be longer lived than edges of natural gaps. Forest fragment edges are in many cases artificially maintained over time. Natural gaps, on the other





17

hand, are usually short lived. The longer the exposure to the edge, the more permanently its effects are likely to persist.
Given these marked differences in scale between gap edges and
fragment edges, the latter are apt to impinge differently on the interactions among plants and animals.













CHAPTER THREE

EDGE EFFECTS ON THE POLLINATION OF TROPICAL CLOUD FOREST PLANTS

Introduction

Pollination is an early step in a series of events leading to plant
reproduction. Changes in pollination levels are likely to cause differences in seed and fruit set, and ultimately affect the plant's distribution and the species composition of a community.
Factors that affect plant-pollinator interactions, and ultimately

pollination, are many, and complex. Here, I refer specifically to those that may derive from proximity to the edge in fragmented habitats. Pollinator behavior may be affected not only by the physical conditions near the edge (direct biological edge effects) but also by factors such as flower density, plant number, and plant neighborhood (indirect biological edge effects; Fig 1). Characteristics of the floral resource could also be responding to the physical conditions (through a direct edge effect), or be the result of reduced reproduction or recruitment on the edge (indirect biological edge effects). Thus the resulting pollination of plants near edges, could be a combination of several direct and indirect edge effects (Fig 1).
Separating the effect of edge on each component of the plant-pollinator interaction, and ultimately on the plants' pollination is not a simple task. As the first step to understanding how edges might affect plant-pollinator interactions, I measured pollination levels as a function of distance to the edge in 13 plant species in a tropical cloud forest. I aimed to override the


18





19


extremely complex factors that affect each plant species by studying many species, while trying to encompass a cross-section as wide as possible of the many plant-pollinator strategies encountered in the forest.
This study is an attempt to answer a key question on edges and their effects: Do physical and direct biological effects translate into effects on the interactions between plants and their pollinators? In this specific case, assuming that the physical conditions on the edge affect the species involved in the plant-pollinator interactions studied: do these effects on plants and their pollinators generate cascading effects that penetrate all levels of the interaction, resulting in measurable departures from conditions typically found in the absence of edges? Because of the complexity of the plantpollinator system I did not attempt to explain the mechanisms behind such responses, rather I concentrated on describing the patterns of intensity and variability of pollination at the community level.
Pollination can be measured at different points, e. g. pollen deposition, pollen tube growth, ovule fertilization. For this study, I measured pollination as the number of pollen tubes that reach the base of the style. This variable provides an accurate estimate of the true potential for fertilization when different breeding systems are involved, because it is not affected by the presence of incompatible pollen, as in self-incompatible or distylous species. I hypothesized that pollination levels near the edges would differ from the pollination levels found in the interior of the forest. Differences in pollination levels associated with the edge would indicate that edge effects on the plants and/or on the pollinators are also reflected in their interaction.
Absolute differences in pollination levels may not be biologically
meaningful. I measured two variables that indicate whether edge effects on the pollination level have the potential to translate into an effect on the




20


EDGE

I Physical effects


Physical Environment I


Direct biological effects


Plant densities <4 A


Pollination, fruit


imal activity nd densities


set, seed dispersal


Figure 1. Types of possible edge effects. Arrows indicate the direction of effects.


0


I Pollination, fruit





21


reproductive output of the plant. In other words, how deep can the edge effect penetrate the reproductive potential of plants through its effect on the plantpollinator interaction. The two variables are the intra-individual coefficient of variation (hereafter CV), and the proportion of flowers in an individual that received enough pollination for seed set (hereafter Per50).
Pollination levels are bound to vary among the flowers of each
individual plant (Feinsinger et al. 1987). This intra-individual variation (CV) in pollination levels can be increased by sporadic or erratic visitation to the flowers caused by low pollinator abundances (Zimmerman 1980), or by medium sized disturbances (Feinsinger et al. 1987). Large variation in the pollination levels of flowers may result in highly unpredictable seed set among the fruits of an individual plant. In addition, it may result in a reduced fruit set, because some flowers may receive enough pollen to fertilize all ovules and still have a surplus, while others may receive pollen loads that are insufficient for fruit set, or no pollen at all. I hypothesized that intraindividual variation would be higher near the edge. I assumed that higher intra-individual variation could result from low densities of forest pollinators near the edge, and/or visitation of the flowers by species from the open areas, which might be less effective as pollen vectors.
A plant's reproductive potential depends on maternal resources (Bawa
and Beach 1981, Whelan and Goldingay 1986), pollination levels (Whelan and Goldingay 1989), or on both (Zimmerman and Pyke 1988). Studying the effect of edges on maternal resources as a limiting factor for fruit set is beyond the scope of this study. Instead, I concentrate on the pollination levels as a way to assess how far the edge effect can penetrate the reproductive dynamics of plants through the plant-pollinator interaction. To assess whether the effect of edge on the plant-pollinator interaction ultimately affects the reproductive





22


potential of the plant (at least in terms of the female function) the amount of pollen tubes that reach the base of the style must be put in the context of the number of ovules that are to be fertilized. If the number of pollen tubes greatly exceeds the number of ovules, such that seed and fruit set are not pollen-limited (Stephenson 1981), then differences in pollination levels caused by an edge effect may not have a major consequence on the plant's reproductive potential in the short term (but see Winsor et al. 1987). If, on the other hand, flowers receive fewer pollen tubes than the number of ovules available for fertilization, then pollination becomes a potential limiting factor to fruit set. To standardize for differences among plant species, I used the number of tubes equal or larger than half the number of their ovules (Per5O) as an indicator of the minimum pollination required by a flower for setting fruit. I used this index to determine the potential for fruit set for individual plants. The potential of an individual for fruit set would then be the proportion of flowers in that individual that received a number of tubes equal or larger than half the number of ovules. I hypothesized that differences in pollination levels between edge and interior plants would translate in differences in potential for fruit set.
I also explored the temporal variation of edge effects on pollination levels of plants. Pollination levels are known to vary within (Busby 1987) and between flowering seasons (Laverty 1992), as a result of variation in the numbers of conspecific flowers (Busby 1987, Laverty 1992, Rathcke 1983) and changes in the abundance of flowers of others species (Feinsinger et al. 1986) (Rathcke 1983). I assessed the temporal consistency of edge effects on pollination levels by measuring the edge effect on pollination levels of several species at different times during their flowering seasons, and in different flowering seasons. The consistency of edge effects should be an indication of





23


the relative importance of edges and their pervasiveness in determining any differences in pollination levels between individuals beyond any natural variation in pollination levels.
Pollinators' behavior, and consequently pollination levels may be influenced by the density of conspecific flowers (Feinsinger et al. 1991, Thomson 1981, Thomson 1983) or by the physical conditions associated with the edge. To discriminate whether changes in pollination levels were caused by direct biological edge effects (response of the pollinators to the physical environment) or indirect biological edge effects (response of the pollinators to flower densities) I carried out two field experiments. The field experiments were designed to measure any differences in pollination levels between the edge and the interior given equal flower densities. The underlying assumption is that given the same resource offer on the edge and the interior, any differences in pollination between the two areas would be result from the effect of the edge's physical conditions on the pollinators.
I also analyzed the distribution of the plants in this study in relation to the edge. This analysis allowed me to assess if differences in pollination levels could be the result (at least partially) of responses of pollinators to differences in the abundances of plants. The abundances of the plants could depend on their interaction with the physical environment (direct edge effect) or differential reproductive success or seed dispersal near the edge (indirect edge effect).
Methods
Study Site
This study was conducted between May 1990 and November 1991 at a cloud forest located in southwestern Colombia, 20 km west of the city of Cali (76038" W, 3030" N). The area lies on the eastern slope of a low pass (2100





24


masl) in the Cordillera Occidental, overlooking a 20 km wide inter-Andean valley. The piedmont of the Cordillera, up to 1700 m in elevation, has been completely deforested and now is covered with highly degraded pastures that are subject to periodic fires, with some patches of second growth along the ravines. Above 1700 m an archipelago of cloud forest remnants covers the mountain tops along the ridge. The cloud forest is in the Lower Montane Very Wet Forest life zone (bmh-MB) according to the Holdrige classification (Espinal 1968). The area receives 2000-4000 mm of precipitation a year, distributed in a bimodal pattern leaving a prolonged dry season in JuneAugust, and a short dry season from December to March. The average monthly temperature fluctuates between 12 and 18*C. During the dry season, temperature fluctuations during the day are caused largely by valleymountain winds. Between 10:00 and 14:00, winds push warm air up from the valley, but in the afternoon they bring down cooler moisture-laden air, cloaking the forest with clouds. During the rainy season, the skies are more or less permanently covered with clouds, with little sunshine and less variable diurnal temperatures.
The area was gradually cleared for small farms during the first half of the century, and in the decades of 1950 and 1960 many houses were built. Currently, the land is used mainly for summer houses that often include gardens, orchards and some pastures or cut feed for horses. In spite of the population growth in the area, many of the forest remnants and their boundaries have remained at their current size and location for at least the past 30 years, as indicated by maps and personal recounts. The forest has been maintained mainly to protect sources of water, and except for occasional extraction of top soil, tree ferns, vines, and moss, the forest is not disturbed. Hunting is rare since very few large mammals and birds remain in the





25


fragments. Since their creation the forest fragments have been fenced; so there has been no interference by livestock. The forest remnants are relegated mainly to the steeper slopes and the mountain tops.
The local topography varies from small plateaus to steep slopes, with a typical inclination of ca. 45 degrees. Forest soils exhibit an A profile 40 cm thick of a sandy clay loam with a pH of 5.4 (Kattan et al. 1984; pers. observ). The B profile is a red laterite that surfaces readily in eroded areas. These soils support a vegetation 15-25 m high, depending on the terrain's inclination.
The canopy in the fragments is composed of trees in the families Moraceae (Ficus spp., Cecropia spp.), Lecythidaceae (Eschweilera U.), Rubiaceae (Ladenbergia a.), Clusiaceae (Clusia spp., Chrysoclamys spp.), Bombacaceae (Spirotheca U.), Lauraceae (Ocotea =., Nectandra .), and Solanaceae (Solanum macrocpa). In the subcanopy there are representatives of the families Rubiaceae (Palicourea spp., Ladenbergia m.) Myrtaceae (Myrcia W.), Anacardiaceae (Tapirira a.), Annonaceae (Guatteria =.), Sapindaceae, Brunelliaceae (Brunellia =.), Proteaceae (Panopsis =.), Flacourtiaceae, Meliaceae (Guarea U.), plus two palms (Prestoea W. and Geonoma W.) and tree ferns (Cvatheaceae spp.). The woody understory contains shrubs in the families Solanaceae, Rubiaceae, Boraginaceae, Piperaceae, Melastomataceae, Monimiaceae, and two palms (Aiphanes . and Chamaedorea .). The herbaceous understory is composed of 60% seedlings of trees (Kattan et al. 1984), plus plants in the families Gesneriaceae, Araceae, Campanulaceae, Piperaceae and Solanaceae. Epiphytes (bromeliads, orchids, ferns and mosses) cover all strata, but are most abundant in the upper vegetation. Vines are not common in this forest and the individuals present are thin and short. Forest parameters like





26


biomass, canopy height, and basal area are comparable to those found in other neotropical cloud forests (Kattan et al. 1984, and references therein). A more detailed description of the vegetation composition and forest structure appears in (Kattan et al. 1984).
Many hardwood trees that typically occur at this elevation (e.g. Aniba ., Cedrella U.) are missing from this forest presumably because they were selectively logged at the beginning of the century. Currently, all logging within the remnants is prohibited and there are no indications of tree removal during the past 20 or 30 years.
The transition between forest and pasture or lawn is very sharp. Fences are located against the base of the trunks of trees on the edge, and lawns and pastures extend to the fences. At these edges, the tree crowns have grown over pastures and lawns, sometimes with overhanging branches reaching down as low as two meters from the ground, in a typical cantilevered edge (Ranney et al. 1981). Thus, the edges are sealed by these overhanging canopies and a palisade of small stems behind the fence. Tree height is usually 10 m on the edge, increasing gradually towards the interior. Vines and tree-fall gaps do not seem to occur more frequently on the edge than in the interior (personal observation).
Sampling Scheme
I selected one edge in each of three fragments: San Antonio (300 ha), San Pablo (75 ha), and Torremolinos (470 ha) (Fig. 2). In each fragment, I selected a stretch of 200 m long along the edge, and defined three sampling zones relative to the forest edge. Starting at the trunk of exposed forest trees I defined the "edge" (0-10 m into the forest), a "transition zone" (10-50 m) and the "interior" (50-250 m). I defined the transition zone because the penetration of the edge effect reported in the literature falls anywhere






























Figure 2. Map of the study area. Edges sampled are indicated in the three fragments with a thicker line and an arrow. SA = San Antonio, SP = San Pablo, and TM = Torremolinos. HatoViejo is off the map, 10 km to the south.






28


BUENAVENTURA





















20 Ho1000 mn IF




10 CoLOMBIA
FStudy SAl
tea








CALI (15 kin)





29


between 10 and 50 m, depending on the variable measured (Williams-Linera 1990; Kapos 1989; Lovejoy et al. 1986; MacDougall and Kellman in press).
In each fragment, when possible I sampled from 10 individuals per plant species per zone. Due to differences in local plant densities and life history traits, and because not all plant species were present in all three fragments or in all three zones, I was not able to obtain this level of replication for all species. Consequently, some species were sampled from only one or two edges; and in several cases, fewer than 10 individuals of a given species were sampled in a zone. I only included in the sample, however, species that were at least present in all three zones of one fragment, with a minimum of three reproductive individuals per zone.
In addition to the three fragments in the main study area, I included samples of Guzmania multiflora that I obtained in November 1990 from a fourth site, HatoViejo, an edge of the National Park "Farallones de Cali" at 1900 m. Sampling in HatoViejo followed the protocol used in the three fragments. Table 1 summarizes the sampling scheme, with the individual plant as the sampling unit. Because all herbs and shrubs in this study were capable of resprouting from fallen branches, or of clonal growth (pers. obs.), I had no certainty that each individual sampled was genetically different from its neighbors. For the purpose of this study, however, I considered an individual any plant growing at least 1 m from its nearest conspecific.
Except for three species, I sampled 10.25 flowers per individual on different dates throughout the flowering season. The exceptions were or Guzmania multiflora with 10 flowers sampled per individual at the peak of its flowering season, and Centropogon solanifolius and Centropogon congestus with only 1- 5 flowers per individual due to their low flower production.





30


Table 1. Plant species and number of individuals sampled in each site and in each flowering season. Marked with an asterisk (*) are three instances in which repeated sampling of all individuals was carried out on two or three different days during the flowering season. The table contains the number of plants sampled each day.


San San Torre- Hato
Antonio Pablo molinos Viejo


Bromeliaceae
Guzmania multiflora 32 29 18
Campanulaceae
Centroogon congestus 35 9
Centropoogn solanifolius 38 66
Gesneriaceae
Besleria . 38
Besleria solanoides 1990 28
Besleria solanoides 1991 60
Columnea anisophyla 6 11 19
Melastomataceae
Miconia acuminifera 1990 38
Miconia acuminifera 1991* 31 50
Leandra, . 12
Rubiaceae
Psychotria hazennii 33 37 15
Palicourea obesifolia 23 38
Paicourea lancifera* 29
Solanaceae
Sola m . 1 Nov 90 29 19
Solanum W. 1 Jan 91* 26
Solanum B. 1 Nov 91 34 17
Solanum ,. 2 17





31


The sampling involved collecting the pistils of flowers that had been
exposed long enough for pollination and pollen tube growth to occur. Due to differences in the floral biology of the species, the collection time was determined individually for each species. Thus, flowers that lasted one day were collected the next morning, and flowers that lasted more than one day were collected several days after anthesis, once the stigma showed signs of senescence, i.e. browning or beginning to decompose. I preserved the styles in FAA (alcohol, acetic acid and formalin in a 9:1:1 ratio) and saved them for further processing. I counted the pollen tubes that reached the base of the style under an epifluorescence microscope, after staining the styles with an aniline-blue solution (Martin 1959). For species with less than 100 tubes per style, I counted all tubes. For species with more than 100 tubes per style, I estimated the total number of tubes per style based on subsamples. The results reported here were obtained from ca. 10600 styles of 782 individuals of 13 plant species.
The Plants
The 13 species sampled represent a partial cross-section of the life forms, pollinators and breeding systems from these cloud forests (Table 2). Dioecious species, canopy trees, and bat- and moth-pollinated species, however, could not be included in the sample because their densities were low and they did not occur often enough on the edges and transition zones for adequate sampling.
Of the species studied, two epiphytes (a vine and a tank bromeliad), three subcanopy trees, a treelet, two shrubs, and five herbaceous species comprise the sample of six plant families. Among these plants, four were selfincompatible, five were self-compatible, two were distylous and one was facultative self-compatible (i.e. self-fertilization is only feasible when loads of








Table 2. Characteristics of the plant species sampled. SI= self-incompatible, SC= self-compatible, SC*= facultative self-compatible (see text), DY=distylous. Numbers in parentheses in the ovule number column are the number of flowers sampled.


LIFE FORM


Bromeliaceae Guzmania multiflora
Campanulacaea Centropogon congestus Centropogfon solanifolius
Gesneraiceae Besleria f. Besleria solanoides Colurnnea anisophyla
Melastomataceae Miconia acuminifera Leandra an. Rubiaceae Psychotria hazennii Palicourea obesifolia Palicourea lancifera
Solanaceae Solanum a. 1 Solanum I. 2


Epiphyte


Herb Herb


Herb Herb Epiphyte

Subcanopy tree Shrub

Shrub Subcanopy tree Subcanopy tree

Treelet Herb


BREEDING SYSTEM


SI SC SC


SC* SC SC

SI


SC
DY DY

SI SI


POLLINATORS


Shortbilled hummingbirds

Longbilled hummingbirds Longbilled hummingbirds

Shortbilled hummingbirds Shortbilled hummingbirds Longbiilled hummingbirds

Large bees
Small bees

Butterflies
Shortbilled hummingbirds Shortbilled hummingbirds, Bees

Shortbilled hummingbirds, Bees Small bees


OVULE NUMBER


178(11)

2461 (4) 2632 (4)

2683 (5) 4087 (2) 623 (3)

300 (5) 272 (5)

2(10) 2 (7) 2(10)

83(18) 32(7)





33


self-pollen are high but below a threshold in the pollen load; the self-pollen tubes fail to grow down the style).
The sample includes hummingbird-, butterfly- and bee-pollinated
species. Among the hummingbird-pollinated species, Centropogon congestus and C. solanifolius have long (>30 mm) curved tubular corollas and are visited almost exclusively by hermit hummingbirds, e.g. Phaethori svrmatophorus and P. guy, and less frequently by Eutoxeres aquila. the sickle-billed hummingbird, and Schistes goffrviD, a short billed hummingbird. Columned anisophya, a vine with straight tubular corollas (= 30 mm) is visited by Coeligena coeligena, a straight long-billed hummingbird.

_Q. multiflora., Besleria solanoides, and Besleria A. were visited by short billed-hummingbirds: mainly Haplophaedia aureliae, and to a much lesser extent by Ocreatus underwoodii and Adelomia melanogenis. i. aureliae was also the main visitor to Palicourea obesifolia, and often established feeding territories that encompassed the crown of two or three conspecific trees. Psychotria hazenni was pollinated by two species of clearwing butterflies (Oleria caucana, Standinger 1885, and Pteronymia zerlina, Hewitson 1855) that began foraging on the one-day flowers at 10:00. Miconia acuminifera is a buzz-pollinated melastome that was visited mainly by large bees, while Leandra g. and Solanum =.2 were both pollinated by small halicteid bees. Finally, Solanum . 1 and Palicourea lancifera were visited by both shortbilled hummingbirds and large bees. Hummingbirds (mainly Q. underwoodii began foraging on these two plant species earlier in the day than the bees. Therefore it is likely that the pollen deposited by the hummingbirds reached the ovules before that deposited by the bees, making the hummingbirds the main pollinators, if not the only ones. It is likely, however, that pollen deposited later in the day by bees also fertilized those ovules not reached by





34


earlier hummingbird pollination, thus I am conservatively considering both birds and bees as pollinators.
Statistical Analyses and Variables
I compared pollination levels between treatments (edge, transition zone and interior) using the median number of pollen tubes that reached the base of the style. I used medians, instead of the traditionally used mean, because zeroes in this system are biologically important since they indicate absence of pollination. Use of means would obscure the zeroes if there was a single flower with some pollination. Thus, the median gives a more biologically meaningful estimate of pollination levels. I analyzed each species separately with a mixed model two-way analysis of variance, where fragment (random) and sampling zone (fixed) were the two factors. For those species present in only one fragment, I used a one-way ANOVA. I used data transformations to correct for heteroscedasticity whenever necessary. To calculate the intraindividual variation I used the coefficient of variation (standard deviation/mean pollination level of each plant).
Determining the minimum pollination levels for fruit set is not straightforward (Stephenson and Bertin 1983). The few studies available have used dose-response experiments to estimate the minimum number of pollen grains per ovule required for seed set varies from a 1:1 pollen grain/ovule ratio to a 10:1 ratio (McDade and Davidar 1984, Murcia 1990, Silander and Primack 1978, Snow 1982). In any of these species, however, the ratio of pollen tube to ovule has not been determined. Furthermore, all of the above studies have been conducted in species with fewer than 10 ovules per flower and capable of developing fruits with as few as one seed. Plants that contain high numbers of ovules require a minimum number of seeds to develop their fruits; otherwise, fruits are aborted (Bertin 1982, Stephenson 1981).





35


In this study, species varied in the number of ovules from two to ca. 4000 (Table 2). Because the number of pollen tubes required for seed and fruit set is not known for these species, I assumed that a number of pollen tubes equal to 50% the number of ovules would be the minimum required for fruit set across all plant species. For the three species of the family Rubiaceae, which have only two ovules per flower, I used the proportion of flowers with two or more pollen tubes, although some species in this family may set fruit with just one pollen tube (Busby 1987, Feinsinger et al. 1988a, Feinsinger et al. 1988b). I called this measure of minimum pollination for fruit set Per5O.
Statistical analyses for the intra-individual CV and Per5O followed the same protocol as the median. Due to the low number of flowers produced by individuals of Q. congestus and Q. solanifolius sometimes as low as one, the within-individual coefficient of variation and Per5O could not be calculated and thus are not included in these analyses.
For each of the three dependent variables (median, CV, Per5O) I carried out a Multivariate Analysis of Variance (MANOVA) to assess the overall effect of edge on pollination at a community-wide level using the 13 species sampled. For this test, I analyzed separately each edge, and only included those species with more than 5 individuals per sampling zone. I performed this analysis on the untransformed data. The robustness of different MANOVA tests to heterogeneity in the dispersion of the matrices varies according to the concentration of the structure (Barker and Barker 1984). I report for each MANOVA analysis the results of Pillai Trace tests because this is the least affected by the heterogeneity of dispersion matrices, yet still retains desirable power (Barker and Barker 1984). Wilk's Lambda tests which are perhaps the most commonly used, and the most sensitive to





36


heterogeneity of dispersion matrices, yielded very similar results to Pillai Trace tests in all cases.
Effect of Sample Size on Standard Deviations.

To assess the adequacy of sample sizes used in the study, I constructed a plot of standard deviation as a function of sample size. I constructed the plot starting with five individuals selected at random from the population sampled, and adding five randomly selected individuals each time. I considered an adequate sample size that which caused no further reduction in standard deviations.
I also explored the effect of increasing the number of pistils on the within-individual standard deviation for one species. For each of five individuals of Solanum =.1, I constructed a plot of the standard deviation as a function of the number of styles sampled, starting with five randomly selected styles, and adding five randomly selected styles from that individual each time.

Temporal Variation of Edge Effect
I explored the temporal variation of edge effect (if any) between
flowering seasons for three plant species: Solanum =. 1 (three flowering seasons), and M. acuminifera and D. solanoides (two flowering seasons each; Table 1). Because not all individuals flower in consecutive seasons the samples of different years may contain different plants. I analyzed each flowering season separately with a two-way ANOVA.
I also evaluated the consistency of edge effects during a flowering season by repeatedly sampling individuals on different dates. For Solanum Z.1 I sampled the same individuals on two dates (29 May and 6 June 1991) at the peak of their flowering season. For E. lancifera and M. acuminifera I sampled the same individuals (in one and two edges, respectively) on three





37


dates, at the beginning, in the middle and three fourths through their flowering season. I calculated the correlation coefficients between pollination levels of individuals on consecutive samples of their flowering seasons (day 1/day 2, and day 2/day 3).

Experiments
For the field experiments, I introduced two species into the forest in planned arrays and measured their pollination levels. I used one commercially available ornamental plant, Cerissa phoetida (Rubiaceae) native to South East Asia, and one native plant that grows at a lower elevation, Salvia W. (Lamiaceae). The experiment controlled for plant density and for any past experience by pollinators that could bias the animals toward certain areas in the forest as a result of prior associations. Thus, I selected plants whose flowers had no resemblance to any of the native species yet were visited by pollinators in the forest.
In each of two forests, I placed three groups of ten plants each at 0-5 m (=edge), 25-30 m (=transition zone), and 85-90 m (=interior) from the edge. I kept the densities constant by covering an area of approximately 2 X 20 m with each set of ten plants, and assigned the plants to the treatments in a stratified-random design blocking for plant size and flower number.
Both plant species are self-incompatible. Because they might have come from only a few individuals, I could not use the number of pollen tubes as a measure of pollination, as these might have reflected compatibility more than pollination. Instead I used the number of pollen grains deposited on the stigma. Given the floral morphology of both species, pollen movement between anthers and stigma in the same flower require an animal vector.
I collected stigmas of the experimental plants on three separate days. A preliminary analysis of variance showed no differences in pollination levels





38


among days, so I combined the data from the three days for each individual for the final analysis. Pollination levels were low in general. Many flowers received zero pollen grains, and as a result the median was zero in most individuals. Consequently, I used the third quartile, rather than the median, in a two-way mixed-model ANOVA with fragment (random) and sampling zone (fixed) as the two factors.
Edge Description
Physical conditions. During the long dry season of 1991 (June-July), I measured several variables to describe the physical conditions along a transect from the edge into the forest. In cloud forests, plants normally experience high relative moisture and low temperatures, and do not seem to show adaptations against effects of drought (Kapos and Tanner 1985, Medina et al. 1981). Thus, the long dry season might be the period of maximum stress (Wright 1991, Wright et al. 1992). Inside the forest, the dry season may not represent any physiological strain to the plants (Kapos and Tanner 1985). Near the edge, however, conditions are more likely to be extreme, and the long dry season would be the time when differences between the conditions of the edge and the interior are most emphasized. If the conditions at the edge fall outside the tolerance limits of the plants then, permanent changes in the vegetation associated with the edge are expected to occur.
In San Antonio and San Pablo, I set up two transects to measure
ambient temperature, air humidity, and soil moisture. The transects were located 25 m apart and ran perpendicular to the edge from 10 m outside the forest to 100 m into the forest. To minimize temporal effects on the measurements, i.e. to avoid confounding distance to the edge with time of the measurement, I took the measurements in the minimum time possible and randomized the starting station (inner or outermost).





