mafia... a
{u

..

mini.

9 6

nu.

,'

‘-

fiwfi
$3mb
.- . r. .
f: L
. h r .

.. JiAM

.snfi

no.

 

‘flm

uni...“-

nA ‘. anmmu
$2 in;

A, ‘01
I: t.) ‘00!

QEWH m. t...
.3. .

 

. ‘l. .l. ‘ .\ . .:
gave? _.... ; ...,..: . , . .4..F-...;.H« L. Li»,

 

THESIS

.1 t.) "D 3

This is to certify that the

dissertation entitled

LANDSCAPE FORAGING ECOLOGY OF GIANT HONEY BEES,
APIS DORSATA,
IN AN INDIAN FOREST

 

presented by

Puja Batra

has been accepted towards fulfillment
of the requirements for

 

Ph. D. eein Zoology/Ecology,
EvolutIgnary Biology, and Behavior

fl( fléQLfl/é

Major professor

 

 

Q/wl/a

MS U is an Affirmative Action/Equal Opportunity Institution 0- 12771

 

LIBRARY
Michigan State
University

 

 

 

PLACE IN RETURN Box to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

AUG 1 1 200

03 25 n3

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 eJCIRC/DateDquGS-ms

LANDSCAPE FORAGING ECOLOGY OF GIANT HONEY BEES, APIS DORSATA,
IN AN INDIAN FOREST

By

Puja Batra

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