39


In each transect I measured air temperature and humidity with a sling psychrometer at 10 m outside the forest and at 0, 10, 20, 25, 50, 75, and 100 m from the edge into the forest. Measurements were taken once a day between 10:30 and 14:00 on 5 days in San Antonio and six days in San Pablo. I measured the soil moisture with electrical resistance sensors (Watermark@). Sensors were buried 0.15 m deep, to sample soil moisture from the area of maximum understory root biomass (Becker and Castillo 1990). Sensors were located at 10 m outside the forest, and at 0, 5, 10, 15, 20, 40, 60 and 100 m into the forest. I set the gypsum-type blocks in place at the beginning of the dry season (14 June) and monitored the desiccation process for 6 weeks as the dry season progressed.
Vegetation analysis. I analyzed the spatial distribution of the 13 focal species in two fragments: San Antonio and San Pablo. In each fragment, I set up a 80 x 80 m plot with one side aligned with the forest edge. I divided each plot in 10 m bands that ran parallel to the edge starting at 0 m from the edge. On each 10 x 80 m band, I counted the number of adult individuals present for the 13 focal species. I used a Correspondence Analysis on the densities of the focal plants to determine the degree of similarity between consecutive samples along the edge to interior gradient, and any association between distance to the edge and plant species. For this procedure, I used the default algorithms to standardize the coordinates (SAS@ 1988).





40


RESULTS
Pollination

Pollination Levels
Overall, there were no significant edge effects in pollination levels in any of the three fragments (Pillai trace = 0.771, p = 0.20; Pillai Trace = 0.880, p =
0.28; Pillai Trace = 1.49, p = 0.12, for San Antonio, San Pablo, and Torremolinos respectively). Of the 17 samples (13 species) only two cases showed a statistically significant response to proximity to the edge (Table 3, Fig. 3). The direction of the effect, however, differed between the two species: Palicourea lancifera had higher pollination levels on the edge than in the interior while Besleria solanoides, examined in 1990, had higher pollination levels in the interior than on the edge. Data from a second flowering season (1991) for f. solanoides, however, failed to show a difference between edge and interior (Table 3, Fig. 3).
The purpose of this community-wide approach is to override the particularities of individual species. However, certain traits of the species may help interpretation of individual responses. E. lancifera is a case in point where the overall differences in pollination between edge and interior is perhaps a result of the distribution of flower morphs in the population. E. lancifera is one species exclusive to Torremolinos. Although equal numbers of individuals in each morph (pin and thrum) were sampled at the edge and in the transition zone, nine of the 11 interior plants found were all the same morph (pin). Thus, it is possible that the low pollen tube count in the interior plants was not the result of lack of pollinator activity but rather a lack of pollen compatibility. Many pollinators do tend to fly very short distances while foraging, thus it is very likely that most pollen came from the nearest





41


Table 3. Summary information of F-values from one- or two-way ANOVAs performed on the median pollen tube numbers of plants located in three sampling zones: edge (E), transition zone (TZ) and interior (I), in two or three fragments. Data from species that occurred in only one fragment were analyzed for effect of sampling zone using a one-way ANOVA, and thus show no fragment or interaction term in this table. Numbers in parentheses are the degrees of freedom. Data transformations are listed under species names. Results of a-posteriori multiple comparisons (Bonferroni/Dunn test) are provided in cases that showed a significant sampling zone effect.


FRAGMENT


SAMPLING
ZONE


INTERACTION


Guzmania multiflora (log x+1)


Centropogon congestus



Centropogon solanifolius (log x+2)

Besleria sp. Besleria solanoides 1990



Besleria solanoides 1991 (In x+1)


Columnea anisophyla (In x+2)

Miconia acuminifera 1990


Miconia acuminifera 1991b (Square root x)


1.450 (2,4) 0.871* (1,1) 7.95
(1,2)


(2,38)


2.379 (2,70)


1.059 (2,39)


1.364 (2,98)


.248


5.178* (2,28)
E=TZ; TZ=I; E
3.268* (2,60)
E=TZ; TZ>I; E=I


0.670
(2,4)


9.679*** (5,10)


1.630
(2,27)

0.630 (2,38)


3.003a (2,182)
E>TZ; E=I; TZ=I


SPECIES


2.473a (4,70)


.001 (1,39)


1.438 (2,98)


1.772
(4,27)


2.556 (10,182)





42


Table 3--continued


FRAGMENT


SAMPLING
ZONE


INTERACTION


Leandra sp. Psychotria hazennii Palicourea obesifolia Palicourea lancifera (In x+0.1)

Solanum sp. N 1990 (log x+1)
Solanum sp. J 1991 Solanum sp. N 1991


21.603**
(2,4)

1.380
(1,2)


0.146 (2,9)

1.5 (2,76)

.827 (2,55)


3.968*
(2,26)
EI; E>I


20.01* (1,2)


6.820
(1,2)


0.884
(2,42)

.773
(2,22)

1.834 (2,45)


Solanum sp.2 (In x+1)


4.191* (2,14)
E=TZ; TZ>I; E=I


SPECIES


0.488
(4,76)

.960 (2,55)


0.631
(2,42)


0.451 (2,45)


* p< 0.05; **p<0.01;*** p<0.001; a 0.06 < p < 0.05, b the model includes day as a nested factor under fragment.






























Figure 3. Median number of pollen tubes (+ SE) of plants in three sampling zones. Sampling zones are: medium grey = edge; light grey = transition zone; dark grey = interior., in three fragments (SA = San Antonio; SP = San Pablo; TM = Torremolinos, or Hatoviejo in the case of Guzmania). All data are untransformed.









QAA


-400 ~ I0


160
Guzmania sp. 80

0M
800


C. congestus P. hazzenn


["


C. solanifolius 400

0 Bsr & . J
1000 iBesleria sp. T _


B. solanoides 90 1000


2008


10
P. obesifolia
-5

... 10
P. lancifera
-2


-20
-10
0


B. solanoides 91 Solanum sp.1 J91 -40

[20


400 C. anisophila

200

408
M. acuminifera 90 200


128
M. acuminifera 91 60-


Solanum sp.1 D91.0
-20 20


Solanum sp.2


SA


SP


TM


SP TM


44


-10
~jItO


Leandra sp.


400-


'-I


800


500

0
9AAA -


a)
0
0


Solanum sp.1 90


1000

0


10


I,


800


-r


SA





45


neighbors which were likely the same morph. To what extent this peculiar distribution of the flower morphs in the interior is determined by the distance to the edge, or by an unrelated event is unknown at this time, since only this fragment contained enough individuals in all three zones for sampling.
Solanum =.2, another species exclusive to Torremolinos also exhibited very low pollination levels in the interior, perhaps due to a very low population density. The four individuals of Solanum. s.2 were scattered through an area of 0.5 ha, while the edge and transition zone plants were all in half that area. As with E. Lancifera. the absence of these species from other fragments prevents any generalizations, and therefore the results from these two species should be interpreted with caution.
Three other species showed significant differences in pollination levels. In all cases, however, the differences were caused by a significant departure in the transition zone, but no statistical differences were found between the edge and the interior.
Within-Individual Coefficient of Variation (CV)
Pollination levels were as variable within individual plants at the forest edge as they were in the forest interior or transition zone (Table 4, Fig. 4). The only exception was Solanum . 1, which in 1990 showed significantly higher intra-individual coefficients of variation in the interior than at the edge. The samples in the analysis of Solanum sp. 1, however, were heteroscedastic and an appropriate transformation could not be found, thus, this result must be interpreted with caution. Community wide, plants showed no significant edge effects for any of the three edges (Pillai Trace = 1.05, p = 0.32; Pillai Trace = 0.82, p = 0.17; Pillai Trace = 0.12, p = 0.82; for San Antonio, San Pablo and Torremolinos, respectively).





46


Table 4. As in Table 3, Summary information of F-values from one- and twoway ANOVAs performed on the coefficients of variation of within-individual pollination levels


FRAGMENT


SAMPLING
ZONE


INTERACTION


Guzmania multiflora Besleria sp. Besleria solanoides 1990 Besleria solanoides 1991 Columnea anisophyla Miconia acuminifera 1990 Miconia acuminifera 1991b Leandra sp. Psychotria hazennii Palicourea obesifolia Palicourea lancifera Solanum sp. N 1990 (log x+1)


Solanum sp. J 1991


SPECIES


1.707
(2,4)


1.445 (2,4)


7.65** (5,10)


1.119
(2,4)

0.028
(1,2)


12.430 (1,2)


1.853 (2,70)

1.385 (2,38)

0.581 (2,28)

0.884 (2,60)

0.353
(2,26)

0.368 (2,39)

1.897 (2,182)

0.725 (2,9)

0.853 (2,71)

1.294 (2,55)

2.149 (2,26)

4.655*a (2,42)


1.478
(4,70)


1.765
(4,26)


1.514 (10,182)


1.252
(4,71)

1.078 (2,55)


1.031
(4,42)


0.132 (2,19)





47


Table 4-- continued


FRAGMENT


SAMPLING
ZONE


INTERACTION


Solanum sp. N 1991


1.391
(1,2)


Solanum sp.2


0.486 (2,45)

0.215
(2,14)


0.350
(2,45)


* p< 0.05; **p<0.01; a variances were homoscedastic and there were no appropriate transformations, thus I excluded the TZ treatment and compared only edge and interior b the model includes day as a nested factor under fragment.


SPECIES





























Figure 4. Intra-individual coefficient of variation in the number of pollen tubes (+ SE) of plants in three sampling zones. Sampling zones are: black = edge; light grey = transition zone; dark grey = interior., in three fragments (SA = San Antonio; SP = San Pablo; TM = Torremolinos, or Hatoviejo in the case of Guzmania). All data are untransformed.





49


2.00 Guzmania sp.
1.00 0.001 I0 A


0


0








I

C3


SP TM


Besleria sp.

0.50

0.00I
I B. solanoides 90
0.50

0.00
0.75T
SB. solaniodes 91
0.50
0.25 0.00
2.00- C. anisophila

1.00 0.00
1.50
I M. acuminifera 90
1.00 -a 0.50
0.00
2.00 M. acuminifera 91


0.00
2.00 Leandra sp.

1.00

0 00


1.50
P. hazzennii
-1.00
T is 0.50

0.00
P. obesifolia 2.00 1.00 0.00
2.00 P. lancifera
-1.00




0.00 Solanum sp. 1 90 -2.00 1.00 0.00 Solanum sp. 1 J91 1.50 1.00
-0.50
-l- - 0.00
2.00 Solanum sp. 2 N91
1.00

0.00
Solanum sp. 2 2.00


100


SA





50


Proportion of Flowers Receiving a Number of Tubes Equal or Larger than Half the Number of Ovules (Per50)
M. acuminifera, D. solanoides (1990, 1991), and Solanum U. (November 1990) exhibited a significant difference between the edge and the interior in the proportion of flowers that received enough pollination for half seed set (Per5O; Fig. 5, Table 5). Two of these four cases, however, showed higher values in the interior, while the other two showed higher values on the edge. Furthermore, the effect or its direction were not consistent from season to season. M. acuminifera, for example, showed a significant difference in 1991, but none the previous year. D. solanoides, also showed a marked inconsistency. In 1990 it had significantly higher values on the interior, but in the following year values on the edge were significantly higher (Fig. 5). Besleria solanoides (1990) was the only case that showed a significant edge effect on the median pollination levels (Fig. 3), and a significant edge effect on its potential for fruit set (Fig. 5). The three other cases, M. acuminifera, D. solanoides (1991) and SolanumW. 1 (Nov 90), with significant differences in their potential for fruit set showed no differences in the pollination levels.
For this variable, the edge did have an effect on the community as a whole in one of the three edges (Pillai Trace = 1.08, p = 0.28; Pillai Trace =
0.995, p = 0.02; Pillai Trace = 0.673, p = 0.37; for San Antonio, San Pablo, and Torremolinos respectively). The significant edge effect on the Per5O of plants in San Pablo reflects the strong edge effects on E. hazennii and D. solanoides (1990). Both species showed a significant reduction in pollination levels near the edge compared to those in the interior in the San Pablo edge (F2,22 = 4.836, p = .018, and F2,22 = .01, for E. hazennii and D. solanoides (1990) respectively). In the case of E. hazennii, however, this significant difference was not consistent among the three edges (Table 5).





51


Table 5. As in Table 3, summary information of F-values from one- and twoway ANOVAs performed on the proportion of flowers with pollination levels equal or higher than 50% the number of ovules (Per50)


FRAGMENT


SAMPLING
ZONE


INTERACTION


Guzmania multiflora Besleria sp.


Besleria solanoides 1990 Besleria solanoides 1991 Columnea anisophyla Miconia acuminifera 1990 Miconia acuminifera 1991b


Leandra sp. Psychotria hazennii (arcsine x) Palicourea obesifolia Palicourea lancifera


17.989**
(2,28)
E=TZ; TZ
4.414**
(2,60)
E=TZ; TZ=I; E>I


0.167
(2,4)


10.861***
(5,10)


0.746 (2,4)


2.626
(1,2)


0.738 (2,27)

2.108 (2,39)

1.836 (2,182)

0.682
(2,10)

2.009 (2,78)


0.347 (2,54)


6.357**
(2,26)
E=TZ; TZ>I; E=I


SPECIES


1.475 (2,4)


1.054 (2,71)

1.203
(2,41)


1.882
(4,71)


0.872
(4,27)


1.176 (10,182)


2.804 (4,78)


0.732
(2,54)





52


Table 5--continued


FRAGMENT


SAMPLING
ZONE


INTERACTION


Solanum sp. N 1990


2.144 (1,2)


Solanum sp. J 1991 Solanum sp. N 1991


3.200
(1,2)


Solanum sp.2


3.640*
(2,42)
E=TZ; TZ=I; E

0.119
(2,22)

0.521 (2,49)

0.977 a (1,10)


* p< 0.05; **p<0.01; *** p

SPECIES


2.811
(4,42)


1.132
(2,49)






























Figure 5. Proportion of flowers in a plant with a number of pollen tubes > half the number of ovules (+ SE) of plants (Per50) in three sampling zones. Sampling zones are: medium grey = edge; light grey = transition zone; dark grey = interior., in three fragments (SA = San Antonio; SP = San Pablo; TM = Torremolinos, or Hatoviejo in the case of Guzmania). All data are untransformed.





54


80


Guzmania sp. P


40 - Besleria sp. 20-


_ _


4.2

-4
04E






C.) V-4
0
04


-1-4



$-4


'A


hazzennii 100

50


P. obesifolia
50


P. lancifera


40


B. solanoides 91 Solanum sp.1 90


40. 201 10


20


I-


C. anisophila


500
100
M. acuminifera 90 50

0


Solanum sp.1 J914U

20


Un


~IrT


M. acuminifera 91 Solanum sp.2


40 . 20.


Leandra sp. 50-


SA


Solanum sp.1 D91
20


0


- w

-40 J0T

TM


SP


SP TM


40 ] B. solanoides 90

20-i


----' ---


-in


I


v


108 I


SA





55


Effect of Sample size on Standard Deviations
In the four species analyzed (one included two flowering seasons),
standard deviation among individuals remained constant with sample sizes larger than 10 individuals (Fig 6). One case of five showed not a decrease, as expected, but an increase in standard deviations as sample size increased beyond 45 plants.
I found no changes in the within-individual standard deviation for sample sizes larger than 5 flowers (Fig 7). Temporal Variation of Edge Effect
Significant changes in pollination levels occurred in some plant species during consecutive flowering seasons, including changes in the intensity and direction of the edge effect (Figs. 3,4,5). To more finely dissect the sources of this temporal variation I considered the extent to which variation was caused by a) sampling different individuals in consecutive seasons, and b) the time of sampling during the flowering season. This allowed me to determine whether the edge effect was more pronounced during a certain time in the flowering season, and whether differences between flowering seasons could have been the result of sampling during a certain time in one flowering season, and during another time in the next.
Individual plants did not flower in all flowering seasons. In fact, as
many as 80% of the individuals sampled in one season could not be sampled during the next mainly because they did not produce flowers. Thus, the fluctuation in pollination levels between seasons could be attributed to the contribution of different sets of individuals to the sample. To explore this possibility, I used data from Solanum A. 1, for which many of the same individuals flowered in two consecutive seasons: June 1991 and November 1991. I used the median pollination levels of those individuals to test





56


800600

400200


0 10 20 30 40 50 60
Number of plants









Figure 6. Standard deviation of the number of pollen tubes as a function of the number of individual plants collected from one fragment, for four species and two seasons of one species: Open squares and triangles: D. solanoides (1990, 1991); closed squares: . anisophyla; open circles: M. acuminifera: closed circles: Solanum aR. 1.





57


40-


0


A


30


20-


10


0


10


20


30


Number of styles








Figure 7. Standard deviation of the number of pollen tubes as a function of the number of styles collected from five individual plants of Solanum W. 1 in June 1991.


5 1 Z 1111114 I I





58


whether, in spite of overall inter-seasonal fluctuations in pollination in a population, the pollination received by one individual was predictable from season to season. I found no significant correlation between the median number of pollen tubes received by individual plants in the two flowering seasons (r = 0.390,n = 19, p=0.08). Thus, a plant that received high pollination levels in one season, relative to the population, did not necessarily receive high pollination levels on the next.
Within a single flowering season, however, pollination levels of
individuals were more predictable. In Solanum m.1, M. acuminifera (1991) and P. lancifera, correlations between samples of the same individuals taken in consecutive weeks of the flowering season were significantly positive in five of seven cases (Table 6). Thus, individuals that one day received high pollination levels relative to the rest of the population were very likely to receive high pollination levels a week later. Experiments
Analyses of variance showed no significant differences in the number of pollen grains (third quartile) received by plants that were located on the edge and those located in the interior for either _. Rhoetida or Salvia V. (F2,48 =
0.178, p>0.8 and F2,51 = 0.771, p>0.4, respectively, Fig. 8). No significant interaction or effect of fragment was observed. In both cases, it was necessary to carry out a (Log +1) transformation to correct for a large heteroscedasticity.

Edge Description
Physical conditions. Above-ground physical conditions measured during the long dry season (June-August) of 1991 showed no edge effects. Both air temperature and air moisture were uniform in the first hundred meters of the





59


Table 6. Correlation coefficients between pollination levels of individuals on different days of their flowering season. Days 1, 2 and 3 correspond to dates in consecutive order, one week apart (see methods). Values in parentheses are the number of plants sampled. For Solanum . 1 I collected only on two days.


Day 1/Day 2


Day 2/Day 3


Solanum =. 1 (June 1991) 0.490 * (22)

Miconia acuminifera (1991)
SP 0.154 (22) 0.317 * (22)
SA 0.532 * (20) 0.493 * (22)

Palicourea lancifera 0.565 ** (22) 0.213 (25)

* p<0.05; ** p<0.01





60


Edge TZ Interior Edge TZ Interior


San Antonio


San Pablo


Fig 8. Pollination levels of plants of Ceriss phoetida and Salvia =. Circles indicate the mean of the average pollination levels of 10 plants located in each sampling zone: edge, transition zone (TZ) and Interior in each of two fragments, San Antonio and San Pablo. Lines connect values corresponding to the three sampling zones in each fragment. Vertical bars indicate one standard deviation. The figures show untransformed data.


24


C. phoetida


a)
0



0


12




0
12


8


4.


Salvia sp.


0


4


I


i---





61


forest in both edges (Fig. 9). Ten meters outside the forest, conditions were only slightly different: air temperature was higher, but air moisture was similar to that inside the forest.
Below ground, physical conditions were different. At the beginning of the dry season, soils were completely saturated at all points along the transects. As the season progressed, soil desiccated faster near the edge than in the interior, and in spite of a few scattered rains during the dry season, the desiccation process was cumulative. On the last sampling date, six weeks into the dry season and 37 days since the last hard rain, soil moisture was much lower over the first five and ten meters of the forest of San Pablo and San Antonio, respectively than elsewhere in the interior (Fig. 10). Vegetation analysis. The correspondence analysis for San Antonio
isolated the samples of the first 10 and 20 meters from all others (Fig. 11). Axis 2 showed a significant correlation with the soil moisture measured on July 19, 1991 (r = 0.859, n = 5, p<0.05), yet no correlation was apparent with soil moistures recorded earlier in the dry season. Two species were associated with the edge samples: F. obesifolia and M. acuminifera, the two subcanopy trees. A third species, _. solanifolius showed an intermediate position with respect to the edge and interior samples (Fig 12). For San Pablo, on the other hand, the correspondence analysis did not discriminate among samples in relation to their distance to the edge. One species, M. acuminifera, was spatially related with the 10 and 20 m samples. These two, however, were close to the most interior samples. Vegetation samples were not related to soil moisture.





62


0 20


40 60


80 100


&
0 a) '.4



a)

I


2221201918-


17


-20 0 20 40 60 80


100


120


Distance to the edge (m)


Figure 9. a) Air moisture and b) air temperature measured on 26 July 1991 at different distances from the edge. Data represent the mean values of two transects ( SD) in each of two fragments: Open squares = San Antonio and closed circles = San Pablo.


0.9 T


a) a)

a) 5.4 '-I


0.8
0.7

0.6
0.5

0.4
0.3

0.2-


14;


0.1


-1


-20


1 '


0


9J e

0 @ee


9.R





63


0

| -0.03. -0.06-0.09-0.12-0.15 , . .
-20 0 20 40 60 80 100 120

0


m -0.04. -0.08-0.12-0.16-0.2-T
-20 0 20 40 60 80 100 120


Distance to the edge (m)



Figure 10. Soil water tension recorded on 26 July 1991, six weeks into the dry season and 38 days since the last rain, for San Antonio and San Pablo. Measurements were taken at 15 cm deep, along two transects, indicated by the two symbol types, located perpendicular to the edge from 10 m outside the forest to 100 m into the forest in two fragments.





64


San Antonio 1- 80
0

Cq 0.5030

0 70 0 50 10
0 0
-0.5- 60
200

1
-1 -0.5 0 0.5 1 1.5 2 2.5

1.5
San Pablo
1 30
0
0.5
40
70
0--0--50
0 0

20
- 0 . 5 - 8 0 1 0 0


-1
-1 -0.5 0 0.5 1 1.5 2


Axis 1
Figure 11. Correspondence analysis on the species composition of samples taken at different distances from the edge in two fragments (SA, SP). Numbers close to the points represent the distance (m) from the edge of the inner most limit of the 10 x 80m bands (see methods). In San Antonio, axis 1 and axis 2 account for 72.7% of the total variation ( 54.0% and 18.7%, respectively). In San Pablo, the two axes account for 71.3% of the variation (49.8% and 21.4%) respectively.





65


'I-


1


0.5


Iq


0


-0.5


-1


1.5


1


0.5


Iq


0


-0.5


-1


-0.5


I


0.5


1


1.5


2


Axis 1

Figure 12. Correspondence analysis on the species composition of samples taken at different distances from the edge in two fragments (SA, SP). Letters close to the points correspond to the species name: Gm: Guzmania multiflora; Cc: Centropogon congestus; Cs: Centropogon solanifolius; Ca: Columnea
asophyla; Bs: Besleria solanoides; Bsp: Besleria a.; Ma: Miconia


Ph
San Antonio



Gm

Bsp

S 0 Ma
Cs 1

Cc PO

- -0.5 0 0.5 1 1.5 2 2.5


0 1
Ca San Pablo




Gm
Cc 0
- Bs
Sge Bsp
Poe


*Mal


-t * I


"





66


acuminifera; Po: Palicourea obesifoai; Ph: Psychotria hazennii; Sg: Solanum AR1.
Discussion
Proximity to the edge does not consistently affect the quantity or
consistency of pollination of plants in the cloud forests near Cali, Colombia. I found distance to the edge to be associated with reductions or increases in pollination levels of plants in only a few species, and a lack of consistency of these differences between flowering seasons. These results suggest that the edge effect is not always as pervasive and consistent as the literature has described (Kapos 1989, Lovejoy et al. 1986, Malcolm 1991). Previous studies have found that conditions at the edge are drastically different from those in the interior, and that the distance into the forest that these conditions are altered varies anywhere between 10 to 50 m. There are several possibilities why the results of this study are different. Study Design
This study was designed to test the edge effect with a factorial design, with the factor "distance to the edge" divided in three categories, and a second factor that allowed for replication on several edges. This design is very powerful but its relevance is contingent upon assigning the categories correctly. It could be argued that the distances that defined the three sampling zones were not scaled appropriately for edge effects that actually occurred. Therefore, I examined a-posteriori the pollination levels of plants as a function of their distance to the edge. For each species and each edge, I plotted the median number of tubes of each individual as a function of its distance to the edge. I visually inspected these plots searching for break points in the distribution of values that would indicate discrete subgroups of values at distances from the edge different from the ones used here. In none





67


of the cases, did I find any reason to consider the distances I originally selected as inappropriate.
As with any other statistical test, the power of these analyses is
dependent on the sample size and the variation among samples. Pollination levels are highly variable (Feinsinger et al. 1986, Herrera 1988). In this study, substantial variation in pollination levels was observed among individuals, as well as among flowers in an individual. In 34 of 87 cases, standard deviations around mean pollination levels were as large as the means in each sampling zone (Fig. 3). Such variation in pollination levels among individuals was paralleled by the variation in pollination levels within individuals. Intra-individual coefficients of variation were in general higher than 0.7.
It is likely that large standard deviations resulted from small sample sizes either in the number of plants sampled or in the number of flowers sampled per plant. Analyses indicated that standard deviations of the plants sampled are apparently representative of the standard deviations of the population, and are not an artifact of small sample sizes. Finally, I used the median as a variable to estimate pollination levels of individuals, instead of the mean, which is more commonly used by other pollination studies. Thus, it could be argued that medians are not as sensitive as means in detecting differences in pollination levels that would be biologically meaningful. To explore this possibility, I performed correlation analyses between the mean and the median for each species, considering each flowering season separately, but combining the results of all edges in each season. The correlation coefficients for all 17 samples ranged from 0.796 to 0.965, suggesting that the results were not an artifact of the variable selected to measure pollination levels.





68


The Plant-Pollinator Interaction System
Because plant-pollinator interactions are generally not species-specific (Feinsinger 1978, Feinsinger 1983, Feinsinger 1987, Roubik 1989 p 320, Waser 1978, Waser 1983), the plant-pollinator interaction may be an inherently robust system. The plant species in this study did not engage in any apparent species-specific interactions with any particular animal species, nor did they show extreme morphological adaptations that would permit exclusive flower visitation by one species. Rather, plants were visited by at least two species in the same taxonomic order, and for some plant species the list of potential pollinators spanned several phyla. Even . congestus. and Q. solanifolius, with their highly modified corollas, were visited by at least four different hummingbird species.
This lack of species specificity, although potentially or theoretically
costly in terms of interference competition for pollination among plant species (Feinsinger 1987, Waser 1983), may have a positive effect in the face of perturbations or in environments where the nectar supply is highly variable. If a pollinator avoids the edge due to its physical conditions, the plants at the edge could still receive visits from other less sensitive species that could make up for any reduction in visitation from former pollinators. Or, some pollinators from the surrounding matrix could enter the forest and visit plants close to the edge. The result would be different pollinator assemblages servicing subsets of a plant's population. Although different pollinator assemblages could produce similar pollination levels on the edge and the interior, behavioral differences between forest and edge pollinators may cause changes in the neighborhood size for edge plants (Levin and Kerster 1974, Schmitt 1980), and in the genetic composition of the pollen delivered





69


(Schmitt 1980). Those effects would be more subtle than simple changes in pollen tube numbers, and were beyond the scope of this study.
In this study, however, plant species not only had several potential
pollinators, but most of the pollinators were generalists in their diet. Thus it is unlikely that the above scenario would apply. Among pollinators, food generalists tend to be tolerant to environmental variation within the forest including the conditions caused by natural perturbations, as well as the conditions encountered in the canopy (Feinsinger et al. 1987, Feinsinger et al. 1988b, Murcia 1987, Roubik 1989 p. 324, Stiles 1975). Therefore, it is more likely that the same pollinators are visiting both the plants on the edge and in the interior, and there is no reason to expect sub-population structuring.
A previous study showed that flower visitation by forest dwelling
hummingbirds is not affected by small natural forest disturbances (mainly treefall gaps), or medium-sized clearcuts (Feinsinger et al. 1987). Edges, in the case of this study, resulted from large scale perturbations, and as such may create conditions that are too inhospitable for forest-dwelling hummingbirds.
The lack of variation in the above-ground physical condition, tied in with casual observations and previous experience with these hummingbirds (Murcia 1987) indicates that this is not the case. Six plant species in this study are pollinated by either HapIophaedia aurelje or Ocreatus underwoodii, two short billed hummingbirds. Neither of these hummingbirds ventures outside of the forest into neighboring gardens (Murcia 1987), but both species have been observed visiting the flowers on the exposed face of the edge (pers. obs.; G. Kattan, pers. comm.). Amazilia saucerottei, a hummingbird from open areas, occasionally entered the forest in San Antonio (Kattan, unpubl. data ), but was only seen associated with a stream. Never





70


did I or other researchers in this area, or in Monteverde (Costa Rica) see A. saucerottei foraging even 1 m inside the forest (G. Kattan, P. Feinsinger, pers. comm).
One set of plants whose pollination levels may be more susceptible to their proximity to edge are those with long, tubular corollas. Phaethornis srmatophorus and P. guy, two hermit hummingbirds that pollinate . congestus and _. solanifolius are restricted to the understory (pers. observation). Furthermore, no other long-billed hummingbird occurs in the area outside the forest, and due to the complex architecture of the flowers, it is unlikely that other non-hermit hummingbird visited the edge plants (with the exception of Schistes geoffroyj which is also a forest dweller). Given that for both plant species the pollination levels were similar on the edge and the interior, one can infer that the hermit hummingbirds visited both sets of plants indiscriminately.
Very little is known on the natural history of highland tropical
hymenopterans, and thus it is premature to predict how conditions on the edge would affect their foraging behavior (Roubik 1989), and ultimately the pollination of plants. Many hymenoptera, however, have been reported as rather vagile and capable of moving across open spaces while foraging (Raw 1989). Also, bees show variation in their spatial distribution in the forest depending on the season. Studies have reported tropical forest bees concentrating on the canopy during the dry season, and moving to the understory during the wet season (Roubik 1989). These results suggest a wide tolerance for biotic and abiotic conditions.
E. hazennii, the only butterfly-pollinated plant in this study was visited exclusively by clearwing butterflies (Ithomiinae), which are found more often in moist and dark parts of the understory (DeVries 1987). In spite of the





71


reported habitat specificity of its pollinators, E. hazennii did not show an edge effect on any of the pollination variables measured. One possible explanation for this lack of effect is the scarcity of F. hazennii on the exposed face of the edge. All plants occurred at least 1 m into the forest, where they were often shadowed by the overhanging canopies of the trees at the edge. Edge Effect on Physical Conditions
Although many of the pollinators in this study have been reported as behaviorally flexible, these pollinators could shy away from the edge under harsh physical conditions like those reported for other edges (Kapos 1989, Williams-Linera 1990b). I found, however, that in these fragments, the conditions were not drastically different between the immediate vicinity of the edge and the interior. Air temperature and moisture were quite uniform over the first 100 m into the forest, and similar to the conditions 10 m outside the forest. Other studies have found differences in air temperature as high as 3-4.5*C between the edge and the interior, but also the transition from these conditions to the interior took place over 15-20 m (Kapos 1989, WilliamsLinera 1990b) respectively. To my knowledge, this is the first study published on edge effects in cloud forest edges, and although it is presently unknown how typical this pattern is for cloud forests, it is very unlikely that environmental conditions interfere with pollinator movement between edge and interior in this study site.
Contrary to the apparent absence of edge effects on the physical
conditions above-ground, below-ground physical conditions did vary with proximity to the edge. At the edge, soils were drier than anywhere along the transects, and the low soil moisture extended into the first 5 or 10 m at SP and SA, respectively. This suggests that environmental conditions on the edge are more likely to affect the pollinator-plant interaction through their





72


effect on the survival of the plants than through their effect on the pollinators' movements. Differential plant survival due to different tolerances to the edge conditions would, in turn, influence the composition of vegetation near the edge, and ultimately pollination levels. In this study, the two edges showed no consistent edge effect in the vegetation composition. One the edge at SA showed a difference in the distribution of plants as a function of the distance to the edge, most likely as a result of changes in soil moisture. In spite of the vegetation being different in the first 20 m, no consistent edge effects in pollination were found in that fragment.
The two-fold reduction in soil moisture found at the edge of SP does not seem to be associated with high temperatures or low air humidity outside the forest, as would be expected from the observed physical effect of recently created edges. In this case, it is likely the result of high evapotranspiration rates associated with high foliar densities, that can result from an increase in light availability at the edge. This drastic reduction in soil moisture, however, seems to have very little or no impact on the vegetation. Oosting and Kramer (1946) also reported a similar reduction in soil moisture at the edge (0 m into forest) although the reduction was not as abrupt as in this study. Time Since Edge Creation
The short distance over which physical conditions associated with the edge disappeared in these fragments is very likely a result of the age of the edges, perhaps more than their location in a cloud forest. Most previous studies of edge effect in the tropics have been carried out along young edges, i.e. less than 5 years old (Kapos 1989, Malcolm 1991, Williams-Linera 1990a). These studies describe drastic changes in the area adjacent to the edge in the physical conditions and in the composition of the fauna and flora subsequent to the creation of the edge. These drastic changes seem to be exacerbated





73


when fire occurs after the creation of the edge (Hester and Hobbs 1992, Lovejoy et al. 1986).
The drastic changes observed on newly created edges, however, might
have led to overestimates of the projections in time of the persistence of these edge effects. To what extent these changes persist in the forest after a few years and continue modifying it, is a question that only recently has begun to be addressed. Two studies have addressed the changes of edge effect over time. In Panama, a study of five edges ranging in age from 10 months to 12 years found that the edge effect on canopy cover and basal area decreased in intensity and penetration into the forest in older edges (Williams-Linera 1990). Similar results were found by another study that examined the edge effect in edges 2-15 years old (Blanchard 1992). In the Ocala National Forest (Florida,USA), edge effects on light intensity and soil temperature were less intense in the 15 year old than in the two year old edges. In addition, a study on 20 year old riparian forest edges, found that the effect of edge on light levels disappears within 12 m into the forest, and any differences in vegetation structure caused by the edge conditions were restricted to this narrow band (MacDougall and Kellman in press). MacDougall and Kellman (in press) suggest that the rapid disappearance of the edge effect into the forest is a consequence of the age of the edge. Their suggestion seems to be supported by the results presented here, and by (Blanchard 1992) and (Williams-Linera 1990). Overall, these four studies indicate that the development of vegetation that seals the forest edge, and the closure of the canopy above and beyond the edge into the open space, play a major role in stabilizing the physical conditions near the edge. This closure of the edge is a consequence of fast growth rate and seedling recruitment spurred by higher





74


light availability (Gysel 1951, Ranney et al. 1981, Sork 1983, Williams-Linera 1990a).
Conclusions
The results of this study imply that even though the concept of edge effect is heavily inscribed in the minds of conservationists, we know very little about the biological consequences of edges. Existing studies are very diverse in the type of forest studied, time since edge creation, variable measured, and latitudinal position. All these parameters are likely to influence the results, and thus, at this point general patterns remain unclear. The results of this study indicate little or no edge effect on plant-pollinator interactions in the edges studied, perhaps as a result of little change in the environmental conditions associated with the edge. It is possible that these results are a consequence of the age of the edges, and that soon after the edges were created the effect was more drastic. It is unknown at this point the extent of the forest resilience to changes in the forest dynamics caused by edges.












CHAPTER FOUR
Toward a Unified Theory of Edges
Contrast as a Determinant of Edge Effects
The current literature on edge effects yields no clear patterns. Studies report detrimental, beneficial or no edge effects (chapter 2). Many variables seem to be involved in this apparent discrepancy among results: edge types, edge age, edge history, edge compass orientation, variables measured, and types of vegetation separated by the edge. There appear to be as many or more kinds of edges as there are researchers.
Without a more unified vision of edges, however, the conservation value of studies in edge effects is limited to the specific circumstances in which the data were collected. In other words, what we know about edge effects from one site may not be applicable for that same site if the matrix surrounding the fragments changes; or may not be applicable to a nearby site if the vegetation types separated by the edge are different from those at the first site. Could we extrapolate from other studies to predict what might happen in those fragments? I think that is unlikely.
In trying to reconcile the empirical information currently available, I propose a model as an initial step towards finding the underlying factors causing variation in responses to edges. This model is based on contrast. Contrast was defined as the difference between two habitats, and the abruptness of the transition between the two (Kuchler 1973). For example, a boundary between a pasture and a mature forest has a higher contrast than one between a pasture and young secondary growth. Also, boundaries


75





76


between two vegetation types may differ in contrast if the transitions differ in abruptness.
This concept of contrast was used by wildlife managers to assess the need for management and creation of edges for game (Thomas et al. 1979). More recently other authors have considered contrast in the context of edge effects (Angelstam 1986, Ratti and Reese 1988, Wienset al. 1985, Yahner et al. 1989), but have not defined the term in their studies, or attempted to explore its general applicability as a predictor of edge effects. Here, I define contrast more precisely and explore its general applicability as a predictor of edge effects in forest-pasture boundaries. I define contrast as the difference in the values of the variables that modulate an edge effect, between two points in the immediate vicinity of each side of the edge (points A and C, in figure 13). I define contrast only in terms of those variables that modulate each individual factor that has the potential to cause an edge effect, and not for the habitat or ecosystem as a whole.

I restrict the measurement of contrast to a few variables at a time
because the permeability of an edge to an edge effect depends only on some variables at a time. Those variables can be abiotic or biotic. Abiotic variables can be related to the vegetation structure, e.g., height, leaf and stem density, and aspect of the face (dripline, advancing or cantilevered (Ranney et al. 1981). These components of vegetation structure act as modulators of abiotic factors such as light, wind, and temperature, and so ultimately determine the strength of physical edge effects. Leaves, for example, intercept incident light that reaches the face of the forest edge. Thus with increasing leaf density at the edge less light penetrates the forest. Leaves also reduce air temperature through evapotranspiration. Thus in edges with high leaf density the difference in air temperature between the edge and the interior occurs over a




77


Pasture


A B C


tt


Forest


I


Figure 13. Diagram illustrating a forest-pasture edge. A and C are points from the pasture and the forest (respectively) in the immediate vicinity of the edge. B is the point of edge creation and maintenance. I is a point in the interior of the forest, unaffected by the edge. d, is the distance the edge effect penetrates.


It




78


shorter distance. Biotic variables, on the other hand, may involve species interactions, e.g., predation, brood parasitism, competition, and mutualisms. Biotic variables are perhaps harder to quantify than abiotic because they must be assessed from the point of view of the species in question (or the biotic factor with the potential to cause an edge effect) .
Using the traditional definition, previous authors had predicted that the intensity of edge effects increases with the contrast between adjacent habitats (Angelstam 1986, Thomas et al. 1979). In the model I present, I modify this prediction at very high contrasts, where the intensity of edge effects decreases (Fig 14). Although edges between different vegetation types yield different contrasts, contrast is not a static parameter. As vegetation structure on the edge changes over time, for example, the contrast between the two zones changes. This, in turn, results in a change in the intensity of edge effects. Therefore edge effects are dynamic, and dependent on the contrast, which is also dynamic. For example, in a recently created forestpasture edge there is a large difference in vegetation height, and in leaf and stem density. At this initial stage, the contrast in leaf and stem density between the pasture and a zone of forest closest to the edge is intermediate (contrast between points A and C, in Fig. 13). Distance d, in figure 13, is the distance that the edge effects penetrate into the forest, and is a measure of the intensity of edge effects. Point E represents the forest interior and its contrast with point A is maximum. To illustrate these effects, consider what happens to a physical condition such as light. Light penetrates a distance d, but its intensity and quality changes as it approaches point C as it is intercepted by leaves and stems. Light penetration is likely to cause changes in the conditions near the edge. What happens to the edge after its creation will determine whether the contrast between points A and C increases or





79


b1
0


U


Low


High


Contrast




Figure 14. General model of intensity of the edge effect (= distance into the forest) on forest next to anthropogenic clearings, as a function of the contrast between the variables that modulate that edge effect.





80


decreases. As a result of high light availability, vegetation growth is spurred and a leaf flush occurs. As leaf and stem density increase, light penetration decreases. Thus, the segment d decreases, eventually approaching zero, i.e., the intensity of the edge effect is decreasing. This , in turn, results in an increase in the contrast between points A and C. That is, conditions at C are returning to the conditions at E, the forest interior. In this case, edge effects brought up by an intermediate contrast resulted in a shift in the contrast towards the high end of the contrast axis in figure 14, and a further reduction in edge effects.
Some forests however, may not move sufficiently to the right on the contrast axis, and they will experience continuously high physical edge effects. Chen et al. (1992), for example, have reported that northwestern rainforest in Washington State (USA) do not develop a thick understory near the edge. Consequently, microcimatic patterns are changed up to 240 m into the forest. By comparison, several other studies on old edges are only changed in the first 20 m (Blanchard 1992, Kapos 1989, MacDougall and Kellman in press, Oosting and Kramer 1946).
Wind is another factor that may cause changes in the edge that result in changes in contrast between points A and C in Fig. 13. In a new forestpasture edge, where there is a difference in contrast in vegetation height between points A and C. Winds may create turbulence as they encounter the forest barrier. This turbulence may topple down those trees closest to the edge (Lovejoy et al. 1986). Consequently, the effect of wind on tree mortality is high and it can penetrate a distance d. With time, dead trees may be replaced by shorter individuals that offer less contrast in height between points A and C, forming a less abrupt transition between pasture and forest,




81


resulting in reduced wind edge effects. In this case, the edge effects may cause a shift in the contrast axis towards low contrast.
Because edge effects cause changes in the vicinity of the edge, those
changes necessarily result in shifts in the contrast. The strongest the effects, the faster the shift would be either towards low or high contrast. As edges approach high or low contrast, the intensity of effects decrease and the shift along the contrast axis decreases. These trends in shifts along the contrast axis suggest that intermediate levels of contrast are unstable while the extremes are most stable.
Biological edge effects also seem to follow the model proposed. At
intermediate levels of contrast, biological effects should also be highest. If there is little contrast between the two habitats, species should have similar adaptations to the physical and biological environment. In a fragment of mesic forest interacting with tall second growth forest, for example, physical effects may be mild. Species from each habitat may freely cross the edge, and their impact on the other habitat will likely be minimal. If, on the other hand, the two ecosystems are drastically different, then species may avoid crossing the edge. Avoidance may result from the edge acting as a physical barrier or from the unsuitable biological environment of the other habitat. At intermediate levels of contrast, species may cross the boundary and have an impact by disrupting processes. Brood parasitism is a case in point.
Cowbirds are typical of open areas with some vegetation cover. In the temperate zone, cowbirds (Molothrus atr) enter the forest fragments in the temperate zone and parasitize nesting birds (Brittingham and Temple 1983, Gates and Gysel 1978). The fairly open understory may provide an intermediate contrast between open areas and temperate forests, and their effect on the forest birds is large. In the tropics, on the other hand, although





82


cowbirds (M. bonariensis) parasitize nesting Henicorhina leucophrys located in edges of secondary growth, cowbirds are never seen in nearby mature forests or forest fragments. Active search of Ii. leucophrs nests have yielded no cowbird eggs inside forests (Kattan and Murcia, unpubl. data). Because cowbirds locate their hosts' nests visually (G. Kattan unpubl. data), the high contrast produced by thick vegetation may prevent cowbirds from finding nests in the forest edge (Gates and Gysel 1978). The high leaf density in high contrast edges is also likely to reduce the searching efficiency of visually oriented predators as well. In edges with low contrast, and high visibility, the impact of both cowbirds and predators should be less because those edges tend to sustain fewer nesting birds (Chasko and Gates 1982).
How changes in biological interactions caused by the edge feed back into the system to cause further changes in the contrast, is unknown at this point. Few studies have explored the patterns of edge effects on some biological processes, but none have explored the mechanisms behind these patterns.
The model I propose is a first attempt to unify concepts. This model is
general and simple. Its heuristic value lies in recognizing that although edges and edge effects are not simple, identifying underlying principles can help us distill our knowledge of edges into a general theory. Because studies published thus far vary substantially in methodology and design, the validity of this model can not be fully tested at this point using the available literature.

Future directions
Perhaps the biggest difficulty in interpreting the results of published
papers lies on the scant description of the study sites, and of the criteria used for determining where the edge is located. Ranney et al (1981) identified two important concepts with respect to the position of the edge: the point of edge





83


creation and the point of edge maintenance. These two factors determine the aspect of the edge, which can act as a modulator of edge effects. In Gates and Gysel (1978), for example, the three edges sampled had different aspects which reflected on the data. In addition, the edge was measured farther out than in many other studies, thus, their conclusions may differ from others' (Ratti and Reese 1988).
The description of the area is also important in assessing whether the portrayed edge effects are independent from other factors associated with landscape heterogeneity that could be causing changes as well, for example, ravines, steep slopes, waterlogged soils. Additional difficulty in interpreting the results stemmed from the lack of replication or the prevalence of pseudoreplication. Few studies incorporate replication into their designs, and thus, conclusions sometimes can be very limited in scope. Carefully designed studies on a diversity of edges and biological systems should allow further exploration of the mechanisms involved in edge effects, in particular, those mechanisms that attenuate the detrimental effects.













LITERATURE CITED


Angelstam, P. 1986. Predation on ground nesting birds' nests in relation to
predator densities and habitat edge. Oikos 47: 365-373

Barker, H. R. and B. M. Barker. 1984. Multivariate Analysis of Variance
(MANOVA). The University of Alabama Press. Alabama, USA.

Bawa, K. S. and J. H. Beach. 1981. Evolution of sexual systems in flowering
plants. Annals of the Missouri Botanical Gardens 68: 254-274

Becker, P. and A. Castillo. 1990. Root architecture of shrubs and saplings in
the understory of a tropical moist forest in lowland Panama. Biotropica
22: 242-249

Bertin, R. I. 1982. Floral biology, hummingbird pollination and fruit
production of trumpet creeper (Campsis radicans; Bignoniaceae).
American Journal of Botany 69: 122-134

Blake, J. G. and W. G. Hoppes. 1986. Resource abundance and microhabitat
use by birds in an isolated east-central Illinois woodlot. Auk 103: 328340

Blanchard, J. D. 1992. Light, vegetation structure, and fruit production on
edges of clearcut sand pine scrub in Ocala National Forest, Florida.
University of Florida. M. Sc. thesis.

Blouin, M. S. and E. F. Connor. 1985. Is there a best shape for nature
reserves? Biological Conservation 32: 277-288

Bradshaw, F. J. 1992. Quantifying edge effect and patch size for multiple-use
silviculture- a discussion paper. Forest Ecology and management 48:
249-264

Brittingham, M. C. and S. A. Temple. 1983. Have cowbirds caused forest
songbirds to decline? BioScience 33: 31-35

Burger, G. V. 1973. Practical Wildlife Management. Winchester Press. New
York, New York, USA.

Busby, W. H. 1987. Flowering phenology and density dependent pollination
success in Cephaelis elata (Rubiaceae). University of Florida. Ph.D.
Dissertation.


84





85


Chasko, G. G. and J. E. Gates. 1982. Avian habitat suitability along a
transmission-line corridor in an oak-hickory forest region. Wildlife
Monographs 82: 1-41

Chazdon, R. L. 1986. Light variation and carbon gain in rain forest
understorey palms. Journal of Ecology 74: 995-1012

Chen, J., J. F. Franklin and T. A. Spies. 1992. Vegetation responses to edge
environments in old-growth Douglas-fir forests. Ecological Applications
2: 387-396

Correll, D. L. 1991. Human impact on the functioning of landscape
boundaries. pages 90-109. in M. M. Holland; P. G. Risser and R. J.
Naiman (editor). Ecotones: The role of landscape boundaries in the
management and restoration of changing environments. Chapman &
Hall. London, England.

Dagg, A. I. 1976. Wildlife Management in Europe. Otter Press. Waterloo,
Ontario, Canada.

Dasmann, R. F. 1964. Wildlife Biology. J. Wiley and Sons, Inc. New York,
New York, USA.

DeVries, P. J. 1987. The butterflies of Costa Rica and their natural history.
Princeton University Press. Princeton, New Jersey, USA.

Diamond, J. M. 1975. The island dilemma: Lessons of modern geographical
studies for the design of habitat reserves. Biological Conservation 7: 129146

Etherington, J. R. 1982. Environment and plant ecology. John Wiley and
Sons. Chichester, United Kingdom.

Feinsinger, P. 1978. Interactions between plants and hummingbirds in a
successional tropical community. Ecological Monographs 48: 269-287

Feinsinger, P. 1983. Coevolution and pollination. pages 282-311. in D.
Futuyma and M. Slatkin (editor). Coevolution. Sinauer Associated Inc.
Boston, Massachusetts, USA.

Feinsinger, P. 1987. Effects of plant species on each other's pollination: Is
community structure influenced. Trends in Ecology and Evolution 2:
123-126

Feinsinger, P., J. H. Beach, Y. B. Linhart, W. H. Busby and K. G. Murray.
1987. Disturbance, pollinator predictability, and pollination success
among Costa Rican cloud forest plants. Ecology 68: 1294-1305

Feinsinger, P., W. H. Busby, K. G. Murray, J. H. Beach, W. Z. Pounds and Y.
B. Linhart. 1988a. Mixed support for spatial heterogeneity in species





86


interactions: Hummingbirds in a tropical disturbance mosaic. American
Naturalist 131: 33-57

Feinsinger, P., W. H. Busby and H. M. Tiebout III. 1988b. Effects of
indiscriminate foraging by tropical hummingbirds on pollination and
plant reproductive success: Experiments with two tropical treelets.
Oecologia (Berlin) 76: 471-474

Feinsinger, P., K. G. Murray, S. Kinsman and W. H. Busby. 1986. Floral
neighborhood and pollination success in four hummingbird-pollinated
cloud forest plant species. Ecology 67: 449-464

Feinsinger, P., H. M. Tiebout III and B. E. Young. 1991. Do tropical birdpollinated plants exhibit density-dependent interactions? Field
experiments. Ecology 72: 1953-1963

Fetcher, N., S. F. Oberhauer and B. R. Strain. 1985. Vegetation effects on
microclimate in lowland tropical forest in Costa Rica. International
Journal of Biometeorology 29: 145-155

Forman, R. T. T. and M. Godron. 1986. Landscape Ecology. John Wiley &
Sons. New York, New York, USA.

Gates, J. E. and L. W. Gysel. 1978. Avian nest dispersion and fledging success
in field-forest ecotones. Ecology 59: 871-883

Geiger, R. 1965. The climate near the ground. Harvard University Press.
Cambridge, Massachusetts, USA.

Gibbs, J. P. 1991. Avian nest predation in tropical wet forest: An
experimental study. Oikos 60: 155-161

Gysel, L. W. 1951. Borders and openings of Beech-Maple woodlands in
southern Michigan. Journal of Forestry 49: 13-19

Harris, L. D. 1988. Edge effects and conservation of biotic diversity.
Conservation Biology 2: 330-332

Herrera, C. 1988. Variation in mutualisms: The spatio-temporal mosaic of
pollinator assemblage. Biological Journal of the Linnean Society 35: 95125

Hester, A. J. and R. J. Hobbs. 1992. Influence of fire and soil nutrients on
native and non-native annuals at remnant vegetation edges in the
western Australian wheatbelt. Journal of Vegetation Science 3: 101-108

Hoppes, W. G. 1988. Seedfall patterns of several species of bird dispersed
plants in an Illinois woodland. Ecology 69: 320-329





87


Kapos, V. 1989. Effect of isolation on the water status of forest patches in the
Brazilian Amazon. Journal of Tropical Ecology 5: 173-185

Kapos, V. and E. V. J. Tanner. 1985. Water relations of Jamaican upper
montane rain forest trees. Ecology 66: 241-250

Kattan, G. H., C. Restrepo and M. Giraldo. 1984. Estructura de un bosque de
niebla en la cordillera occidental, Valle del Cauca, Colombia. Cespedesia
13: 23-43

Kevan, P. G. 1975. Pollination and environmental conservation.
Environmental Conservation 2: 293-298

Kuchler, A. W. 1973. Problems in classifying and mapping vegetation for
ecological regionalization. Ecology 54: 512-523

Laurance, W. F. and E. Yensen. 1991. Predicting the impacts of edge effects
in fragmented habitats. Biological Conservation 55: 77-92

Laverty, T. M. 1992. Plant interactions for pollinator visits--A test of the
magnet species effect. Oecologia 89: 502-508

Leopold, A. 1933. Game management. Charles Scribner's Sons. New York,
New York, USA.

Levey, D. J. 1988. Treefall gaps and the distribution of understory birds and
shrubs in a tropical wet forest. Ecology 69: 1076-1089

Levin, D. A. and H. W. Kerster. 1974. Gene flow in seed plants. Evolutionary
Biology 7: 139-220

Lovejoy, T. E., R. 0. Bierregaard Jr., A. B. Rylands, J. R. Malcolm, C. E.
Quintela, L. H. Harper, K. S. Brown, A. H. Powell, G. N. V. Powell, 0. R.
Schubart and M. B. Hays. 1986. Edge and other effects of isolation on
amazon forest fragments. pages 257-285. in M. E. Sould (editor).
Conservation biology: the science of scarcity and diversity. Sinauer
Associates. Sunderland, Massachusetts, USA.

MacDougall, A. S. and M. Kellman. in press. The understory light regime and
patterns of tree seedlings in tropical riparian forest patches. Journal of
Biogeography

Malcolm, J. R. 1991. The small mammals of Amazonian forest fragments:
Pattern and process. University of Florida. Ph.D. Dissertation.

Margules, C., A. J. Higgs and R. W. Rafe. 1982. Modern biogeographic theory:
Are there any lessons for nature reserve design? Biological Conservation
24: 115-128





88


McClanahan, T. R. 1986. Pollen dispersal and intensity as criteria for the
minimum viable population and species reserves. Environmental
Management 10: 381-383

McCoy, E. D., S. S. Bell and K. Walters. 1986. Identifying biotic boundaries
along environmental gradients. Ecology 67: 749-759

McDade, L. A. and P. Davidar. 1984. Determinants of fruit and seed set in
Pavonia dasvpetala (Malvaceae). Oecologia 64: 61-67
Medina, E., E. Cuevas and P. L. Weaver. 1981. Composici6n foliar y
transpiraci6n de especies leinosas de Pico del Este, Sierra de Luquillo,
Puerto Rico. Acta Cientifica Venezolana 32: 159-165

Moller, A. P. 1989. Nest site selection across field-woodland ecotones: The
effect of nest predation. Oikos 56: 240-246

Murcia, C. 1987. Estructura y dindmica del gremio de colibries (Aves:
Trochilidae) en un bosque andino. Humboldtia 1: 29-64

Murcia, C. 1990. Effect of floral morphology on pollen receipt and removal in
Ipomoea trichocara. Ecology 71: 1098-1109
Murray, K. G. 1988. Avian seed dispersal of three neotropical gap-dependent
plants. Ecological Monographs 58: 271-298

Oosting, H. J. and P. J. Kramer. 1946. Water and light in relation to pine
reproduction. Ecology 27: 47-53

Palik, B. J. and P. G. Murphy. 1990. Disturbance versus edge effects in sugarmaple/beech forest fragments. Forest Ecology and Management 32: 187202

Quintela, C. E. 1986. Forest fragmentation and differential use of natural
and man-made edges by understory birds in central Amazonia.
University of Illinois. M.Sc. Thesis.

Ranney, J. W., M. C. Bruner and J. B. Levenson. 1981. The importance of
edge in the structure and dynamics of forest islands. pages 57-95. in R.
L. Burgess and D. M. Sharpe (editor). Forest island dynamics in mandominated landscapes. Springer-Verlag. New York, New York, USA.

Rathcke, B. 1983. Competition and facilitation among plants for pollination.
pages 305-329. in L. Real (editor). Pollination biology. Academic Press,
Inc. Orlando, Florida, USA.

Ratti, J. T. and K. P. Reese. 1988. Preliminary test of the ecological trap
hypothesis. Journal of Wildlife Management 52: 484-491





89


Raw, A. 1989. The dispersal of euglossine bees between isolated patches of
eastern Brazilian wet forest (Hymenoptera, Apidae). Revista Brasileira
de Entomologia 33: 103-107

Romano, G. B. 1990. Invasibility of a mixed hardwood forest by Eupatorium
capijlifQlium and B. compositifolium. University of Florida. M.Sc. Thesis.

Roubik, D. W. 1989. Ecology and natural history of tropical bees. Cambridge
University Press. New York, New York, USA.

Santos, T. and J. L. Tellerfa. 1992. Edge effects on nest predation in
Mediterranean fragmented forests. Biological Conservation 60: 1-5

SAS�, S. I. I. 1988. Technical report P-179. Additional SAS/STATTM
Procedures, Release 6.03. SAS Institute Inc. Cary, North Carolina, USA.

Saunders, D. A., R. Hobbs J. and C. Margules R. 1991. Biological
consequences of ecosystem fragmentation: A review. Conservation
Biology 5: 18-32

Schmitt, J. 1980. Pollinator foraging behavior and gene dispersal in Senecio
(Compositae). Evolution 34: 934-943

Silander, J. A. and R. B. Primack. 1978. Pollination intensity and seed set in
the evening primrose (Oenothera fruticosa). American Midland
Naturalist 100: 213-216

Simberloff, D. and R. N. Gotelli. 1984. Effects of insularization on plant
richness in the prairie-forest ecotone. Biological Conservation 29: 27-46

Simberloff, D. S. and L. G. Abele. 1976. Island biogeography theory and
conservation practice. Science 191: 285-286

Snow, A. A. 1982. Pollination intensity and potential seed set in Passiflora
vitifolia. Oecologia 55: 231-237

Sork, V. L. 1983. Distribution of pignut hickory (Carva glabra) along a forest
edge transect, and factors affecting seedling recruitment. Bulletin of the
Torrey Botanical Club 110: 491-506

Soul6, M. E. 1986. The effects of fragmentation. pages 233-236. in M. E. Sou16
(editor). Conservation biology: The science of scarcity and diversity.
Sinauer Associates, Inc. Sunderland, Massachusetts, USA.

Stephenson, A. G. 1981. Flower and fruit abortion: Proximal causes and
ultimate functions. Annual Review of Ecology and Systematics 12: 253279





90


Stephenson, A. G. and R. I. Bertin. 1983. Male competition, female choice,
and sexual selection in plants. pages 109-149. in L. Real (editor).
Pollination biology. Academic Press, Inc. Orlando, Florida, USA.
Stiles, F. G. 1975. Ecology, flowering phenology and hummingbird pollination
of some Costa Rican Heliconia species. Ecology 56: 285-301

Stiles, F. G. 1985. On the role of birds in the dynamics of Neotropical forests.
pages 49-59. in A. W. Diamond and T. Lovejoy (editor). Conservation of
tropical forest birds. Paston Press. Norwich, United Kingdom.

Terborgh, J. 1976. Island biogeography and conservation: Strategies and
limitations. Science 193: 1028-1029

Thomas, J. W., C. Maser and J. E. Rodiek. 1979. Edges. pages 48-59. in J. W.
Thomas (editor). Wildlife habitats in managed forests: The Blue
Mountains of Washington and Oregon. U. S. Forest Service Handbook
No. 553. U.S. Government Printing Office. Washington, D. C., USA.

Thomson, J. D. 1981. Spatial and temporal components of resource
assessment by flower-feeding insects. Journal of Animal Ecology 50: 4959

Thomson, J. D. 1983. Component analysis of community-level interactions in
pollination systems. pages in C. E. Jones and R. J. Little (editor).
Handbook of experimental pollination biology. Van Nostrand Reinhold.
New York, New York, USA.

Wales, B. A. 1972. Vegetation analysis of north and south edges in a mature
oak-hickory forest. Ecological Monographs 42: 451-471

Waser, N. M. 1978. Interspecific pollen transfer and competition between cooccurring plant species. Oecologia (Berlin) 36: 223-236

Waser, N. M. 1983. Competition for pollination and floral character
differences among sympatric plant species: a review of evidence. pages in
E. C. Jones and R. J. Little (editor). Handbook of experimental
pollination biology. Van Nostrand Reinhold Company. New York, New
York, USA.

Whelan, R. J. and R. L. Goldingay. 1986. Do pollinators influence seed set in
Banksia paludosa Sm. and B. spinulosa R. Br.? Australian Journal of
Ecology 11: 181-186

Whelan, R. J. and R. L. Goldingay. 1989. Factors affecting fruit-set in Telopea
peioiim (Proteaceae): The importance of pollen limitation. Journal
of Ecology 77: 1123-1134





91


Wiens, J. A., C. S. Crawford and J. R. Gosz. 1985. Boundary dynamics: A
conceptual framework for studying landscape ecosystems. Oikos 45: 421427
Wilcove, D. S., C. H. McLellan and A. P. Dobson. 1986. Habitat fragmentation
in the temperate zone. pages 237-256. in M. E. Soul6 (editor).
Conservation Biology. The science of scarcity and diversity. Sinauer
Associates, Inc. Sunderland, Massachusetts, USA.

Wilcox, B. A. 1980. Insular ecology and conservation. pages 95-117. in M. E.
Soul and B. A. Wilcox (editor). Conservation ecology: An evolutionaryecological perspective. Sinauer Associates. Sunderland, Massachusetts,
USA.

Williams-Linera, G. 1990a. Origin and early development of forest edge
vegetation in Panama. Biotropica 22: 235-241

Williams-Linera, G. 1990b. Vegetation structure and environmental
conditions of forest edges in Panama. Journal of Ecology 78: 356-373

Williamson, M. 1975. The design of wildlife preserves. Nature 256: 519

Willson, M. F. and F. H. J. Crome. 1989. Patterns of seed rain at the edge of a
tropical Queensland rain forest. Journal of Tropical Ecology 5: 301-308

Winsor, J. A., L. E. Davis and A. G. Stephenson. 1987. The relationship
between pollen load and fruit maturation and the effect of pollen load on
offspring vigor in Cucurbita 9Z&. American Naturalist 129: 643-656

Wright, S. J. 1991. Seasonal drought and the phenology of understory shrubs
in a tropical moist forest. Ecology 72: 1643-1657

Wright, S. J., J. L. Machado, S. S. Mulkey and S. A. P. 1992. Drought
acclimation among tropical forest shrubs (Psychotria, Rubiaceae).
Oecologia 89: 457-463

Yahner, R. H. 1988. Changes in wildlife communities near edges.
Conservation Biology 2: 333-339

Yahner, R. H., T. E. Morrell and J. S. Rachael. 1989. Effects of edge contrast
on depredation of artificial avian nests. Journal of Wildlife Management
53: 1135-1138

Yahner, R. H. and A. L. Wright. 1985. Depredation on artificial ground nests:
Effects of edge and plot age. Journal of Wildlife Management 49: 508513
Yoakum, J., W. P. Dasmann, H. R. Sanderson, C. M. Nixon and H. S.
Crawford. 1980. Habitat improvement techniques. pages 329-403. in S.





92

D. Schemnitz (editor). Wildlife management techniques manual. The
Wildlife Society. Washington, D. C., USA.

Yoakum, J. P. and W. P. Dasmann. 1971. Habitat manipulation practices.
pages 173-231. in R. H. Giles (editor). Wildlife management techniques.
The Wildlife Society. Washington, D. C., USA.

Zimmerman, M. 1980. Reproduction in Polemonium: Competition for
pollinators. Ecology 61: 497-501
Zimmerman, M. and G. H. Pyke. 1988. Reproduction in Polemonium:
Assessing the factors limiting seed set. American Naturalist 131: 723738












BIOGRAPHICAL SKETCH


Carolina Murcia was born on 25 October 1960 in BogotA, Colombia. She graduated as a zoologist from Universidad del Valle in Cali, Colombia, in 1983. In 1984 she married Gustavo Kattan, a fellow biologist, and started graduate school at the University of Florida. She obtained her M. Sc. in Zoology in 1987 under the direction of Dr. Peter Feinsinger, working on the floral morphology of morning glories. Her Ph.D. work was also conducted under the direction of Dr. Feinsinger. Her research interests are in plant reproductive biology, plant-pollinator interactions and conservation biology.


93




Full Text

PAGE 1

EDGE EFFECTS ON THE POLLINATION OF TROPICAL CLOUD FOREST PLANTS By CAROLINA MURCIA A DISSERTATION PRESENTED TO THE GRADUATE SCHOOT OF THE m^IVERSITY OF FLORIDA IN plSX FULfK^^ OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 1993

PAGE 2

1 ACKNOWLEDGMENTS I would like to thank the members of my doctoral committee, Drs. Peter Feinsinger, Frank G. Nordlie, Douglas J. Levey, Richard A. Kiltie, Francis E. Putz and Kent H. Redford, for their advice during all phases of the project. I would Hke to thank especially Peter Feinsinger for his support and encouragement through all my graduate student years. I would also like to extend my appreciation to the following persons: Natalia Arango for her assistance and companionship during the field work phase; Ronald Edwards for graciously walking me through the nuances of IBM computing for the Correspondence Analysis, G. Kattan, R. Edwards and Dr. Carmine Lanciani for commenting on portions of earUer drafts of this dissertation; Mr. Luis Felipe Carvajal, Mr. Eduardo Calderon and Mr. Antonio Gonzalez Caicedo for permission to work on their properties; Mr. Luis Miguel Constantino for identifying some butterflies; Mr. German Parra for the identification of bees; Mr. Stinger Guala and Dr. Walter Judd for their help in determinating the plant material. Many of the ideas presented here benefitted greatly from discussions with fellow students in the Department of Zoology, to all of them I thank. Financial support for this project was provided by the Department of Zoology, University of Florida, the Underbill Foundation, and the Fundacion para la Promoci6n de la Investigacidn y la Tecnologia, Banco de la Republica, Bogot^i, Colombia. i

PAGE 3

Finally and most importantly, I would like to acknowledge the support and encouragement of my husband, Gustavo Kattan, who put up with my ups and downs with love and a smile. • • lU

PAGE 4

TABLE OF CONTENTS page ACKNOWLEDGEMENTS ii ABSTRACT vi CHAPTERS 1 INTRODUCTION 1 2 BACKGROUND 5 Landscape Ecology and Edges 5 Historical Perspectives on Edges 7 Why Is Exposure to the Edge Deletereous for the Fragments? 10 Physical Edge Effects 10 Direct Biological Edge Effects 12 Indirect Biological Edge Effects 14 Edges and Plant-Animal Interactions 15 3 EDGE EFFECTS ON THE POLLINATION OF TROPICAL CLOUD FOREST PLANTS Introduction 18 Methods 23 Study Site 23 Sampling Scheme 26 The Plants 31 Statistical Analyses and Variables 34 Effect of Sample Size on Standard Deviation 36 Temporal Variation of Edge Effect 36 Experiments 37 Edge Description 38 Results 40 Pollination 40 Effect of Sample Size on Standard Deviation 55 Temporal Variation of Edge Effect 55 Experiments 58 Edge Description 58 iv

PAGE 5

Discussion 66 Study Design 66 The Plant-Pollinator Interaction System 68 Edge Effect on Physical Conditions 71 Time Since Edge Creation 72 Conclusions 74 4 TOWARDS A UNIFIED THEORY OF EDGES Contrast as a Determinant of Edge effects 75 Future Directions 82 LITERATURE CITED 84 BIOGRAPHICAL SKETCH 93 V

PAGE 6

Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy EDGE EFFECTS ON THE POLLINATION OF TROPICAL CLOUD FOREST PLANTS By Carolina Murcia May 1993 Chairperson: Dr. Frank Nordlie Major Department: Zoology Exposure to edge is one consequence of habitat fragmentation that can result in detrimental effects on the fauna and flora of fragmented habitats. In this study, edge is defined as a sharp transition between natural and anthropogenic habitats, such as between forest and pastures. I assessed the effect of proximity to the edge on the polUnation levels of 13 cloud forest plant species in Colombia. I collected the styles of plants located in three zones of each of three forest fi-agments: edge (0-10 m), transition zone (10-50 m), and interior (50-200 m). PolUnation levels, measured as the median number of pollen tubes that reached the base of the styles, were not consistently affected in their pollination by the proximity to the edge. Intra-individual coefficient of variation, i.e. vEiriation in pollination levels among flowers in indidividuals was also not affected by the proximity to the edge. Few plant species were significantly affected by the edge, and those that were affected differed in the direction of the edge effects. To measure the edge effect on the potential for vi

PAGE 7

limitation of seed set and its penetration on successive levels of the plantpollinator interaction, I measured the proportion of flowers in each individual with a nimiber of pollen tubes eqvial to or larger than half the number of ovules. In one of two cases where there was an edge effect on pollination levels, the potential for seed set was also affected. Edge effects, when present, were inconsistent between flowering seasons. Field experiments with two introduced plant species showed no differences in pollination levels between the plants placed at the edge and in the forest interior. There are several possible explanations for these results. First, abovegroimd environmental conditions do not differ between the edge and the interior, thus making it imlikely that the polHnators are affected by the edge. Second, below-groimd conditions are affected only for the first 10 m into the forest, and this affects the densities of the focal plants in only one of the fragments. Finally, plant-pollinator interactions can be robust to some pertiu'bations because generally, these interactions have little speciesspecificity. vii

PAGE 8

CHAPTER ONE INTRODUCTION When a natiiral ecosystem is modified by humans the result is often a mosaic of isolated remnants of native vegetation interspersed with large areas of modified habitat. This process of habitat fi-agmentation has deleterious consequences for most of the native biota. First and most obviously the totsd area of the original vegetation is reduced. This reduction in area often results in species extinction. Because the process of extinction in fragments is analogous to that occurring on islands, the theory of island biogeography provided the initial conceptual fi*amework for studying this process in terrestrial ecosystems (Diamond 1975, SimberlofiF and Gotelli 1984, Simberloff and Abele 1976, Terborgh 1976, Wilcox 1980), but see (Margules, et al. 1982). In addition to a reduction in total area, fragmentation can have £in impact on the organisms that remain in the fragment through what has been termed "edge eflfects" (Lovejoy, et al. 1986, Saiinders, et al. 1991, Soul6 1986, Wilcove, et al. 1986). Edge effects result from the interaction between two adjacent habitats. In fi*agmented habitats, edge effects are exacerbated as a consequence of bidimensional geometry. With the reduction in fragment area, the area/perimeter ratio decreases. As a result, a larger proportion of the habitat isleind is exposed to the edge, or line of contact with the other habitat. The intensity of edge effects is strongest in the immediate vicinity of the edge, and declines away fi-om the edge (Blanchard 1992, Gates and Gysel 1978, Hester and Hobbs 1992, Kapos 1989, Ranney, et al. 1981, Wales 1972,

PAGE 9

2 Williams-Linera 1990a, Williams-Linera 1990b). In addition to distance from the edge, several factors seem to mitigate the strength of the effect: edge age (Blanchard 1992, Williams-Linera 1990b), aspect (Gysel 1951, Ranney et al. 1981), orientation (Kapos 1989, Palik and Miuphy 1990, Ranney et al. 1981, Wales 1972), and also the type of vegetation on both sides of the edge (Chasko and Gates 1982). Attention to edge effects has focused mainly on the fragmentation of \. forests and their replacement with simple agricultural or pastoral ecosystems. The general pattern of changes for forest-field edges that can be extracted from the Uterature indicates that forest edges tend to have higher solar incidence than the interior of the fragments (Blanchard 1992, Kapos 1989, Costing and Kramer 1946). Consequently, air temperatures near the edge are usually higher and more variable (Kapos 1989, Williams-Linera 1990a), and soils tend to be dryer (Kapos 1989, Costing and Kramer 1946) thein in the forest interior. With higher light availability, plant growth is enhanced. Therefore edges tend to have higher foliar densities (Blanchard 1992, Malcolm 1991), and higher plant densities (Chen, et al. 1992, Palik and Murphy 1990, Ranney et al. 1981, Sork 1983, Wales 1972, Williams-Linera 1990a), in spite of higher tree mortality caused by exposvire and wind throw (Chen et al. 1992, Lovejoy et al. 1986). This abundance in cover and plant density attracts herbivorous insects (Blanchard 1992), and nesting birds (Chasko and Gates 1982), which in turn attract nest predators (Gates and Gysel 1978), and brood parasites (Brittingham and Temple 1983). Although organisms from outside the fragments may be attracted to the edge (Brittingham and Temple 1983, Chasko and Gates 1982, Hester £uid Hobbs 1992, Laurance and Yensen 1991), the change in the physical environment associated with edges may also cause some plants and animals to avoid the

PAGE 10

edges (Chen et al. 1992, Malcolm 1991, Ranney et al. 1981). Thus, changes associated with edges C£in have effects at many levels. At the individual level edges can affect the reproductive potential of plants (Blanchard 1992) and animals (Angelstam 1986, Brittingham and Temple 1983, Chasko and Gates 1982, Gibbs 1991, Wilcove et al. 1986); at the population level, edges can affect the presence of species near the edge (Blanchard 1992, Lovejoy et al. 1986, Quintela 1986, Ranney et al. 1981); and at the community level changes in the interaction among species can be affected as well. Little is known on the effects of edge at the community level because, with the exception of avian nest predation and brood parasitism, edge effects on processes that involve species interactions have rarely been studied. Of particular interest are plant-animal interactions (such as polUnation, frugivory, herbivory, or seed dispersal), due to their importance in the dynamics of many forests. Plant-animal interactions can be very sensitive to disturbance because of the dual effect on both the plants and the animals, and because of the speciesspecific nature of many of these interactions (Kevan 1975, McClsuiahan 1986). In this community-level study, I address how interactions among plants and their pollinators are affected by their proximity to a forest-pastiu"e edge. Because researchers perceive and define the concepts of edge and their effects differently, I begin (chapter 2) by reviewing the evolution of the concepts, and their current definitions. I confine this review to edges between forests and field/pastures, because these edges are the most relevant to the conservation of forest fragments. I also draw general patterns on edge effects at all levels of organization (individuals, populations or communities), and divide edge effects into three classes: effects on the physical environment, direct biological effects, and indirect biological effects. This context allows me

PAGE 11

4 to discuss the importance of plant-animal interactions, and in particular plant-pollinator interactions, in the dynamics of tropical forests, and their potential susceptibility to edges. In chapter 3, 1 present a study on the effect of proximity to the edge on the pollination of 13 tropical cloud-forest plant species. I foimd no overall effects on pollination at the commimity level. At the species level a few species showed edge effects, yet these effects were inconsistent between flowering seasons. This lack of edge effects was also confirmed by field experiments using two introduced plant species. Finally, edge effects on the physical environment, when present, extended only 10 m into the forest. These results do not accord well with the predictions made in the Uterature. This lack of agreement is perhaps due to two factors acting in concert: a) the intensity of edge effects varies depending on the types of ecosystems on each side of the boimdary, and age, orientation or aspect of the edge; and b) there is little agreement among researchers on what is an edge and how to measure its effects. In the final chapter, I propose a model that predicts the intensity of edge effects based on the degree of contrast between the ecosystems separated by the edge. The concept of contrast is an old one, and has been mentioned as one of the factors that may determine the intensity of the edge effect (Santos and Telleria 1992, Wiens, et al. 1985), yet it has not been formally defined and its role on modulating edge effects has been overlooked. I finish by suggesting future directions for research on edges. In this final section, I also point out common problems in the design of the studies and interpretation of the resvdts that, in my concept, may obstruct the advancement of our knowledge on edges.

PAGE 12

CHAPTER TWO BACKGROUND Landscape Ecology and Edges A landscape is an sirea composed of interacting ecosystems (Forman and Godron 1986, p. 11). In natural landscapes, a variety of ecosystems may result from heterogeneity in the spatial distribution of critical resources and distiu-bsinces. Because resources tend to vary spatially following gradients, ecosystems do not have sharp boimdaries but merge gradually (McCoy, et al. 1986). The area of contact between ecosystems is an ecotone. When boundsiries between ecosystems are abrupt, the ecotone is almost imidimensional and it is called an edge. Natural edges are rare and occur mostly as a result of abrupt changes in the soil as occurs, for example, in rocky outcrops, or serpentine soils. In contrast, edges are more prevalent in himian-made ecosystems or when those ecosystems are created and maintained next to natural vegetation. Ecotones and edges sepeirate ecosystems, but are not impermeable to fluxes. Edges may function as semi-permeable membranes that filter flows of energy, nutrients and species between ecosystems (Forman and Godron 1986, Wiens et al. 1985). Flows across these membranes cause transitory or permanent changes in the composition and abimdance of species in the area of contact. These changes are called edge effects. The intensity of an edge effect has been measured as the distance (d) into an ecosystem that these changes penetrate (Laurance and Yensen 1991). Changes due to edge effects 5

PAGE 13

6 modify the ecosystem beyond any intrinsic natioral variation of the ecosystem. In other words, the conditions at the edge occur only when an edge is present. The area near the edge that is modified by edge effects has the potential of becoming an ecotone if no fmlJier human intervention occurs (Gysel 1951). In that case, species firom both ecosystems would gradually invade the other, "smoothing" the transition between the two. If, on the other hand, the invasion of at least one ecosystem by the other is £irtificially suppressed, the result is the maintenance of an abrupt edge. The points of edge creation and edge maintenance have been used by Ranney et al. (1981) to define three different types of edges: cantilevered, canopy dripline, and advancing. Cantilevered edges are maintained at their point of creation. These edges are characterized by the overhsmging canopy of the branches that grow tow£irds the open space, and a thick understory among large tree tnmks. Canopy dripline edges are those maintained at the outer tips of the horizontal branches of the canopy trees. These edges have a dense imderstory, shaded by the branches of the outermost canopy trees, but have no large tree trunks exposed to the edge. Canopy dripline edges represent an abrupt change in height and plemt density between fields and forests. Advancing edges are msiintained several meters away fi*om their point of creation, or not maintained at all. They are characterized by a dense vegetation that gradually declines in height between the point of edge creation £ind the point of edge meiintenance. Advancing edges are not abrupt as are cantilevered or canopy dripline edges, and consequently I categorize them as ecotones rather than edges. The effects of ecosystems on one another across an ecotone are likely to differ fi-om the effects across an edge. This difference is perhaps a result of the longer distances separating ecosystems and the presence of a set of plants and emimals that are be more tolerant to the conditions of both

PAGE 14

7 ecosystems. For that reason, I will restrict the remainder of the dissertation to abrupt ecotones, i.e., cantilevered and dripline edges. The effects of ecosystems on one another across the edge may be asymmetrical depending on the relative sizes of the ecosystems. In a landscape, the ecosystem with the largest total area and highest connectivity among patches is called the matrix, and it dominates the dynamics of the landscape (Forman and Godron 1986, p. 159). Elements in the landscape that are less connected than the matrix are called patches or fragments, and are considered to be "immersed" in the matrix. Because flows of energy, nutrients and species are dominated by the matrix, edge effects caused by the contact between the fragments and the matrix are likely to affect the fragments more strongly than the matrix. This asymmetry in the effects has historically directed the interest of ecologists towards the ecosystem that experiences the strongest effects. Historical Perspective on Edge Effects In North America, wildlife managers concerned themselves with edges when they noticed that most game species thrived where forest, brush, pastures, and crops were in contact. The increase in the populations of game species at these edges was termed the "edge effect" by Aldo Leopold (Leopold 1933)(p. 131). In the interest of increasing game populations, Leopold (1933, p. 131) formulated his Law of Interspersion. It stated that "...{it is highly desirable to grant game} simultaneous access to more than one environmental type". Leopold's recommended management technique required interspersion of habitats maximizing the amoimt of edge for any ecosystem area. To m a x i m ize the area of contact between ecosystems or habitats some management manuals recommend the interspersion of fields and croplands

PAGE 15

8 with sedges and vegetation that would make an otherwise depauperate area more attractive to wildlife (Burger 1973, Yoaktun, et al. 1980, Yoakum and Dasmann 1971). Although this management technique could potentiaUy increase the are available for forest habitat, the recommendations were not always interpreted to the benefit of the forests. In some instances, for example, large expanses of forest were considered depauperate (Dasmann 1964), and consequently manuals recommended the interspersion of forest with other ecosystems at the expense of the forest (Dasmann 1964, Thomas, et al. 1979). Interestingly, European wildlife managers seem unconcerned with edges as a wildlife management technique (Dagg 1976), perhaps because the degree of fragmentation present in Europe in the last centuries has made £iny further creation of edges unnecesseuy. Foresters and silviculturalists, on the other hand, have perceived edges as detrimental, although the relative importance given to the edge effects varies between the two schools of silviculture (Bradshaw 1992). Since the Middle Ages, the school that advocates clearcutting, and the school that advocates selective felling have debated the relative importance of edges in maintaining harvestable forests. Those in favor of clearcutting claim that the technique minimizes the amoimt of forest that could be exposed to the edge, in spite of a costly reduction in the rate of recovery of the logged area (Bradshaw 1992). Advocates of selective felling, on the other hand, are aware that this technique increases the area exposed to edge and the problems associated with the creation of edge, yet favor the technique because the rates of recovery of the cleared area are higher that with clearcutting (Bradshaw 1992). As forested areas shrink faster than they can recover, the ongoing controversy between silvicultural schools is of utmost interest. As the total

PAGE 16

9 forest area decreases, the rates of recovery of the cleared land decrease due to the shift in the dynamics of the landscape. From a forest dominated landscape, where clearcuts were the fragments and were readily affected by the recolonization of the forest, the landscape has changed to one dominated by clearcuts. In the latter scenario, fluxes of energy and species are dominated by the clearcut matrix to the detriment of the forest fragments. As a result of this shift, it is likely that the school of forestry that advocates clearcutting may loose support, and the controversy may be resolved. In Australia and USA, for example, there is increasing pressure for eliminating clearcutting (Bradshaw 1992). With the increase in the rates of forest fragmentation, the concepts of edges and edge effects have also changed among wildlife managers. In the past two decades, the concern for game species has given way to concern about the edge effects on forest wildlife and its preservation. Consequently, the focus of wildlife managers has also switched towards non-game wildlife (Harris 1988, Yahner 1988), and their perception of edges as management tools has changed. Efforts have been made to identify the factors that determine the potential for species conservation in habitat remnants. Under the influence of island biogeography theory, the initial emphasis was on "area." as the major factor determining the conservation potential of a site (Diamond 1975). Many heated debates, however, convinced ecologists that area per se is not the only, or best ,predictor of the nimiber of species that can be supported by a particular reserve (Margules et al. 1982, Simberloflf and GotelH 1984, Simberlofif and Abele 1976). Other factors associated with fragmentation but partially independent from area, have been identified as contributing to the deleterious effect of fragmentation on species conservation (Blouin and

PAGE 17

10 Connor 1985, Lovejoy et al. 1986, Satinders et al. 1991, Soul6 1986). Among these factors is the increased exposure of the fragment species to the edge (Lovejoy et al. 1986, Saimders et al. 1991, Soul6 1986, Williamson 1975). Edge effects have been recognized qualitatively for many decades, yet quantification of these effects (with a few exceptions) has only taken place on the last 12 years (Gates and Gysel 1978, Gysel 1951, Costing and Kramer 1946, Wales 1972). This qioantification has focused on measuring the distance that edge effects penetrate the fi*agment with the ultimate purpose of determining what proportion of the fi-agment still contains typical forest conditions, and thus its potential for conservation. Why Ta Exposure to the Edge Considered Deleterious f or the Fragments? Forest fi^agmentation shifts the fluxes of energy, nutrients and species in the landscape from being forest-dominated to being dominated by the new matrix. These shifts are boimd to cause edge effects on the forest. I divide the types of edge effects in three categories. 1) Changes in the environmental conditions that result from the proximity to the matrix are physical edge effects. 2) Changes in the abundance and distribution of species caused directly by the physical conditions near the edge, e. g. desiccation, windthrow, plant growth, are direct biological edge effects. These are determined by the physiological tolerances of species to the conditions on and near the edge. 3) Changes in species interactions, e.g. predation, brood parasitism, competition, herbivory, biotic pollination, zoochorous seed disperssd, are the result of indirect biological edge effects. Phvsical Edge Effects Compared to a structursdly complex vegetation such as a forest, structiu-ally simple vegetation, like crops and pastures, aUow a higher amount of solar radiation to reach the groimd during the day, and higher

PAGE 18

11 reradiation to the atmosphere at night (Geiger 1965). Consequently, diumal temperatures tend to be higher, and daily temperatures fluctuate more widely (Fetcher et al. 1985). High temperatures cause a reduction in air moisture (Fetcher et al. 1985, Geiger 1965). Under taller and more complex vegetation, on the other hand, there are lower air temperatures and narrower temperature fluctuations due to the larger biomass that c£in absorb solar radiation (Etherington 1982, Fetcher et al. 1985) as well as lower Ught levels (Chazdon 1986). These differences between the vegetation on each side of the edge are likely to create a gradient perpendicular to the edge. Changes in the physical conditions associated with the edge show cleeir patterns. Measurements of incident light, air temperature, air moisture, soil temperature, and soil moisture at the edge (0 m) differ from the interior, displaying intermediate values between the interior of the fragment and the matrix. Differences in these factors between the edge and the interior usually disappear over a distance of 5 to 20 m in a variety of tropical and temperate zone forests (Blanchard 1992, Kapos 1989, MacDougall and Kellman in press, Costing and Krsimer 1946). The less the exposure to solar radiation, the weaker are the physical edge effects. Thus, for north or northeastern facing edges in the Brazilian Amazon and New Jersey, for example, differences between the edge and the interior were smaller than for edges facing other directions, and edge effects penetrated less into the forest (Kapos 1989, Wales 1972). Physical edge effects can also result from the movement of chemic£d compounds across the edge, that can alter environmental conditions. Chemical fertilizers from adjacent croplands can penetrate up to 5 m into the wheatbelt shrubland and up to 50 m into the wheatbelt woodland of Australia (Hester and Hobbs 1992). Also, nitrates, sulfates and herbicides from

PAGE 19

12 adjacent croplands are known to penetrate into riparian forests in Maryland (Correll 1991). In both cases, values at the edge were highest, and declined with distance into the forest fragment. Direct biological edge effects Changes in the physical environment caused by edges have different effects on the fauna and flora of the fragments depending on the org£inisms' physiological tolerances. In the case of plants, edge effects could ultimately determine their densities near the edge. A variety of responses have been observed: some plant species are never foimd near edges, while others have higher densities near the edge than in the interior, and yet others show no changes in densities as a fimction of distance to the edge (MacDougall and Kellman in press, Ranney et al. 1981). In the case of forest animals, density and activity have showed diverse responses to the edge, from avoidsuice to preference (Chasko and Gates 1982, Quintela 1986). In addition, species from the matrix may react to the edge as well. A favorable environment (Hester and Hobbs 1992), or dispersal by abiotic vectors (Romano 1990, Willson and Crome 1989) may cause species from the matrix to converge on the edge, and even penetrate some distance into the fragment. In general, individusJ plants show a positive response to the increase in incident light availability near the edge, in spite of the potential dehydration that can reach 5-10 into the forest (Kapos 1989). Leaf density (Blanchard 1992, Malcolm 1991), stem densities and basal areas tend to be higher within 20 m of the edge in a variety of tropical and temperate zone forests (Palik and Murphy 1990, Ranney et al. 1981, Wales 1972, WilUams-Linera 1990bX but see )(Chen et al. 1992). In some ecosystems such as undisturbed sugar maplebeach forests in Michigan, however, the increased plant biomass response is stronger among canopy and subcanopy trees (PaHk and Murphy 1990), while

PAGE 20

13 in second growth forests of sugar maple-beach forests in Michigan, and in the lowland rainforests of the Brazilian Amazon the response is stronger among imderstory plants (Malcolm 1991). Plant growth is also spurred by the conditions near the edge in a variety of forests in the United States and Panama (Chen et al. 1992, Sork 1983, Wales 1972, Williams-Linera 1990a). In contrast to the greening caused by increased growth and leaf production, increased mortality can also result from proximity to edge. Tree mortality can result from windthrow (Chen et al. 1992), and possibly as a result of fire following the creation of the edge (Lovejoy et al. 1986). Changes in the physical environment also result in shifts in species composition. Some plants show lower densities or are absent near the edge (Chen et al. 1992, MacDougall and Kellman in press, Ranney et al. 1981, Sork 1983, Wales 1972), while others show higher densities (Chen et al. 1992, MacDougall and Kellman in press, Ranney et al. 1981, Wales 1972), or no changes in density in association with distance to the edge (Blanchard 1992, MacDougall and Kellman in press, Ranney et al. 1981). As a result of these different responses, species composition may differ between the edge and the interior. Tree species composition, for example, differed in one study between the first 5-45 m and the interior in undisturbed sugar maple/beach forest fragments in Michigan (Pedik and Murphy 1990). Species composition in insects (Malcolm 1991), and birds (Quintela 1986) edso have shown differences between the edge (050 m) and the interior of lowland rain forests in Brazil. In other cases, however, differences in species composition do not occur. Studies on tree and seedling species composition have foimd no differences in species composition as a result of proximity to the edge in second growth sugar/beach forests in Michigan and in undisturbed lowland

PAGE 21

14 rain forests in Panama (Palik and Murphy 1990, Williams-Linera 1990a, Williams-Linera 1990b). Although some general patterns in edge effects can be extracted from the literature, direct biological effects are not as clear as those observed in the physical environment. Perhaps the most consistent responses are those spurred by the increase in Ught availability, i.e., leaf and stem growth. Any other effects are less clear, especially when idiosyncratic responses of particular species are involved. This variability in responses may result from a combination of direct and indirect biological edge effects. In other words, different responses observed could be due to some species responding to direct biological edge effects, while others are responding to indirect biological edge effects, or to both. Indirect Biological Edge Effects Changes in the distribution and abimdance of species near the edge may alter the dynamics of species' interactions near the edge. For example, a leaf flush that results from increased light incidence may attract herbivorous insects. These, in turn, may attract nesting birds, which in turn could attract nest predators and brood parasites. Thus, the indirect effect of Ught availability on herbivorous insects may initiate a series of cascading effects, that can spread across the fabric of the ecosystem. What this imphes, is that the density of one species, could be determined by its response to the physical conditions near the edge, but it could also result from its response to other species. Most studies only describe changes in densities and species compositions near the edge, but few have explored the causes for these changes. Studies that concentrate on the interactions among species may shed some light on the importeince of indirect effects on the species near the edge.

PAGE 22

15 The species interactions that have received most attention are nest predation and brood parasitism in birds. The results have been inconsistent, however. Some studies have foimd higher rates of nest predation on or near the edge in an oak-hickory forest in Michigan, a lowland rain forest in Costa Rica, and in Tennessee (Gates and Gysel 1978, Gibbs 1991, Wilcove et al. 1986). In such forests, increased rates of nest predation can occur up to 20 m into the fragments in Costa Rica (Gibbs 1991), 45 m in Michigan (Gates and Gysel 1978), and 300-600 m in Tennessee (Wilcove et al. 1986) as opposed to the forest interior. Other studies have foimd inconsistent or no significant effects in a variety of temperate zone forests in North America and Europe (Chasko and Gates 1982, Gibbs 1991, M0ller 1989, Ratti and Reese 1988, Santos and Tellen'a 1992, Yahner and Wright 1985). Avian brood parasitism in the temperate zone seems to follow the same patterns as nest predation, with higher parasitism near the edge and decUning away from it. In one study in a Wisconsin deciduous forest, cowbirds ( Molothrus ater) parasitized nests up to 300 m into the forest fragments (Brittingham and Temple 1983). In continuous forests, however, cowbirds did not normally parasitize forest interior birds. Studies on other types of species interactions are scarce. These studies have foimd lower post-dispersal seed predation (Sork 1983), higher herbivory (Sork 1983) and zoochorous dispersal of seeds from the matrix (WiUson £ind Crome 1989) to a distance up to 80 m into the forest. Rdgfts and Plant-Animal Interactinna Plant-animal interactions are crucial to the d5niamics of tropical forests. Animals are involved in the pollination, seed dispersEd, seed predation and herbivory of a large percentage of tropical species. In the highlands of Costa Rica, for example, animals disperse 70-75 % of forest plants (Stiles 1985).

PAGE 23

16 Plant-animal interactions can be very sensitive processes because disturbances can aifect the interaction by affecting the plants, the animals, or both (Kevan 1975, McClanahan 1986). Plants and their associated animals normally encoimter a certain degree of environmental variation caused by small scale perturbations in the forest, such as treefall gaps. The effects of canopy gaps on the dynamics of plantanimal interactions have been explored for severed systems, such as: plantfrugivorous birds (Blake and Hoppes 1986, Hoppes 1988, Levey 1988, Murray 1988), and plants and their pollinating himMningbirds (Feinsinger, et al. 1987, Feinsinger, et al. 1988a). Large scale perturbations such as those caused by forest fragmentation and exposure to the edge, are apt to affect plant-animal interactions. Extrapolation of potential edge effects from the literature on gaps, however, may not be realistic due to the different spatial and temporal scales involved. At the spatial scale, gaps represent very small clearings compared to those created by forest fragmentation. Therefore climatic conditions may be substantially different. A study in a lowland tropical rainforest, at La Selva, Costa Rica, for example, revealed that microclimatic conditions in a 400 m^ gap are closer to those of the forest interior than to those of a 5,000 m2 clearing (0.5 ha) (Fetcher et al. 1985). Even a 0.5 ha clearing is small when compgired to the open areas that typically surroimd forest fragments. Climatic differences between canopy gaps and the clearings that result from forest fragmentation are likely to be pronotmced. At the temporal scale, edges that result from forest fragmentation tend to be longer lived than edges of natural gaps. Forest fragment edges are in many cases artificially maintained over time. Natural gaps, on the other

PAGE 24

17 h£ind, are usually short lived. The longer the exposure to the edge, the more permanently its effects are likely to persist. Given these marked differences in scale between gap edges and fragment edges, the latter are apt to impinge differently on the interactions among plants and animals.

PAGE 25

CHAPTER THREE EDGE EFFECTS ON THE POLLINATION OF TROPICAL CLOUD FOREST PLANTS Introduction Pollination is an early step in a series of events leading to plant reproduction. Changes in pollination levels are likely to cause differences in seed and finiit set, and ultimately affect the plant's distribution and the species composition of a community. Factors that affect plant-polUnator interactions, and ultimately pollination, are many, and complex. Here, I refer specifically to those that may derive from proximity to the edge in fragmented habitats. PolHnator behavior may be affected not only by the physical conditions near the edge (direct biologiced edge eflfects) but also by factors such as flower density, plant number, and plant neighborhood (indirect biological edge effects; Fig 1). Characteristics of the floreJ resource could also be responding to the physical conditions (through a direct edge effect), or be the result of reduced reproduction or recruitment on the edge (indirect biological edge effects). Thus the resulting pollination of plants near edges, could be a combination of several direct and indirect edge effects (Fig 1). Separating the effect of edge on each component of the plant-pollinator interaction, and ultimately on the plants' pollination is not a simple task. As the first step to understanding how edges might affect plant-pollinator interactions, I measured pollination levels as a fimction of distance to the edge in 13 plant species in a tropical cloud forest. I aimed to override the 18

PAGE 26

19 extremely complex factors that affect each plant species by studying many species, while trying to encompass a cross-section as wide as possible of the many plant-pollinator strategies encoimtered in the forest. This study is an attempt to answer a key question on edges sind their effects: Do physical and direct biological effects translate into effects on the interactions between plants and their pollinators? In this specific case, assiuning that the physical conditions on the edge affect the species involved in the plant-pollinator interactions studied: do these effects on plants and their pollinators generate cascading effects that penetrate all levels of the interaction, resulting in measurable departures from conditions typically foxmd in the absence of edges? Because of the complexity of the plantpollinator system I did not attempt to explain the mechanisms behind such responses, rather I concentrated on describing the patterns of intensity and variability of polUnation at the community level. PolHnation can be measured at different points, e. g. pollen deposition, pollen tube growth, ovule fertilization. For this study, I measured pollination as the nimiber of pollen tubes that reach the base of the style. This variable provides an accurate estimate of the true potential for fertilization when different breeding systems are involved, because it is not affected by the presence of incompatible pollen, as in self-incompatible or distylous species. I hypothesized that pollination levels near the edges would differ from the pollination levels found in the interior of the forest. Differences in pollination levels associated with the edge would indicate that edge effects on the plants and/or on the pollinators are also reflected in their interaction. Absolute differences in polUnation levels may not be biologically meaningfvil. I measured two variables that indicate whether edge effects on the pollination level have the potential to translate into an effect on the

PAGE 27

20 EDGE 1 Physical effects Physical Environment Direct biological effects Plant densities Animal activity and densities Pollination, fruit set, seed dispersal a 1 Figure 1. Types of possible edge effects. Arrows indicate the direction of effects.

PAGE 28

21 reproductive output of the plant. In other words, how deep can the edge effect penetrate the reproductive potential of plants through its effect on the plantpollinator interaction. The two variables are the intra-individual coefficient of variation (hereafter C V), and the proportion of flowers in an individual that received enough pollination for seed set (hereafter Per50). Pollination levels are bound to vary among the flowers of each individual plant (Feinsinger et al. 1987). This intra-individual variation (CV) in pollination levels can be increased by sporadic or erratic visitation to the flowers caused by low pollinator abundances (Zimmerman 1980), or by medivun sized disturbsinces (Feinsinger et al. 1987). Large variation in the pollination levels of flowers may result in highly impredictable seed set among the fruits of an individusd plant. In addition, it may result in a reduced fruit set, because some flowers may receive enough pollen to fertilize all ovules and still have a surplus, while others may receive pollen loads that are insufficient for finiit set, or no pollen at all. I h3^thesized that intraindividual variation would be higher near the edge. I assumed that higher intra-individual variation could result from low densities of forest pollinators near the edge, and/or visitation of the flowers by species from the open areas, which might be less effective as pollen vectors. A plant's reproductive potential depends on maternal resoiu*ces (Bawa and Beach 1981, Whelan and Goldingay 1986), pollination levels (Whelan and Goldingay 1989), or on both (Zimmerman and Fyke 1988). Studying the effect of edges on maternal resources as a limiting factor for fruit set is beyond the scope of this study. Instead, I concentrate on the pollination levels as a way to assess how far the edge effect can penetrate the reproductive dynamics of plants through the plant-pollinator interaction. To assess whether the effect of edge on the plant-pollinator interaction ultimately affects the reproductive

PAGE 29

22 potential of the plant (at least in terms of the female fimction) the amoimt of pollen tubes that reach the base of the style must be put in the context of the number of ovules that are to be fertilized. If the nimaber of pollen tubes greatly exceeds the nmnber of ovules, such that seed £ind finiit set are not pollen-limited (Stephenson 1981), then differences in poUination levels caused by an edge effect may not have a major consequence on the plant's reproductive potential in the short term (but see Winsor et al. 1987). If, on the other hand, flowers receive fewer pollen tubes than the number of ovules available for fertilization, then pollination becomes a potential limiting factor to fruit set. To standardize for differences among plant species, I used the nimiber of tubes equal or larger than half the number of their ovules (Per50) as an indicator of the minimtim pollination required by a flower for setting fruit. I used this index to determine the potential for fiuit set for individual plants. The potential of an individual for fruit set would then be the proportion of flowers in that individual that received a number of tubes equal or larger than half the nmnber of ovtdes. I hypothesized that differences in pollination levels between edge and interior plants would translate in differences in potential for fruit set. I also explored the temporal variation of edge effects on polUnation levels of plants. Pollination levels are known to vary within (Busby 1987) and between flowering seasons (Laverty 1992), as a result of variation in the nimibers of conspecific flowers (Busby 1987, Laverty 1992, Rathcke 1983) and changes in the abimdance of flowers of others species (Feinsinger et al. 1986) (Rathcke 1983). I assessed the temporal consistency of edge effects on pollination levels by measuring the edge effect on pollination levels of several species at different times during their flowering seasons, and in different flowering seasons. The consistency of edge effects should be an indication of

PAGE 30

23 the relative importance of edges and their pervasiveness in determining any differences in pollination levels between individuals beyond any natural vfiriation in pollination levels. Pollinators' behavior, £ind consequently pollination levels may be influenced by the density of conspecific flowers (Feinsinger et al. 1991, Thomson 1981, Thomson 1983) or by the physical conditions associated with the edge. To discriminate whether changes in pollination levels were caused by direct biological edge eflfects (response of the pollinators to the physicsd environment) or indirect biological edge effects (response of the polhnators to flower densities) I carried out two field experiments. The field experiments were designed to measure any differences in pollination levels between the edge and the interior given equal flower densities. The imderlying assimiption is that given the same resource offer on the edge and the interior, any differences in pollination between the two areas would be result from the effiect of the edge's physical conditions on the polUnators. I also analyzed the distribution of the plants in this study in relation to the edge. This analysis allowed me to assess if differences in pollination levels could be the result (at least partially) of responses of pollinators to differences in the abundances of plants. The abundances of the plants could depend on their interaction with the physical environment (direct edge effect) or differential reproductive success or seed dispersed near the edge (indirect edge effect). Methods Studv Site This study was conducted between May 1990 £ind November 1991 at a cloud forest located in southwestern Colombia, 20 km west of the city of Cali (76°38" W, 3°30" N). The area Hes on the eastern slope of a low pass (2100

PAGE 31

24 masl) in the Cordillera Occidental, overlooking a 20 km wide inter-Andean valley. The piedmont of the Cordillera, up to 1700 m in elevation, has been completely deforested and now is covered with highly degraded pastures that are subject to periodic fires, with some patches of second growth along the ravines. Above 1700 m an archipelago of cloud forest remneints covers the moimtain tops along the ridge. The cloud forest is in the Lower Montane Very Wet Forest life zone (bmh-MB) according to the Holdrige classification (Espinal 1968). The area receives 2000-4000 mm of precipitation a year, distributed in a bimodal pattern leaving a prolonged dry season in JuneAugust, and a short dry season fi-om December to March. The average monthly temperature fluctuates between 12 and 18°C. During the dry season, temperature fluctuations during the day are caused lEirgely by valleymountain winds. Between 10:00 and 14:00, winds push warm air up from the valley, but in the afternoon they bring down cooler moisture-laden air, cloaking the forest with clouds. During the rainy season, the skies are more or less permsinently covered with clouds, with little sunshine and less variable diurnal temperatures. The area was gradually cleared for small farms during the first hedf of the centtuy, and in the decades of 1950 and 1960 many houses were built. Currently, the land is used mainly for siunmer houses that often include g£irdens, orchards and some pastures or cut feed for horses. In spite of the population growth in the area, many of the forest remnants and their boimdaries have remained at their current size and location for at least the past 30 years, as indicated by maps and personal recoimts. The forest has been maintained mainly to protect sources of water, and except for occasional extraction of top soil, tree ferns, vines, and moss, the forest is not disturbed. Hunting is rare since very few large mammals and birds remain in the

PAGE 32

25 fragments. Since their creation the forest fragments have been fenced; so there has been no interference by Uvestock, The forest remnants are relegated mainly to the steeper slopes and the mountain tops. The local topography varies from small plateaus to steep slopes, with a t5^ical inclination of ca. 45 degrees. Forest soils exhibit an A profile 40 cm thick of a sandy clay loam with a pH of 5.4 (Kattan et al. 1984; pers. observ). The B profile is a red laterite that surfaces readily in eroded areas. These soils support a vegetation 15-25 m high, depending on the terrain's inchnation. The CEinopy in the fragments is composed of trees in the families Moraceae (Eicua spp., Cecropia spp.), Lecythidaceae (Eachweilera ap.), Rubiaceae (Ladenbergia ac ), Clusiaceae (Clusia spp., Chrvsoclamvs spp.), Bombacaceae (Spirotheca SE-), Lauraceae (Ocotea ssi-, Necteindra sp .), and Solanaceae (Rolannm macrocarpa) . In the subcanopy there sire representatives of the families Rubiaceae (Palicourea spp., Ladenbergia SD.-) Myrtaceae ( Mvrcia an ), Anacardiaceae (Ta pirira ac ), Annonaceae (Guatteria sp.X Sapindaceae, Brunelliaceae ( Brunellia sp.). Proteaceae (Panopsis sp.), Flacourtiaceae, Meliaceae (Guarea afi ), plus two psdms (Prestoea SHand Geonoma ac ) and tree ferns (Cvatheaceae spp.). The woody imderstory contains shrubs in the families Solanaceae, Rubiaceae, Boraginaceae, Piperaceae, Melastomataceae, Monimiaceae, and two palms ( Aiphanes sp. and Chamaedorea aji.). The herbaceous understory is composed of 60% seedlings of trees (Kattan et al. 1984), plus plants in the families Gesneriaceae, Araceae, Campantdaceae, Piperaceae and Solanaceae. Epiph5rtes (bromeliads, orchids, ferns and mosses) cover all strata, but are most abvmdant in the upper vegetation. Vines are not common in this forest £ind the individuals present are thin and short. Forest parameters like

PAGE 33

26 biomass, canopy height, and basal area are comparable to those found in other neotropical cloud forests (Kattan et al. 1984, and references therein). A more detailed description of the vegetation composition and forest structure appears in (Kattan et al. 1984). Many hardwood trees that typically occur at this elevation (e.g. Aniba sp., Cedrella sq ) are missing from this forest presvimably because they were selectively logged at the beginning of the century. Currently, £dl logging within the remnemts is prohibited and there are no indications of tree removal during the past 20 or 30 years. The transition between forest and pasture or lawn is very sharp. Fences are located against the base of the tnmks of trees on the edge, and lawns and pastures extend to the fences. At these edges, the tree crowns have grown over pastures and lawns, sometimes with overhanging branches reaching down as low as two meters from the ground, in a typical cantilevered edge (Ranney et al. 1981). Thus, the edges are sealed by these overhanging canopies and a palisade of small stems behind the fence. Tree height is usually 10 m on the edge, increasing gradually towards the interior. Vines £ind tree-faU gaps do not seem to occur more frequently on the edge than in the interior (personal observation). Sampling Scheme I selected one edge in each of three fragments: San Antonio (300 ha), San Pablo (75 ha), and Torremolinos ( 470 ha) (Fig. 2). In each fragment, I selected a stretch of 200 m long along the edge, and defined three sampling zones relative to the forest edge. Stsirting at the trunk of exposed forest trees I defined the "edge" (0-10 m into the forest), a "treinsition zone" (10-50 m) and the "interior" (50-250 m). I defined the transition zone because the penetration of the edge effect reported in the literature falls anywhere

PAGE 34

Figure 2. Map of the study area. Edges sampled are indicated in the three fragments with a thicker hne and an arrow. SA = San Antonio, SP = San Pablo, and TM = Torremolinos. HatoViejo is off the map, 10 km to the south.

PAGE 36

29 between 10 and 50 m, depending on the variable measured (Williams-Linera 1990; Kapos 1989; Lovejoy et al. 1986; MacDougall and Kellman in press). In each fragment, when possible I sampled from 10 individuals per plant species per zone. Due to differences in local plant densities and life history traits, and because not aU plant species were present in all three fragments or in £dl three zones, I was not able to obtain this level of replication for all species. Consequently, some species were sampled from only one or two edges; and in several cases, fewer than 10 individuals of a given species were sampled in a zone. I only included in the sample, however, species that were at least present in all three zones of one fragment, with a minimimi of three reproductive individuals per zone. In addition to the three fragments in the main study area, I included samples of Guzmania miiltiflora that I obtedned in November 1990 from a fourth site, HatoViejo, an edge of the National Park "Farallones de Cali" at 1900 m. Sampling in HatoViejo followed the protocol used in the three fragments. Table 1 summarizes the sampling scheme, with the individual plant as the sampling imit. Because all herbs and shrubs in this study were capable of resprouting from fallen branches, or of clonal growth (pers. obs.), I had no certainty that each individual s£mipled was geneticEdly different from its neighbors. For the purpose of this study, however, I considered an individual any plant growing at least 1 m from its nearest conspecific. Except for three species, I sampled 10-»25 flowers per individual on different dates throughout the flowering season. The exceptions were or Guzmania multiflora with 10 flowers sampled per individual at the peak of its flowering season, and Centropogon solanifohus and Centropogon congestus with only 15 flowers per individual due to their low flower production.

PAGE 37

30 Table 1. Plant species and number of individuals sampled in each site and in each flowering season. Marked with an asterisk (*) Eire three instances in which repeated sampling of all individuEds was carried out on two or three different days during the flowering season. The table contains the number of plants sampled each day. San San TorreHato Antonio Pablo molinos Viejo Bromeliaceae Guzmania mtdtiflora 32 29 Campanulaceae CentroDoeon con?estus 35 9 Centropof on solanifolius 38 66 Gesneriaceae Besleria sp. 38 Besleria solanoides 1990 28 Besleria solanoides 1991 60 Cnliimnfia anisophvla 6 11 19 Melastomataceae Miconia actmiinifera 1990 38 Miconia acuminifera 1991 31 50 Leandra so. 12 Rubiaceae Psvchotria hazennii 33 37 15 Palicourea obesifolia 23 38 Palicourea lancifera* 29 Solanaceae Solanum sp. 1 Nov 90 29 19 Solanum sp. 1 Jan 91 26 Solanum sp. 1 Nov 91 34 17 Solanum sp. 2 17

PAGE 38

31 The sampling involved collecting the pistils of flowers that had been exposed long enough for pollination and pollen tube growth to occur. Due to differences in the floral biology of the species, the collection time was determined individiially for each species. Thus, flowers that lasted one day were collected the next morning, and flowers that lasted more than one day were collected several days after anthesis, once the stigma showed signs of senescence, i.e. browning or beginning to decompose. I preserved the styles in FAA (sdcohol, acetic acid and formaUn in a 9:1:1 ratio) and saved them for further processing. I counted the pollen tubes that reached the base of the style under an epifluorescence microscope, after staining the styles with an aniline-blue solution (Martin 1959). For species with less than 100 tubes per style, I coimted all tubes. For species with more than 100 tubes per style, I estimated the total number of tubes per style based on subsamples. The results reported here were obtsiined from ca. 10600 styles of 782 individuals of 13 plant species. The Plants The 13 species S£uiipled represent a partial cross-section of the life forms, pollinators and breeding systems from these cloud forests (Table 2). Dioecious species, canopy trees, and batand moth-pollinated species, however, could not be included in the sample because their densities were low and they did not occur often enough on the edges and transition zones for adequate sampling. Of the species studied, two epiph5rtes (a vine and a tank bromeliad), three subcsinopy trees, a treelet, two shrubs, £ind five herbaceous species comprise the sample of six plant famiHes. Among these plants, four were selfincompatible, five were self-compatible, two were distylous and one was facultative self-compatible (i.e. self-fertilization is only feasible when loads of

PAGE 39

32 CO 0) II o (xi: >^ ^ — ' U5 (N eo o t> o »-( ^ 1-1 ^-^ CX3 t> iH 2461 2632 2683 4087 623 CO (M 00 t> i-H eg CO a 8 8 >l C o 0) 3 (3) > 3 ° 'ii a.s d « -e CD a S3 I ••H 3 ^ 03 H OS B o QQ QD ® Jl O ^ QQ "S •c « ^ -0 a a on O »H u a> (M •5 5*3 ^ OQ Ch c I 2g 00 (1> o i •a CO u o (m cm d d o o oo a. 2 CO CO '© s i s So PQ CO CO u CO CO ft 6 CO O 0) d o u C o b4 t>i o o O o u d O) o o ^ -d 03 03 6oo o o d CO ID U •rH CO u d !» (V O a' 03 0) C0| CO c -c 0) Ol
PAGE 40

33 self-pollen are high but below a threshold in the pollen load; the self-pollen tubes fail to grow down the style). The sample includes hummingbird-, butterflyand bee-pollinated species. Among the hummingbird-pollinated species, Centrop ogon congestus £uid Q. solanifolius have long (>30 mm) curved tubular corollas and are visited almost exclusively by hermit hummingbirds, e.g. Phaethomis syrmatophorus and £. guv, and less frequently by Eutoxeres a quilar the sickle-billed hummingbird, and Schistes geoffroyi . a short billed himmiingbird. flnlnmnpfl anisophyla . a vine with straight tubular corollas (= 30 mm) is visited by Coeligena coeligena , a straight long-billed hummingbird. Gmultiflora .. Besleria solanoides . and Besleria acwere visited by short billed-hummingbirds: mainly Haplophaedia aureliae . and to a much lesser extent by Ocreatus imderwoodii and Adelomvia melanogenis . li. aureliae was also the main visitor to Palicourea obesifolia . and often established feeding territories that encompassed the crown of two or three conspecific trees. Psvchotria hazenni was pollinated by two species of clearwing butterflies (Qleria caucana . Standinger 1885, and Pteronvmia zeriina . Hewitson 1855) that began foraging on the one-day flowers at 10:00. Miconia acnminifftra i s a buzz-pollinated melastome that was visited mainly by large bees, while Leandra SD.and Snlannm 2^.2 were both pollinated by small halicteid bees. Finally, Solanum sSi. 1 and Palicourea lancifera were visited by both shortbilled hu mmin gbirds and large bees. Hvmimingbirds (mainly Q. imderwoodii ) began foraging on these two plant species earlier in the day than the bees. Therefore it is likely that the pollen deposited by the himamingbirds reached the ovules before that deposited by the bees, making the himuningbirds the main pollinators, if not the only ones. It is likely, however, that pollen deposited later in the day by bees also fertilized those ovules not reached by

PAGE 41

34 earlier hiommingbird pollination, thus I £im conservatively considering both birds and bees as pollinators. Statistical Analyses and Variables I compeired pollination levels between treatments (edge, transition zone and interior) using the median number of pollen tubes that reached the base of the style. I used medians, instead of the traditionally used mean, because zeroes in this system are biologically important since they indicate absence of pollination. Use of me£ms would obscure the zeroes if there was a single flower with some pollination. Thus, the median gives a more biologically meaningful estimate of pollination levels. I analyzed each species separately with a mixed model two-way analysis of variance, where fragment (random) and sampling zone (fixed) were the two factors. For those species present in only one fragment, I used a one-way ANOVA. I used data transformations to correct for heteroscedasticity whenever necessary. To calculate the intraindividual variation 1 used the coefficient of variation (stsmdard deviation/mean pollination level of each plant). Determining the minimimi pollination levels for finrit set is not strjiightforward (Stephenson and Bertin 1983). The few studies available have used dose-response experiments to estimate the minimiun nimiber of pollen grains per ovule required for seed set varies fi-om a 1:1 pollen grain/ovtde ratio to a 10:1 ratio (McDade and Davidar 1984, Murcia 1990, Silander and Primack 1978, Snow 1982). In any of these species, however, the ratio of pollen tube to ovule has not been determined. Furthermore, all of the above studies have been conducted in species with fewer than 10 ovules per flower and capable of developing fruits with as few as one seed. Plants that contain high numbers of ovxiles require a minimum nimiber of seeds to develop their finiits; otherwise, fi-uits are aborted (Bertin 1982, Stephenson 1981).

PAGE 42

35 In this study, species varied in the number of ovules from two to ca. 4000 (Table 2). Because the number of pollen tubes required for seed and fruit set is not known for these species, I assumed that a nimiber of pollen tubes equal to 50% the ntunber of ovules would be the minimum reqtiired for fruit set across all plant species. For the three species of the family Rubiaceae, which have only two ovules per flower, I used the proportion of flowers with two or more pollen tubes, although some species in this family may set fruit with just one pollen tube (Busby 1987, Feinsinger et al. 1988a, Feinsinger et al. 1988b). I called this measure of minimiun pollination for fruit set Per50. Statistical analyses for the intra-individual CV and Per50 followed the same protocol as the median. Due to the low nmnber of flowers produced by individuals of G. congestua and Q. solanifolius sometimes as low as one, the within-individu2d coefficient of variation and Per50 could not be calculated and thus are not included in these analyses. For each of the three dependent variables (median, CV, Per50) I carried out a Multivariate Analysis of Variance (MANOVA) to assess the overall effect of edge on pollination at a commimity-wide level using the 13 species sampled. For this test, I analyzed separately each edge, and only included those species with more than 5 individuals per sampling zone. I performed this analysis on the untransformed data. The robustness of different MANOVA tests to heterogeneity in the dispersion of the matrices varies according to the concentration of the structure (Barker and Barker 1984). I report for each MANOVA analysis the results of Pillai Trace tests because this is the least affected by the heterogeneity of dispersion matrices, yet still retains desirable power (Barker and Barker 1984). Wilk's Lambda tests which are perhaps the most commonly used, and the most sensitive to

PAGE 43

36 heterogeneity of dispersion matrices, )delded very similar results to Pillai Trace tests in all cases. Effect of Sample Size on Standard Deviations. To assess the adequacy of sample sizes used in the study, I constructed a plot of standard deviation as a fimction of sample size. I constructed the plot starting with five individuals selected at random from the poptdation sampled, and adding five randomly selected individuals each time. I considered an adequate sample size that which caused no fiirther reduction in standard deviations. I also explored the effect of increasing the nimiber of pistils on the within-individual standard deviation for one species. For each of five individuals of wSnlannm gj^.l, I constructed a plot of the standard deviation as a fimction of the number of styles sampled, starting with five randomly selected styles, and adding five randomly selected styles from that individual each time. Temporal Variation of Edye Effect I explored the temporal variation of edge effect (if any) between flowering seasons for three plant species: Snlannm gfi.! (three flowering seasons), and Mactmiinifera and fi. solanoides (two flowering seasons each; Table 1). Because not all individuals flower in consecutive seasons the samples of different years may contain different plants. I ansdyzed each flowering season separately with a two-way ANOVA. I also evaluated the consistency of edge effects during a flowering season by repeatedly sampling individuals on different dates. For Snlannm ay.l I sampled the same individuals on two dates (29 May and 6 Jime 1991) at the peak of their flowering season. For £. lancifera and Mamminifpr^ I sampled the same individuals (in one and two edges, respectively) on three

PAGE 44

37 dates, at the beginning, in the middle and three fourths through their flowering season. I calculated the correlation coefficients between pollination levels of individuals on consecutive samples of their flowering seasons (day 1/day 2, and day 2/day 3). Ex periments For the field experiments, I introduced two species into the forest in planned arrays and measured their pollination levels. I used one commercially available ornamental plant, Cerissa phoetida (Rubiaceae) native to South East Asia, and one native plant that grows at a lower elevation. Salvia aC(Lamiaceae). The experiment controlled for plant density and for any past experience by pollinators that could bias the animals toward certain areas in the forest as a result of prior associations. Thus, I selected plants whose flowers had no resemblance to any of the native species yet were visited by pollinators in the forest. In each of two forests, I placed three groups often plants each at 0-5 m (=edge), 25-30 m (=transition zone), and 85-90 m (=interior) from the edge. I kept the densities constant by covering an area of approximately 2 X 20 m with each set of ten plants, and assigned the plants to the treatments in a stratified-random design blocking for plant size and flower number. Both plant species are self-incompatible. Because they might have come from only a few individuals, I could not use the number of pollen tubes as a measure of pollination, as these might have reflected compatibility more than pollination. Instead I used the niraiber of pollen grains deposited on the stigma. Given the floral morphology of both species, pollen movement between anthers and stigma in the same flower require an £inimal vector. I collected stigmas of the experimental plants on three separate days. A preliminary analysis of vsiriance showed no differences in pollination levels

PAGE 45

38 among days, so I combined the data from the three days for each individual for the final anjdysis. PolUnation levels were low in general. Many flowers received zero pollen grains, and as a result the median was zero in most individuals. Consequently, I used the third quartile, rather than the median, in a two-way mixed-model ANOVA with fragment (random) and sampUng zone (fixed) as the two factors. Edge Description Physical conditions . During the long dry season of 1991 (Jime-July), I measured several vgiriables to describe the physical conditions along a transect fi:om the edge into the forest. In cloud forests, plants normally experience high relative moisture and low temperatures, and do not seem to show adaptations against effects of drought (Kapos and Tanner 1985, Medina et al. 1981). Thus, the long dry season might be the period of maximum stress (Wright 1991, Wright et al. 1992). Inside the forest, the dry season may not represent any physiological strain to the plants (Kapos and Tanner 1985). Near the edge, however, conditions are more likely to be extreme, and the long dry season would be the time when differences between the conditions of the edge and the interior are most emphasized. If the conditions at the edge fall outside the tolerance limits of the plants then, permanent changes in the vegetation associated with the edge are expected to occur. In San Antonio and San Pablo, I set up two transects to measure ambient temperature, air himiidity, and soil moisture. The transects were located 25 m apart and ran perpendicular to the edge fi-om 10 m outside the forest to 100 m into the forest. To minimize temporal effects on the measurements, i.e. to avoid confoimding distemce to the edge with time of the measurement, I took the measurements in the minimum time possible and randomized the starting station (inner or outermost).

PAGE 46

39 In each transect I measured air temperature and humidity with a sHng psychrometer at 10 m outside the forest and at 0, 10, 20, 25, 50, 75, and 100 m from the edge into the forest. Measurements were taken once a day between 10:30 and 14:00 on 5 days in San Antonio and six days in San Pablo. I measured the soil moisture with electrical resistance sensors (Watermark®). Sensors were buried 0.15 m deep, to sample soil moisture from the area of maximum understory root biomass (Becker and Castillo 1990). Sensors were located at 10 m outside the forest, and at 0, 5, 10, 15, 20, 40, 60 and 100 m into the forest. I set the gypsvun-type blocks in place at the beginning of the dry season (14 Jime) and monitored the desiccation process for 6 weeks as the dry season progressed. Vegetation analysis . I analyzed the spatial distribution of the 13 focal species in two fragments: San Antonio and San Pablo. In each fragment, I set up a 80 X 80 m plot with one side aligned with the forest edge. I divided each plot in 10 m bands that ran parallel to the edge starting at 0 m from the edge. On each 10 x 80 m band, I counted the number of adult individuals present for the 13 focal species. I used a Correspondence Analysis on the densities of the focal plants to determine the degree of similarity between consecutive samples along the edge to interior gradient, and any association between distance to the edge and plant species. For this procedure, I used the default algorithms to standardize the coordinates (SAS® 1988).

PAGE 47

40 RESULTS Pollination Pollination Levels Overall, there were no significant edge effects in pollination levels in any of the three fragments (Pillai trace = 0.771, p = 0.20; Pillai Trace = 0.880, p = 0.28; Pillai Trace = 1.49, p = 0.12, for San Antonio, San Pablo, and Torremolinos respectively). Of the 17 samples (13 species) only two cases showed a statistically significant response to proximity to the edge (Table 3, Fig. 3). The direction of the effect, however, differed between the two species: Palicourea lancifera had higher pollination levels on the edge than in the interior while Besleria solanoides , examined in 1990, had higher pollination levels in the interior than on the edge. Data from a second flowering season (1991) for E. solanoides . however, failed to show a difference between edge and interior (Table 3, Fig. 3). The purpose of this commimity-wide approach is to override the particularities of individual species. However, certain traits of the species may help interpretation of individual responses. £. lancifera is a case in point where the overall differences in pollination between edge and interior is perhaps a result of the distribution of flower morphs in the population. £. lancifera is one species exclusive to Torremolinos. Although equal nimibers of individuals in each morph (pin and thrum) were sampled at the edge and in the transition zone, nine of the 11 interior plants foimd were all the same morph (pin). Thus, it is possible that the low pollen tube coimt in the interior plants was not the result of lack of pollinator activity but rather a lack of pollen compatibiUty. Many pollinators do tend to fly very short distances while foraging, thus it is very Ukely that most pollen came from the nearest

PAGE 48

41 Table 3. Summary information of F-values from oneor two-way ANOVAs performed on the median pollen tube nimibers of plants located in three sampling zones: edge (E), transition zone (TZ) and interior (I), in two or three fragments. Data from species that occtured in only one fragment were analyzed for effect of sampUng zone using a one-way ANOVA, and thus show no fragment or interaction term in this table. Numbers in parentheses £ire the degrees of freedom. Data transformations are listed imder species names. Results of a-posteriori mtiltiple comparisons (Bonferroni/Dunn test) are provided in cases that showed a significant sampling zone effect. SPECIES FRAGMENT SAMPLING ZONE INTERACTION Guzmania multiflora (log x+1) 1.450 (2,4) 2.379 (2,70) 2.473a (4,70) Centropogon congestus 0.871* (1,1) 1.059 (2,39) .001 (1,39) Centropogon solanifolius (logx+2) Besleria sp. 7.95 (1,2) (2,38) 1.364 (2,98) .248 1.438 (2,98) Besleria solanoides 1990 5.178* (2,28) E=TZ; TZ=I; EI; E=I Columnea anisophyla 0.670 (hi x+2) (2,4) Miconia acuminifera 1990 1.630 (2,27) 0.630 (2,38) 1.772 (4,27) Miconia acuminifera 1991^ 9.679*** (Square root x) (5,10) 3.003a (2,182) E>TZ; E=I; TZ=I 2.556 (10,182)

PAGE 49

42 Table S-continued SAMPLING SPECIES FRAGMENT ZONE INTERACTION Leeindra sp. 0.146 Psychotria hazennii 21.603** (2 A) 1.5 0.488 Palicourea obesifolia 1.380 .827 .960 (9 Kf^\ Palicourea lancifera (lnx+0.1) 3.968* (2,26) EI; E>I Solanum sp. N 1990 (log x+1) 20.01* (1,2) 0.884 (2,42) 0.631 (2,42) Solanum sp. J 1991 .773 (2,22) Solanum sp. N 1991 6.820 (1,2) 1.834 (2,45) 0.451 (2,45) Solanum sp.2 (In x+1) 4.191* (2,14) E=TZ; TZ>I; E=I * p< 0.05; **p<0.01;*** p<0.001; a 0.06 < p < 0.05, b the model includes day as a nested factor imder fragment.

PAGE 50

Figure 3. Median number of pollen tubes (+ SE) of plants in three sampling zones. Sampling zones are: medium grey = edge; light grey = transition zone; dark grey = interior., in three fragments (SA = San Antonio; SP = San Pablo; TM = Torremolinos, or Hatoviejo in the case of Guzmania). All data are un transformed.

PAGE 51

44 160 800 -400 500 0 2000 Besleria sp. P. lancifera TM

PAGE 52

45 neighbors which were Ukely the same morph. To what extent this peculisir distribution of the flower morphs in the interior is determined by the distance to the edge, or by an mirelated event is lanknown at this time, since only this fragment contained enough individuals in all three zones for sampling. Solamim SD. 2, another species exclusive to TorremoUnos also exhibited very low pollination levels in the interior, perhaps due to a very low population density. The four individuals nf Snlannm sfp 2 were scattered through an area of 0.5 ha, while the edge and transition zone plants were all in hsdf that area. As with £. lancifera. the absence of these species from other fragments prevents any generalizations, and therefore the results from these two species should be interpreted with caution. Three other species showed significant differences in pollination levels. In all cases, however, the differences were caused by a significant departure in the transition zone, but no statistical differences were foimd between the edge £ind the interior. Within-Individual Coefficient of Variation (CV) PolUnation levels were as variable within individual plants at the forest edge as they were in the forest interior or transition zone (Table 4, Fig. 4). The only exception was .Solannm sjj.l, which in 1990 showed significantly higher intra-individual coefficients of variation in the interior than at the edge. The samples in the analysis of Solanum sp.l, however, were heteroscedastic and an appropriate transformation could not be found, thus, this result must be interpreted with caution. Commimity wide, plants showed no significant edge effects for any of the three edges (Pillai Trace = 1.05, p = 0.32; Pillai Trace = 0.82, p = 0.17; Pillai Trace = 0.12, p = 0.82; for San Antonio, San Pablo and Torremolinos, respectively).

PAGE 53

46 Table 4. As in Table 3, Svimmary information of F-values from one£ind twoway ANOVAs performed on the coefficients of variation of within-individual pollination levels SAMPLING SPECIES FRAGMENT ZONE INTERACTION Guzmania multiflora 1.707 (2,4) 1.853 (2,70) 1.478 (4,70) Besleria sp. 1.385 (2,38) Besleria solanoides 1990 0.581 Besleria solanoides 1991 0.884 (2,60) Colvminea anisophyla 1.445 0.353 U,iiD) 1.765 (4,26) Miconia acuminifera 1990 0.368 (2,39) Miconia acimoinifera 1991^ 7.65** (.0,1U) 1.897 (2,182) 1.514 (10,182) Leandra sp. 0.725 (2,9) Psychotria hazennii 1.119 (2,4) 0.853 (2,71) 1.252 (4,71) Palicourea obesifolia 0.028 (1,2) 1.294 (2,55) 1.078 (2,55) Palicourea lancifera 2.149 (2,26) Solanum sp. N 1990 (log x+1) 12.430 (1,2) 4.655*a (2,42) 1.031 (4,42) Solanum sp. J 1991 0.132 (2,19)

PAGE 54

47 Table 4c ontinued SAMPLING SPECIES FRAGMENT ZONE INTERACTION Solanum sp. N 1991 1.391 0.486 0.350 (1,2) (2,45) (2,45) Solanum sp.2 0.215 (2,14) * p< 0.05; **p<0.01; a variances were homoscedastdc and there were no appropriate transformations, thus I excluded the TZ treatment and compared only edge and interior b the model includes day as a nested factor imder fragment.

PAGE 55

Figure 4. Intra-individual coefficient of variation in the number of pollen tubes (+ SE) of plants in three sampling zones. Sampling zones are: black = edge; light grey = transition zone; dark grey = interior., in three fragments (SA = San Antonio; SP = San Pablo; TM = Torremolinos, or Hatoviejo in the case of Guzmania). All data are untransformed.

PAGE 56

49 o •i-H -4-) 5 0) QJ o I 2 00 ^^^2™*^* sp 1.000.00 1.00 0.50-1 0.00 0.75 B. solanoides 90 T p. landfera -1.00 0.00 SA SP TM

PAGE 57

50 Proportion of Flowers Receiving a Number of Tubes Equal or Larger than Half the Nu mber of Ovules (PerfiO) M. an^miniferf^, B. solanoides (1990, 1991), and Snlannm s^. (November 1990) exhibited a significant difference between the edge and the interior in the proportion of flowers that received enough pollination for half seed set (Per50; Fig. 5, Table 5). Two of these four cases, however, showed higher values in the interior, while the other two showed higher values on the edge. Furthermore, the effect or its direction were not consistent from season to season. Macuminifera . for example, showed a significant difference in 1991, but none the previous year. jg. solanoides . also showed a marked inconsistency. In 1990 it had significantly higher values on the interior, but in the following year vsdues on the edge were significantly higher (Fig. 5). Besleria solanoides (1990) was the only case that showed a significant edge effect on the median pollination levels (Fig. 3), and a significant edge effect on its potential for fruit set (Fig. 5). The three other cases, Mamminifpra solanoides (1991) and Snlamim g^. 1 (Nov 90), with significant differences in their potential for fruit set showed no differences in the pollination levels. For this variable, the edge did have an effect on the commimity as a whole in one of the three edges (Pillai Trace = 1.08, p = 0.28; Pillai Trace = 0.995, p = 0.02; Pillai Trace = 0.673, p = 0.37; for San Antonio, San Pablo, and Torremolinos respectively). The significant edge effect on the Per50 of plants in San Pablo reflects the strong edge effects on £. hazennii and Bsolanoides (1990). Both species showed a significant reduction in pollination levels near the edge compared to those in the interior in the San Pablo edge (F2,22 = 4.836, p = .018, and •2,22 = 01, for £. hazfimui and £. solanoides (1990) respectively). In the case of £. hazfimui, however, this significant difference was not consistent among the three edges (Table 5).

PAGE 58

51 Table 5. As in Table 3, summary information of F-values from oneand twoway ANOVAs performed on the proportion of flowers with pollination levels eqiial or higher than 50% the nimiber of ovules (Per50) SPECIES SAMPLING FRAGMENT ZONE INTERACTION Guzmania midtiflora Besleria sp. Besleria solanoides 1990 Besleria solanoides 1991 Colimmea smisophyla Miconia acuminifera 1990 Miconia acuminifera 199 1^ Leandra sp. Psychotria h£izennii (arcsine x) 1.475 1.054 1.882 (2,4) (2,71) (4,71) 1.203 (2,41) 17.989** (2,28) E=TZ; TZI 0.167 0.738 0.872 (2,4) (2,27) (4,27) 2.108 (2,39) 10.861*** 1.836 (5,10) (2,182) 0.682 (2,10) 0.746 2.009 2.804 (2,4) (2,78) (4,78) 1.176 (10,182) Palicourea obesifolia Palicourea lancifera 2.626 0.347 (1,2) (2,54) 6.357** (2,26) E=TZ; TZ>I; E=I 0.732 (2,54)

PAGE 59

52 Table 5-continued SAMPLING SPECIES FRAGMENT ZONE INTERACTION Solanum sp. N 1990 2.144 3.640* 2.811 (1,2) (2,42) (4,42) E=TZ; TZ=I; E
PAGE 60

Figure 5. Proportion of flowers in a plant with a number of pollen tubes > half the number of ovules (+ SE) of plants (Per50) in three sampling zones. Sampling zones are: medium grey = edge; light grey = transition zone; dark grey = interior., in three fragments (SA = San Antonio; SP = San Pablo; TM = Torremolinos, or Hatoviejo in the case of Guzmania). All data are untransformed.

PAGE 61

54 SA SP TM

PAGE 62

55 Effect of Sample size on Standard Deviationa In the four species analyzed (one included two flowering seasons), standard deviation among individuals remained constant with sample sizes Ifirger than 10 individuals (Fig 6). One case of five showed not a decrease, as expected, but an increase in standard deviations as sample size increased beyond 45 plsints. I found no changes in the within-individual standsird deviation for sample sizes larger than 5 flowers (Pig 7). Temporal Variation of Edge Effect Significant changes in pollination levels occurred in some plsmt species during consecutive flowering seasons, including changes in the intensity and direction of the edge effect (Figs. 3,4,5). To more finely dissect the sources of this temporal variation I considered the extent to which variation was caused by a) sampling different individuals in consecutive seasons, and b) the time of sampling during the flowering season. This allowed me to determine whether the edge effect was more pronounced during a certain time in the flowering season, and whether differences between flowering seasons could have been the result of sampUng during a certain time in one flowering season, and during another time in the next. Individual plants did not flower in all flowering seasons. In fact, as many as 80% of the individuals sampled in one season could not be sampled during the next mainly because they did not produce flowers. Thus, the fluctuation in pollination levels between seasons could be attributed to the contribution of different sets of individuals to the sample. To explore this possibility, I used data from Snlannm sQ.l, for which many of the same individuals flowered in two consecutive seasons: June 1991 and November 1991. 1 used the median poUination levels of those individuals to test

PAGE 63

56 Figure 6. Standard deviation of the number of pollen tubes as a function of the number of individual plants collected from one fragment, for four species and two seasons of one species: Open squares and triangles: B. solanoidea (1990, 1991); closed squares: Canisophvla : open circles: Macuminifera : closed circles: Solannm sp 1,

PAGE 64

57 Figure 7. Standard deviation of the number of pollen tubes as a function of the number of styles collected from five individued plants of Solannm gj^1 in Jime 1991.

PAGE 65

58 whether, in spite of overall inter-seasonal fluctuations in pollination in a population, the pollination received by one individual was predictable from season to season. I found no significeint correlation between the median nimiber of pollen tubes received by individual plants in the two flowering seasons (r = 0.390,n = 19, p=0.08). Thus, a plant that received high pollination levels in one season, relative to the population, did not necessarily receive high pollination levels on the next. Within a single flowering season, however, pollination levels of individuals were more predictable. In Solanum SE-l, Maniminifftra (1991) and £. lancifera , correlations between s£miples of the same individuals taken in consecutive weeks of the flowering season were significantly positive in five of seven cases (Table 6). Thus, individuals that one day received high pollination levels relative to the rest of the population were very likely to receive high pollination levels a week later. Experiments Analyses of variance showed no significant differences in the number of pollen grains (third quartile) received by plants that were located on the edge and those located in the interior for either Q. phoetida or Salvia ss.. (^2,48 = 0.178, p>0.8 and F2,51 = 0.771, p>0.4, respectively. Fig. 8). No significant interaction or effect of fragment was observed. In both cases, it was necessary to carry out a (Log +1) trsinsformation to correct for a Ifirge heteroscedasticity. Edge Deacriptinn Phvsical conditions. Above-groimd physical conditions measured during the long dry season (June-August) of 1991 showed no edge effects. Both air temperatm-e and air moisture were uniform in the first himdred meters of the

PAGE 66

59 Table 6. Correlation coefficients between pollination levels of individuals on different days of their flowering season. Days 1, 2 and 3 correspond to dates in consecutive order, one week apart (see methods). Values in parentheses are the number of plants sampled. For Snlannm gj^.H collected only on two days. Day 1/Day 2 Day 2/Day 3 Snlannm fij^. 1 (Jime 1991) 0.490* (22) MiCimia ammimfprfl (1991) SP SA 0.154 0.532 * (22) (20) 0.317 * (22) 0.493 * (22) Palicourea lancifera 0.565 ** (22) 0.213 (25) * p<0.05; ** p<0.01

PAGE 67

60 I a o Edge TZ Interior Edge TZ Interior San Antonio San Pablo Fig 8. Pollination levels of plants of Cerissa phoetida and Salvia gp. Circles indicate the mean of the average pollination levels of 10 plants located in each sampling zone: edge, transition zone (TZ) and Interior in each of two fragments, San Antonio £ind San Pablo. Lines connect values corresponding to the three sampling zones in each fragment. Vertical bars indicate ± one standard deviation. The figures show imtransformed data. 1

PAGE 68

61 forest in both edges (Fig. 9). Ten meters outside the forest, conditions were only slightly different: air temperature was higher, but air moisture was similar to that inside the forest. Below ground, physical conditions were different. At the beginning of the dry season, soils were completely saturated at all points along the transects. As the season progressed, soil desiccated faster near the edge than in the interior, and in spite of a few scattered rains during the dry season, the desiccation process was cumulative. On the last sampling date, six weeks into the dry season and 37 days since the last hard rain, soil moisture was much lower over the first five and ten meters of the forest of San Pablo and San Antonio, respectively than elsewhere in the interior (Fig. 10). Vegetation analysis . The correspondence analysis for San Antonio isolated the samples of the first 10 and 20 meters from all others (Fig, 11). Axis 2 showed a significant correlation with the soil moisture measured on July 19, 1991 (r = 0.859, n = 5, p<0.05), yet no correlation was apparent with soil moistures recorded earlier in the dry season. Two species were associated with the edge samples: £. obesifolia and Macnminifpra ^ the two subcanopy trees. A third species, Q. solanifolius showed an intermediate position with respect to the edge and interior samples (Fig 12). For San Pablo, on the other hand, the correspondence analysis did not discriminate among samples in relation to their distance to the edge. One species, Maf^iiminifpr^^ , w£is spatially related with the 10 and 20 m samples. These two, however, were close to the most interior samples. Vegetation samples were not related to soil moisture.

PAGE 69

62 Distance to the edge (m) Figure 9. a) Air moisttire and b) air temperature measured on 26 July 1991 at different distances from the edge. Data represent the mean values of two transects (± SD) in each of two fragments: Open squares = San Antonio and closed circles = San Pablo.

PAGE 70

63 -20 0 20 40 60 80 100 120 0 120 Distance to the edge (m) Figure 10. Soil water tension recorded on 26 July 1991, six weeks into the dry season and 38 days since the last rain, for San Antonio and San Pablo. Measurements were taken at 15 cm deep, along two transects, indicated by the two symbol types, located perpendicular to the edge from 10 m outside the forest to 100 m into the forest in two fragments.

PAGE 71

64 (N s 1.510.50-0.5San Antonio 80 0 O 30 50( o 60 ' a 20O 1.5 3 10.50-0.5 0 0.5 1.5 2.5 30 0 40 i\ San Pablo 70 -e S O 0 80 IOq 20 50 -0.5 0.6 1.5 Axis 1 Figure 11. Correspondence analysis on the species composition of samples taken at different distances from the edge in two fragments (SA, SP). Nimibers close to the points represent the distance (m) from the edge of the inner most limit of the 10 x 80m bands (see methods). In San Antonio, axis 1 and axis 2 accoimt for 72.7% of the total variation ( 54.0% and 18.7%, respectively). In San Pablo, the two axes accoimt for 71.3% of the variation (49.8% and 21.4%) respectively.

PAGE 72

66 1.510.50 -0.5 H -1Bsp Cc — I— -0.5 Ph San Antonio Gm Ma Cs Po T 1 — I— 1.6 0 0.5 2.5 1.50-0.5• Ca San Pablo Gm Cc • • 9Bs Sg« Po# • Ma -0.5 0 0.5 1.5 Axis 1 Figure 12. Correspondence analysis on the species composition of S£imples taken at different distances from the edge in two fragments (SA, SP). Letters close to the points correspond to the species name: Gm: Guzmania multiflora : Cc: CentroPOgon COngestus; Cs: Centropogon solanifolius : Ca: Coltmmea amSQPhvla; Bs: Besleria solanoides: Bsp: Besleria sfi.; Ma: Miconia

PAGE 73

66 acumnifera; Po: Palicourea obegifoUa; Ph: Psychotria hazsmui; Sg: Solanum aiii. Discussion Proximity to the edge does not consistently affect the quantity or consistency of pollination of plants in the cloud forests near Cali, Colombia. I found distance to the edge to be associated with reductions or increases in pollination levels of plants in only a few species, and a lack of consistency of these differences between flowering seasons. These results suggest that the edge effect is not always as pervasive and consistent as the literature has described (Kapos 1989, Lovejoy et al. 1986, Malcolm 1991). Previous studies have fotind that conditions at the edge are drasticsdly different from those in the interior, and that the distance into the forest that these conditions are gdtered varies an3rwhere between 10 to 50 m. There are several possibilities why the results of this study are different. Studv Design This study was designed to test the edge effect with a factorial design, with the factor "distance to the edge" divided in three categories, and a second factor that allowed for replication on several edges. This design is very powerful but its relevance is contingent upon assigning the categories correctly. It could be argued that the distances that defined the three sampling zones were not scaled appropriately for edge effects that actuEdly occurred. Therefore, I examined a-posteriori the pollination levels of plants as a fimction of their distance to the edge. For each species and each edge, I plotted the median number of tubes of each individual as a fimction of its distance to the edge. I visually inspected these plots searching for break points in the distribution of values that would indicate discrete subgroups of values at distances fi"om the edge different from the ones used here. In none

PAGE 74

67 of the cases, did I find any reason to consider the distances I originally selected as inappropriate. As with £iny other statistical test, the power of these analyses is dependent on the sample size and the variation among ssimples. Pollination levels are highly variable (Feinsinger et al. 1986, Herrera 1988). In this study, substantial variation in pollination levels was observed among individuals, as well as among flowers in an individual. In 34 of 87 cases, standard deviations aroimd mean pollination levels were as large as the means in each sampling zone (Fig. 3). Such variation in pollination levels among individuals was paralleled by the variation in pollination levels within individuals. Intra-individual coefficients of variation were in genered higher than 0.7. It is likely that large standard deviations resulted from small s£imple sizes either in the nimiber of plants sampled or in the nimiber of flowers sampled per plant. Analyses indicated that standard deviations of the plants sampled are apparently representative of the standard deviations of the population, and are not an artifact of smgdl sample sizes. Finally, I used the median as a variable to estimate pollination levels of individuals, instead of the me£in, which is more commonly used by other pollination studies. Thus, it could be argued that medians £ire not as sensitive as means in detecting differences in pollination levels that would be biologically meaningful. To explore this possibility, I performed correlation analyses between the mean and the median for each species, considering each flowering season separately, but combining the results of all edges in each season. The correlation coefficients for all 17 samples ranged from 0.796 to 0.965, suggesting that the results were not an artifact of the variable selected to measure pollination levels.

PAGE 75

68 The PlantPollinator Interaction System Because plant-pollinator interactions are generally not species-specific (Feinsinger 1978, Feinsinger 1983, Feinsinger 1987, Roubik 1989 p 320, Waser 1978, Waser 1983), the plant-pollinator interaction may be an inherently robust system. The plant species in this study did not engage in any apparent species-specific interactions with any particular animal species, nor did they show extreme morphological adaptations that would permit exclusive flower visitation by one species. Rather, plants were visited by at least two species in the same taxonomic order, and for some plant species the list of potential pollinators spanned several phyla. Even Q. congestus . and Q. solanifolius . with their highly modified corollas, were visited by at least four different himmiingbird species. This lack of species specificity, although potentially or theoretically costly in terms of interference competition for pollination among plant species (Feinsinger 1987, Waser 1983), may have a positive effect in the face of perturbations or in environments where the nectar supply is highly variable. If a pollinator avoids the edge due to its physical conditions, the plants at the edge could still receive visits from other less sensitive species that could make up for any reduction in visitation from former pollinators. Or, some pollinators from the surrounding matrix could enter the forest and visit plsuits close to the edge. The result would be different pollinator assemblages servicing subsets of a plant's population. Although different pollinator assemblages coiild produce similar pollination levels on the edge and the interior, behavioral differences between forest and edge polHnators may cause changes in the neighborhood size for edge plants (Levin and Kerster 1974, Schmitt 1980), and in the genetic composition of the pollen deUvered

PAGE 76

69 (Schmitt 1980). Those effects would be more subtle than simple changes in pollen tube numbers, and were beyond the scope of this study. In this study, however, plant species not only had seversd potential pollinators, but most of the pollinators were generalists in their diet. Thus it is imlikely that the above sceneirio would apply. Among pollinators, food generalists tend to be tolerant to environmentsd variation within the forest including the conditions caused by natural perturbations, as well as the conditions encoimtered in the canopy (Feinsinger et al. 1987, Feinsinger et al. 1988b, Murcia 1987, Roubik 1989 p. 324, Stiles 1975). Therefore, it is more likely that the same pollinators are visiting both the plants on the edge and in the interior, and there is no reason to expect sub-population structuring. A previous study showed that flower visitation by forest dwelUng hummingbirds is not affected by small natural forest disturbances (mainly treefall gaps), or medium-sized clearcuts (Feinsinger et al. 1987). Edges, in the case of this study, resulted from large scale perturbations, and as such may create conditions that are too inhospitable for forest-dwelling hummingbirds . The lack of variation in the above-ground physical condition, tied in with casual observations and previous experience with these hummingbirds (Murcia 1987) indicates that this is not the case. Six plant species in this study are pollinated by either Haolophaedia aureUae or Ocreatua imderwoodii, two short billed hummingbirds. Neither of these hummingbirds ventures outside of the forest into neighboring gardens (Murcia 1987), but both species have been observed visiting the flowers on the exposed face of the edge (pers. obs.; G. Kattan, pers. comm.). Amazilia saucerottei , a hummingbird from open areas, occasionally entered the forest in San Antonio (Kattan, impubl. data ), but was only seen associated with a stream. Never

PAGE 77

70 did I or other researchers in this area, or in Monteverde (Costa Rica) see Asaucerottei foraging even 1 m inside the forest (G. Kattan, P. Feinsinger, pers. comm). One set of plants whose pollination levels may be more susceptible to their proximity to edge are those with long, tubular corollas. Phaethomis svrmatophorus and £. guy, two hermit hummingbirds that pollinate Q. COPgestmg and Q. solanifolius are restricted to the understory (pers. observation). Furthermore, no other long-billed hummingbird occurs in the area outside the forest, and due to the complex architecture of the flowers, it is imlikely that other non-hermit himmiingbird visited the edge plants (with the exception of Sduaks geofifrovi which is also a forest dweller). Given that for both plant species the pollination levels were similar on the edge and the interior, one can infer that the hermit hummingbirds visited both sets of plants indiscriminately. Very little is known on the natural history of highland tropical hymenopterans, and thus it is premature to predict how conditions on the edge would affect their foraging behavior (Roubik 1989), and ultimately the pollination of plants. Many hymenoptera, however, have been reported as rather vagile and capable of moving across open spaces while foraging (Raw 1989). Also, bees show variation in their spatial distribution in the forest depending on the season. Studies have reported tropical forest bees concentrating on the canopy during the dry season, and moving to the understory during the wet season (Roubik 1989). These results suggest a wide tolerance for biotic and abiotic conditions. £. hazfinnii, the only butterfly-pollinated plant in this study was visited exclusively by clearwing butterflies (Ithomiinae), which are foimd more ofl«n in moist and dark parts of the understory (DeVries 1987). In spite of the

PAGE 78

71 reported habitat specificity of its pollinators, £. hazeimii did not show an edge effect on any of the polUnation variables measured. One possible explanation for this lack of effect is the scarcity of £. hazennii on the exposed face of the edge. All plants occurred at least 1 m into the forest, where they were often shadowed by the overhanging canopies of the trees at the edge. Edge Effect on Physi cal Conditions Although many of the pollinators in this study have been reported as behaviorsdly flexible, these pollinators could shy away from the edge under harsh physical conditions like those reported for other edges (Kapos 1989, Wilhams-Linera 1990b). I foimd, however, that in these fragments, the conditions were not drastically different between the immediate vicinity of the edge and the interior. Air temperature and moisture were quite uniform over the first 100 m into the forest, and similar to the conditions 10 m outside the forest. Other studies have foimd differences in air temperature sis high as 3-4.5°C between the edge and the interior, but also the transition from these conditions to the interior took place over 15-20 m (Kapos 1989, WilhamsLinera 1990b) respectively. To my knowledge, this is the first study pubHshed on edge effects in cloud forest edges, and although it is presently imknown how typical this pattern is for cloud forests, it is very unlikely that environmental conditions interfere with pollinator movement between edge and interior in this study site. Contrary to the apparent absence of edge effects on the physical conditions above-groimd, below-groimd physical conditions did vary with proximity to the edge. At the edge, soils were drier than anywhere along the transects, and the low soil moistiu-e extended into the first 5 or 10 m at SP and SA, respectively. This suggests that environmental conditions on the edge are more likely to affect the pollinator-plant interaction through their

PAGE 79

72 effect on the survival of the plants than through their effect on the pollinators' movements. Differential plant survival due to different tolereinces to the edge conditions would, in turn, influence the composition of vegetation near the edge, and ultimately pollination levels. In this study, the two edges showed no consistent edge effect in the vegetation composition. One the edge at SA showed a difference in the distribution of plants as a function of the distance to the edge, most likely as a result of changes in soil moisture. In spite of the vegetation being different in the first 20 m, no consistent edge effects in polUnation were found in that fi-agment. The two-fold reduction in soil moisture found at the edge of SP does not seem to be associated with high temperatures or low air hmnidity outside the forest, as would be expected from the observed physical effect of recently created edges. In this case, it is Ukely the result of high evapotranspiration rates associated with high fohar densities, that can result from an increase in hght availabiUty at the edge. This drastic reduction in soil moisture, however, seems to have very little or no impact on the vegetation. Costing and Kramer (1946) also reported a similar reduction in soil moisture at the edge (0 m into forest) although the reduction was not as abrupt as in this study. Time Since Edge Creation The short distance over which physical conditions associated with the edge disappeared in these fragments is very hkely a result of the age of the edges, perhaps more than their location in a cloud forest. Most previous studies of edge effect in the tropics have been carried out along yoimg edges, i.e. less than 5 years old (Kapos 1989, Malcohn 1991, Williams-Linera 1990a). These studies describe drastic changes in the area adjacent to the edge in the physical conditions and in the composition of the fauna and flora subsequent to the creation of the edge. These drastic changes seem to be exacerbated

PAGE 80

73 when fire occurs after the creation of the edge (Hester and Hobbs 1992, Lovejoy et al. 1986), The drastic changes observed on newly created edges, however, might have led to overestimates of the projections in time of the persistence of these edge effects. To what extent these changes persist in the forest sifter a few years and continue modifying it, is a question that only recently has begim to be addressed. Two studies have addressed the changes of edge effect over time. In Panama, a study of five edges ranging in age fi-om 10 months to 12 years foimd that the edge effect on canopy cover and basal area decreased in intensity and penetration into the forest in older edges (Williams-Linera 1990). Similar results were found by another study that examined the edge effect in edges 2-15 years old (Blanchard 1992). In the Ocala National Forest (Florida,USA), edge effects on light intensity and soil temperature were less intense in the 15 year old than in the two year old edges. In addition, a study on 20 year old riparian forest edges, found that the effect of edge on light levels disappears within 12 m into the forest, and any differences in vegetation structure caused by the edge conditions were restricted to this narrow band (MacDougall and Kellman in press). MacDougall and Kellman (in press) suggest that the rapid disappearance of the edge effect into the forest is a consequence of the age of the edge. Their suggestion seems to be supported by the results presented here, and by (Blanchard 1992) and (WilKams-Linera 1990). OveraU, these four studies indicate that the development of vegetation that seals the forest edge, and the closure of the canopy above and beyond the edge into the open space, play a major role in stabilizing the physical conditions near the edge. This closure of the edge is a consequence of fast growth rate and seedling recruitment spurred by higher

PAGE 81

74 light availability (Gysel 1951, Ranney et al. 1981, Sork 1983, Williams-Linera 1990a). Conclusions The results of this study imply that even though the concept of edge effect is heavily inscribed in the minds of conservationists, we know very little about the biological consequences of edges. Existing studies are very diverse in the type of forest studied, time since edge creation, variable measured, and latitudinal position. All these parameters are likely to influence the results, and thus, at this point general patterns remain unclear. The results of this study indicate little or no edge effect on plant-polhnator interactions in the edges studied, perhaps as a result of Uttle change in the environmental conditions associated with the edge. It is possible that these results are a consequence of the age of the edges, and that soon after the edges were created the effect was more drastic. It is imknown at this point the extent of the forest resilience to changes in the forest dynamics caused by edges.

PAGE 82

CHAPTER FOUR Toward a Unified Theory of Edges Contrast as a Determinant of Edge Effects The current Uterature on edge effects 3delds no clear patterns. Studies report detrimental, beneficial or no edge effects (chapter 2). Memy variables seem to be involved in this apparent discrepancy among results: edge types, edge age, edge history, edge compass orientation, variables measured, and types of vegetation separated by the edge. There appear to be as many or more kinds of edges as there are researchers. Without a more unified vision of edges, however, the conservation value of studies in edge effects is limited to the specific circumstances in which the data were collected. In other words, what we know about edge effects from one site may not be applicable for that same site if the matrix surrounding the fragments changes; or may not be applicable to a nearby site if the vegetation types separated by the edge are different from those at the first site. Could we extrapolate fi-om other studies to predict what might happen in those fragments? I think that is unlikely. In trying to reconcile the empirical information currently available, I propose a model as an initial step towards finding the imderlying factors causing variation in responses to edges. This model is based on contrast. Contrast was defined as the difference between two habitats, and the abruptness of the transition between the two (Kuchler 1973). For example, a boundary between a pasture and a mature forest has a higher contrast than one between a pasture and young secondary growth. Also, boimdaries 75

PAGE 83

76 between two vegetation t)T)es may differ in contrast if the transitions diflfer in abruptness. This concept of contrast was used by wildlife managers to assess the need for management find creation of edges for game (Thomas et al. 1979). More recently other authors have considered contrast in the context of edge effects (Angelstam 1986, Ratti and Reese 1988, Wienset al. 1985, Yahner et al. 1989), but have not defined the term in their studies, or attempted to explore its general applicability as a predictor of edge effects. Here, I define contrast more precisely and explore its general applicabiUty as a predictor of edge effects in forest-pasture boimdaries. I define contrast as the difference in the values of the variables that modulate an edge effect, between two points in the immediate vicinity of each side of the edge (points A £ind C, in figure 13). I define contrast only in terms of those variables that modulate each individual factor that has the potential to cause an edge effect, and not for the habitat or ecosystem as a whole. I restrict the measurement of contrast to a few variables at a time because the permeability of an edge to an edge effect depends only on some variables at a time. Those variables can be abiotic or biotic. Abiotic variables can be related to the vegetation structure, e.g., height, lesif and stem density, and aspect of the face (dripline, advancing or cantilevered (Ranney et al. 1981). These components of vegetation structure act as modulators of abiotic factors such as light, wind, and temperature, and so ultimately determine the strength of physical edge effects. Leaves, for example, intercept incident light that reaches the face of the forest edge. Thus with increasing leaf density at the edge less Hght penetrates the forest. Leaves also reduce air temperature through evapotranspiration. Thus in edges with high leaf density the difference in air temperature between the edge and the interior occurs over a

PAGE 84

77 Figure 13. Diagram illustrating a forest-pasture edge. A and C are points from the pasture and the forest (respectively) in the immediate vicinity of the edge. B is the point of edge creation and maintenance. I is a point in the interior of the forest, imaffected by the edge, d, is the distance the edge effect penetrates.

PAGE 85

78 shorter distance. Biotic variables, on the other hand, may involve species interactions, e.g., predation, brood parasitism, competition, and mutualisms. Biotic variables are perhaps harder to quantify than abiotic because they must be assessed from the point of view of the species in question (or the biotic factor with the potential to cause an edge effect) . Using the traditional definition, previous authors had predicted that the intensity of edge effects increases with the contrast between adjacent habitats (Angelstam 1986, Thomas et al. 1979). In the model I present, I modify this prediction at very high contrasts, where the intensity of edge effects decreases (Fig 14). Although edges between different vegetation types yield different contrasts, contrast is not a static parameter. As vegetation structure on the edge changes over time, for example, the contrast between the two zones changes. This, in turn, results in a change in the intensity of edge effects. Therefore edge effects are dynamic, and dependent on the contrast, which is also dynamic. For example, in a recently created forestpasture edge there is a large difference in vegetation height, and in leaf and stem density. At this initial stage, the contrast in leaf and stem density between the pasture and a zone of forest closest to the edge is intermediate (contrast between points A and C, in Fig. 13). Distance d, in figure 13, is the distance that the edge effects penetrate into the forest, and is a measure of the intensity of edge effects. Point E represents the forest interior and its contrast with point A is maximum. To illustrate these effects, consider what happens to a physical condition such as hght. Light penetrates a distance d, but its intensity and quality changes as it approaches point C as it is intercepted by leaves and stems. Light penetration is likely to cause changes in the conditions near the edge. What happens to the edge after its creation will determine whether the contrast between points A and C increases or

PAGE 86

79 Figure 14. General model of intensity of the edge eflFect (= distance into the forest) on forest next to anthropogenic clearings, as a function of the contrast between the variables that modulate that edge effect.

PAGE 87

80 decreases. As a resvilt of high Hght availability, vegetation growth is spurred and a leaf flush occurs. As \ea£ and stem density increase, light penetration decreases. Thus, the segment d decreases, eventually approaching zero, i.e., the intensity of the edge effect is decreasing. This , in turn, results in an increase in the contrast between points A and C. That is, conditions at C are returning to the conditions at E, the forest interior. In this case, edge effects brought up by an intermediate contrast resulted in a shift in the contrast towards the high end of the contrast axis in figure 14, and a further reduction in edge effects. Some forests however, may not move sufficiently to the right on the contrast axis, and they will experience continuously high physicsJ edge effects. Chen et al. (1992), for example, have reported that northwestern rainforest in Washington State (USA) do not develop a thick understory near the edge. Consequently, microclimatic patterns are changed up to 240 m into the forest. By comparison, several other studies on old edges are only changed in the first 20 m (Blanchard 1992, Kapos 1989, MacDougall and Kellman in press. Costing and Kramer 1946). Wind is another factor that may cause chsmges in the edge that result in changes in contrast between points A and C in Fig, 13. In a new forestpasture edge, where there is a difference in contrast in vegetation height between points A and C. Winds may create turbulence as they encounter the forest barrier. This turbulence may topple down those trees closest to the edge (Lovejoy et al. 1986). Consequently, the effect of wind on tree mortality is high £ind it can penetrate a distance d. With time, dead trees may be replaced by shorter individuals that offer less contrast in height between points A and C, forming a less abrupt transition between pasture and forest.

PAGE 88

81 resulting in reduced wind edge e£fects. In this case, the edge effects may cause a shift in the contrast axis towards low contrast. Because edge effects cause changes in the vicinity of the edge, those changes necessarily result in shifts in the contrast. The strongest the effects, the faster the shift would be either towards low or high contrast. As edges approach high or low contrast, the intensity of effects decrease and the shift along the contrast axis decreases. These trends in shifts along the contrast axis suggest that intermediate levels of contrast are unstable while the extremes are most stable. Biological edge effects also seem to follow the model proposed. At intermediate levels of contrast, biological effects shoiild also be highest. If there is little contrast between the two habitats, species should have similar adaptations to the physical and biological environment. In a fragment of mesic forest interacting with tall second growth forest, for example, physical effects may be mild. Species from each habitat may freely cross the edge, and their impact on the other habitat will likely be minimal. If, on the other hand, the two ecosystems are drastically different, then species may avoid crossing the edge. Avoidance may result from the edge acting as a physical barrier or from the imsuitable biological environment of the other habitat. At intermediate levels of contrast, species may cross the boimdary £md have an impact by disrupting processes. Brood parasitism is a case in point. Cowbirds are typical of open areas with some vegetation cover. In the temperate zone, cowbirds (Molothrus ater ) enter the forest fragments in the temperate zone and parasitize nesting birds (Brittingham and Temple 1983, Gates and Gysel 1978). The fairly open xmderstory may provide an intermediate contrast between open areas and temperate forests, and their effect on the forest birds is large. In the tropics, on the other hand, although

PAGE 89

82 cowbirds (Mbonariensis) parasitize nesting Henicorhina leucophrvs located in edges of secondary growth, cowbirds are never seen in nearby mature forests or forest fragments. Active search of Hleucophrys nests have yielded no cowbird eggs inside forests (Kattan and Murcia, impubl. data). Because cowbirds locate their hosts' nests visually (G. Kattan impubl. data), the high contrast produced by thick vegetation may prevent cowbirds from finding nests in the forest edge (Gates and Gysel 1978). The high leaf density in high contrast edges is also Ukely to reduce the searching efficiency of visually oriented predators as well. In edges with low contrast, and high visibility, the impact of both cowbirds and predators should be less because those edges tend to sustain fewer nesting birds (Chasko £md Gates 1982). How changes in biological interactions caused by the edge feed back into the system to cause further changes in the contrast, is imknown at this point. Few studies have explored the patterns of edge effects on some biological processes, but none have explored the mechanisms behind these patterns. The model I propose is a first attempt to imify concepts. This model is general and simple. Its heuristic value lies in recognizing that although edges and edge effects are not simple, identifying underlying principles can help us distill our knowledge of edges into a general theory. Because studies published thus far vEiry substantially in methodology and design, the validity of this model can not be fully tested at this point using the available literature. Future directions Perhaps the biggest difficulty in interpreting the results of pubHshed papers Ues on the scant description of the study sites, and of the criteria used for determining where the edge is located. Ranney et al (198 1) identified two important concepts with respect to the position of the edge: the point of edge I !

PAGE 90

83 creation and the point of edge maintenance. These two factors determine the aspect of the edge, which can act as a modulator of edge effects. In Gates and Gysel (1978), for example, the three edges sampled had different aspects which reflected on the data. In addition, the edge was measured farther out than in many other studies, thus, their conclusions may differ from others' (Ratti and Reese 1988). The description of the area is also important in assessing whether the portrayed edge effects are independent from other factors associated with landscape heterogeneity that could be causing changes as well, for example, ravines, steep slopes, waterlogged soils. Additional difficulty in interpreting the results stemmed from the lack of repHcation or the prevalence of pseudoreplication. Few studies incorporate replication into their designs, and thus, conclusions sometimes can be very limited in scope. Carefully designed studies on a diversity of edges and biological systems should allow further exploration of the mechanisms involved in edge effects, in peuticular, those mechanisms that attenuate the detrimental effects.

PAGE 91

LITERATURE CITED Angelstam, P. 1986. Predation on ground nesting birds' nests in relation to predator densities and habitat edge. Oikos 47: 365-373 Barker, H. R. and B. M. Barker. 1984. Mtiltivariate Analysis of Vgiriance (MANOVA). The University of Alabama Press. Alabsima, USA. Bawa, K. S. and J. H. Beach. 1981. Evolution of sexual systems in flowering plants. Annals of the Missouri Botanical Gardens 68: 254-274 Becker, P. and A. Castillo. 1990. Root £irchitecture of shrubs and sapUngs in the understory of a tropical moist forest in lowland Panama. Biotropica 22: 242-249 Bertin, R. I. 1982. Floral biology, hunmiingbird pollination and fixiit production of tnmipet creeper (Campsis radicans : Bignoniaceae). American Journal of Botany 69: 122-134 Blake, J. G. and W. G. Hoppes. 1986. Resource abundance and microhabitat use by birds in an isolated east-central Illinois woodlot. Auk 103: 328340 Blanchsird, J. D. 1992. Light, vegetation structure, and firuit production on edges of clearcut sand pine scrub in Ocala National Forest, Florida. University of Florida. M. Sc. thesis. Blouin, M. S. and E. F. Connor. 1985. Is there a best shape for nature reserves? Biological Conservation 32: 277-288 Bradshaw, F. J. 1992. Quantifying edge effect and patch size for multiple-use silviculturea discussion paper. Forest Ecology and management 48: 249-264 Brittingham, M. C. and S. A. Temple. 1983. Have cowbirds caused forest songbirds to decline? Bioscience 33: 31-35 Burger, G. V. 1973. Practical Wildlife Management. Winchester Press. New York, New York, USA. Busby, W. H. 1987. Flowering phenology and density dependent pollination success in Cephaelis elata (Rubiaceae). University of Florida. Ph.D. Dissertation. 84

PAGE 92

85 Chasko, G. G. and J. E, Gates. 1982. Avian habitat suitability along a transmission-line corridor in an oak-hickory forest region. Wildlife Monographs 82: 1-41 Chazdon, R. L. 1986. Light variation and carbon gain in rain forest understorey palms. Journal of Ecology 74: 995-1012 Chen, J., J. F. Franklin and T. A. Spies. 1992. Vegetation responses to edge environments in old-growth Douglas-fir forests. Ecological Applications 2: 387-396 Correll, D. L. 1991. Human impact on the functioning of Isindscape boimdaries. pages 90-109. in M. M. Holland; P. G. Risser and R. J. Naiman (editor). Ecotones: The role of landscape boundaries in the management and restoration of changing environments. Chapman & Hall. London, England. Dagg, A. I. 1976. Wildlife Management in Europe. Otter Press. Waterloo, Ontario, Canada. Dasmann, R. F. 1964. Wildlife Biology. J. Wiley and Sons, Inc. New York, New York, USA. DeVries, P. J. 1987. The butterflies of Costa Rica and their natural history. Princeton University Press. Princeton, New Jersey, USA. Diamond, J. M. 1975. The island dilemma: Lessons of modem geographical studies for the design of habitat reserves. Biological Conservation 7: 129146 Etherington, J. R. 1982. Environment and plant ecology. John Wiley and Sons. Chichester, United Kingdom. Feinsinger, P. 1978. Interactions between plants and himmiingbirds in a successional tropical community. Ecological Monographs 48: 269-287 Feinsinger, P. 1983. Coevolution and pollination, pages 282-311, in D. Futuyma and M. Slatkin (editor). Coevolution. Sinauer Associated Inc. Boston, Massachusetts, USA. Feinsinger, P. 1987. Effects of plant species on each other's pollination: Is community structure influenced. Trends in Ecology and Evolution 2123-126 Feinsinger, P., J. H. Beach, Y. B. Linhart, W. H. Busby and K. G. Murray. 1987. Disturbance, pollinator predictability, and pollination success among Costa Rican cloud forest plants. Ecology 68: 1294-1305 Feinsinger, P., W. H. Busby, K G. Murray, J. H. Beach, W. Z. Pounds and Y. B. Linhart. 1988a. Mixed support for spatial heterogeneity in species

PAGE 93

86 interactions: Hummingbirds in a tropical disturbance mosaic. Americsin Naturalist 131: 33-57 Feinsinger, P., W. H. Busby and H. M. Tiebout III. 1988b. EflFects of indiscriminate foraging by tropical himmiingbirds on pollination and plant reproductive success: Experiments with two tropical treelets. Oecologia (Berlin) 76: 471-474 Feinsinger, P., K. G. Murray, S. Kinsman and W. H. Busby. 1986. Floral neighborhood and pollination success in four hummingbird-polUnated cloud forest plant species. Ecology 67: 449-464 Feinsinger, P., H. M. Tiebout III and B. E. Yoimg. 1991. Do tropical birdpollinated plants exhibit density-dependent interactions? Field experiments. Ecology 72: 1953-1963 Fetcher, N., S. F. Oberhauer and B. R. Strain. 1985. Vegetation effects on microclimate in lowland tropiced forest in Costa Rica. International Joiu-nal of Biometeorology 29: 145-155 Forman, R. T. T. and M. Godron. 1986. Landscape Ecology. John Wiley & Sons. New York, New York, USA. Gates, J. E. and L. W. Gysel. 1978. Avian nest dispersion and fledging success in field-forest ecotones. Ecology 59: 871-883 Geiger, R. 1965. The climate near the ground. Harvard University Press. Cambridge, Massachusetts, USA. Gibbs, J. P. 1991. Avian nest predation in tropical wet forest: An experimental study. Oikos 60: 155-161 Gysel, L. W. 1951. Borders and openings of Beech-Maple woodlands in southern Michigan. Journal of Forestry 49: 13-19 Harris, L. D. 1988. Edge effects and conservation of biotic diversity. Conservation Biology 2: 330-332 Herrera, C. 1988. Variation in mutualisms: The spatio-temporal mosaic of pollinator assemblage. Biological Journal of the Linnean Society 35: 95Hester, A. J. and R. J. Hobbs. 1992. Influence of fire and soil nutrients on native and non-native annuals at remnant vegetation edges in the western Australian wheatbelt. Journal of Vegetation Science 3: 101-108 Hoppes, W. G. 1988. Seedfall patterns of several species of bird dispersed plants in an Illinois woodland. Ecology 69: 320-329

PAGE 94

87 Kapos, V. 1989. Effect of isolation on the water status of forest patches in the Brazilian Amazon. Journal of Tropical Ecology 5: 173-185 Kapos, V. and E. V. J. Tanner. 1985. Water relations of Jsungdcan upper montane rain forest trees. Ecology 66: 241-250 Kattan, G. H., C. Restrepo and M. Giraldo. 1984. Estructura de un bosque de niebla en la cordillera occidental, Valle del Cauca, Colombia. Cespedesia 13: 23-43 Kevan, P. G. 1975. Pollination and environmental conservation. Environmental Conservation 2: 293-298 Kuchler, A. W. 1973. Problems in classifying and mapping vegetation for ecological regionalization. Ecology 54: 512-523 Laur£ince, W. F. sind E. Yensen. 1991. Predicting the impacts of edge effects in fragmented habitats. Biological Conservation 55: 77-92 Laverty, T. M. 1992. Plant interactions for pollinator visitsA test of the magnet species effect. Oecologia 89: 502-508 Leopold, A. 1933. Game management. Charles Scribner's Sons. New York, New York, USA. Levey, D. J. 1988. Treefall gaps and the distribution of understory birds and shrubs in a tropical wet forest. Ecology 69: 1076-1089 Levin, D. A. and H. W, Kerster. 1974. Gene flow in seed plants. Evolutionary Biology 7: 139-220 Lovejoy, T. E., R. O. Bierregaard Jr., A. B. Rylands, J. R. Malcolm, C. E. Quintela, L. H. Harper, K. S. Brown, A. H. Powell, G. N. V. Powell, O. R. Schubart and M. B. Hays. 1986. Edge and other effects of isolation on amazon forest fragments, pages 257-285. in M. E. Soul6 (editor). Conservation biology: the science of scarcity and diversity. Sinauer Associates. Sunderleind, Massachusetts, USA. MacDougall, A. S. and M. Kellman. in press. The understory light regime and patterns of tree seedlings in tropical rip£irian forest patches. Journal of Biogeography Malcolm, J. R. 1991. The small mammals of Amazonian forest fragments: Pattern and process. University of Florida. Ph.D. Dissertation. Margules, C, A. J. Higgs and R. W. Rafe. 1982. Modem biogeographic theory: Are there any lessons for nature reserve design? Biological Conservation 24: 115-128

PAGE 95

d8 McClanahan, T. R. 1986. Pollen dispersal and intensity as criteria for the minimum viable population and species reserves. Environmentsd Management 10: 381-383 McCoy, E. D., S. S. Bell and K. Walters. 1986. Identifying biotic bovmdaries along environmental gradients. Ecology 67: 749-759 McDade, L. A. and P. Davidar. 1984. Determinants of fruit and seed set in Pavonia dasvpetala (Malvaceae). Oecologia 64: 61-67 Medina, E., E. Cuevas and P. L. Weaver. 1981. Composici6n foliar y transpiracidn de especies lenosas de Pico del Este, Sierra de Luquillo, Puerto Rico. Acta Cienti'fica Venezolana 32: 159-165 M0ller, A. P. 1989. Nest site selection across field-woodland ecotones: The effect of nest predation. Oikos 56: 240-246 Murcia, C. 1987. Estructura y din^tmica del gremio de colibries (Aves: Trochilidae) en un bosque andino. Humboldtia 1: 29-64 Murcia, C. 1990. Effect of floral morphology on pollen receipt and removal in Ipomoea trichocarpa . Ecology 71: 1098-1109 Murray, K, G. 1988. Avian seed dispersal of three neotropical gap-dependent pl£ints. Ecological Monographs 58: 271-298 Costing, H. J. and P. J. Kramer. 1946. Water and light in relation to pine reproduction. Ecology 27: 47-53 Palik, B. J. and P. G. Murphy. 1990. Disturbance versus edge effects in sug£irmaple/beech forest fragments. Forest Ecology and Management 32: 187202 Quintela, C. E. 1986. Forest fragmentation and differential use of natural and man-made edges by understory birds in central Amazonia. University of Illinois. M.Sc. Thesis. Ranney, J. W., M. C. Bruner and J. B. Levenson. 1981. The importance of edge in the structure and dynamics of forest islands, pages 57-95. in R. L. Burgess and D. M. Sharpe (editor). Forest islsmd dynamics in mandominated landscapes. SpringerVerlag. New York, New York, USA. Rathcke, B. 1983. Competition and facilitation among plants for pollination, pages 305-329. in L. Real (editor). Pollination biology. Academic Press, Inc. Orlando, Florida, USA. Ratti, J. T. and K. P. Reese. 1988. Prehminary test of the ecological trap hypothesis. Journal of Wildlife Management 52: 484-491

PAGE 96

89 Raw, A. 1989. The dispersal of euglossine bees between isolated patches of eastern Brazilian wet forest (Hymenoptera, Apidae). Revista Brasileira de Entomologia 33: 103-107 Romano, G. B. 1990. Invasibility of a mixed hardwood forest by Eupatorium rapillifnliiim and £. nnmpnaitifnlinm . University of Florida. M.Sc. Thesis. Roubik, D. W. 1989. Ecology and natural history of tropical bees. Cambridge University Press. New York, New York, USA. Santos, T. and J. L. Telleria. 1992. Edge effects on nest predation in Mediterranean fragmented forests. Biologiced Conservation 60: 1-5 SAS®, S. 1. 1. 1988. Technical report P-179. Additional SAS/STAT^" Procedures, Release 6.03. SAS Institute Inc. Cary, North CaroHna, USA. Saimders, D. A., R. Hobbs J. and C. Margules R. 1991. Biological consequences of ecosystem fragmentation: A review. Conservation Biology 5: 18-32 Schmitt, J. 1980. Pollinator foraging behavior and gene dispersal in Senecio (Compositae). Evolution 34: 934-943 Silander, J. A. and R. B. Primack. 1978. Pollination intensity and seed set in the evening primrose (Oenothera fruticosa) . American Midland Naturalist 100: 213-216 Simberloff, D. and R. N. Gotelli. 1984. Effects of insularization on plant richness in the prairie-forest ecotone. Biological Conservation 29: 27-46 Simberloff, D. S. and L. G. Abele. 1976. Island biogeography theory and conservation practice. Science 191: 285-286 Snow, A. A. 1982. Pollination intensity and potential seed set in Passiflora vitifolia . Oecologia 55: 231-237 Sork, V. L. 1983. Distribution of pignut hickory (Carya glabra ) along a forest edge transect, and factors affecting seedling recruitment. Bidletin of the Torrey Botanical Club 110: 491-506 Soule, M. E. 1986. The effects of fragmentation, pages 233-236. in M. E. Soule (editor). Conservation biology: The science of scarcity £ind diversity. Sinauer Associates, Inc. Simderland, Massachusetts, USA. Stephenson, A. G. 1981. Flower and fruit abortion: Proximal causes and ultimate functions. Annual Review of Ecology and Systematics 12: 253279

PAGE 97

do Stephenson, A. G. and R. I. Bertin. 1983. Male competition, female choice, and sexual selection in plants, pages 109-149. in L. Real (editor). Polhnation biology. Academic Press, Inc. Orlando, Florida, USA. Stiles, F. G. 1975. Ecology, flowering phenology and hummingbird pollination of some Costa Rican Heliconia species. Ecology 56: 285-301 Stiles, F. G. 1985. On the role of birds in the dynamics of Neotropical forests, pages 49-59. in A. W. Diamond and T. Lovejoy (editor). Conservation of tropical forest birds. Paston Press. Norwich, United Kingdom. Terborgh, J. 1976. Island biogeography and conservation: Strategies £ind limitations. Science 193: 1028-1029 Thomas, J. W., C. Maser and J. E. Rodiek. 1979. Edges, pages 48-59. in J. W. Thomas (editor). Wildlife habitats in managed forests: The Blue Mountains of Washington and Oregon. U. S. Forest Service Handbook No. 553. U.S. Government Printing Office. Washington, D. C, USA. Thomson, J. D. 1981. Spatial and temporal components of resource assessment by flower-feeding insects. Journal of Animal Ecology 50: 4959 Thomson, J. D. 1983. Component analysis of community-level interactions in pollination systems, pages in C. E. Jones and R. J. Little (editor). Handbook of experimental pollination biology. Van Nostrand Reinhold. New York, New York, USA. Wales, B. A. 1972. Vegetation analysis of north and south edges in a mature oak-hickory forest. Ecological Monographs 42: 451-471 Waser, N. M. 1978. Interspecific pollen transfer and competition between cooccurring plant species. Oecologia (Berlin) 36: 223-236 Waser, N. M. 1983. Competition for pollination and floral character differences among sympatric plant species: a review of evidence, pages in E. C. Jones and R. J. Little (editor). Handbook of experimental pollination biology. Van Nostrand Reinhold Company. New York, New York, USA. Whelan, R. J. and R. L. Goldingay. 1986. Do pollinators influence seed set in B^lksia pal^dosa Sm. and B. sninuloaa R. Br.? AustraHan Journal of Ecology 11: 181-186 Whelan, R. J. and R. L. Goldingay. 1989. Factors affecting finiit-set in Telopea gpgciosissima (Proteaceae): The importance of pollen limitation. Journal of Ecology 77: 1123-1134

PAGE 98

91 Wiens, J. A., C. S. Crawford and J. R. Gosz. 1985. Boundary dynamics: A conceptual framework for studying landscape ecosystems, Oikos 45: 421427 Wilcove, D. S., C. H. McLellan and A. P. Dobson. 1986. Habitat fragmentation in the temperate zone, pages 237-256. in M. E. Soul6 (editor). Conservation Biology. The science of scarcity and diversity. Sinauer Associates, Inc. Sunderland, Massachusetts, USA. Wilcox, B. A. 1980. Insular ecology and conservation, pages 95-117. in M. E. Soul6 and B. A. Wilcox (editor). Conservation ecology: An evolutionaryecological perspective. Sinauer Associates. Simderland, Massachusetts. USA. Wilhams-Linera, G. 1990a. Origin and early development of forest edge vegetation in Panama. Biotropica 22: 235-241 WilHams-Linera, G. 1990b. Vegetation structure and environmental conditions of forest edges in Panama. Journal of Ecology 78: 356-373 WilUamson, M. 1975. The design of wildlife preserves. Nature 256: 519 Willson, M. F. and F. H. J. Crome. 1989. Patterns of seed rain at the edge of a tropical Queensland rain forest. Journal of Tropical Ecology 5: 301-308 Winsor, J. A., L. E. Davis and A. G. Stephenson. 1987. The relationship between pollen load and fruit maturation and the effect of pollen load on offspring vigor in Cucurhita SSSSI. American NaturaHst 129: 643-656 Wright, S. J. 1991. Seasonal drought and the phenology of imderstory shrubs in a tropical moist forest. Ecology 72: 1643-1657 Wright, S. J., J. L. Machado, S. S. Mulkey and S. A. P. 1992. Drought acclimation among tropical forest shrubs ( Psvchotria . Rubiaceae) Oecologia 89: 457-463 Yahner, R. H. 1988. Changes in wildlife commimities near edges. Conservation Biology 2: 333-339 Yahner, R. H., T. E. Morrell and J. S. Rachael. 1989. Effects of edge contrast on depredation of artificial avian nests. Journal of Wildlife Management 53: 1135-1138 Yahner^ R. H. and A. L. Wright. 1985. Depredation on artificial ground nestsEffects of edge and plot age. Journal of Wildlife Management 49: 508513 Yoakum, J., W. P. Dasmann, H. R. Sanderson, C. M. Nixon and H. S. Crawford. 1980. Habitat improvement techniques, pages 329-403. in S.

PAGE 99

92 D. Schemnitz (editor). Wildlife management techniques manual. The Wildlife Society. Washington, D. C, USA. Yoakum, J. P. and W. P. Dasmann. 1971. Habitat manipulation practices, pages 173-231. in R. H. Giles (editor). Wildlife management techniques. The WildUfe Society. Washington, D. C, USA. Zimmerman, M. 1980. Reproduction in Pnlemnnii^] [X)r Competition for pollinators. Ecology 61: 497-501 Zimmerman, M. and G. H. Pyke. 1988. Reproduction in PnlemnmnnT Assessing the factors limiting seed set. American Naturalist 131: 723738

PAGE 100

BIOGRAPHICAL SKETCH CaroKna Murcia was bom on 25 October 1960 in Bogotd, Colombia. She gradiiated as a zoologist from Universidad del Valle in Cali, Colombia, in 1983. In 1984 she married Gustavo Kattan, a fellow biologist, and started graduate school at the University of Florida. She obtained her M. Sc. in Zoology in 1987 imder the direction of Dr. Peter Feinsinger, working on the floral morphology of morning glories. Her Ph.D. work was also conducted imder the direction of Dr. Feinsinger. Her research interests are in plant reproductive biology, plant-polUnator interactions and conservation biology. 93

PAGE 101

I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fiilly adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Frank G. Nordlie, Chairman Professor of Zoology I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Richard A. Kiltie Associate Professor of Zoology I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Dotiglas J. Levey Assistant Professor of Zoology I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy.

PAGE 102

I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. This dissertation was submitted to the Graduate Faculty of the Department of Zoology in the College of Liberal Arts and Sciences and to the Graduate School and was accepted as partial fulfillment of the requirements for the degree of Doctor of Philosophy. May 1993 Kent H. Rediord \ Associate Professor of Latin American Studies Dean, Graduate School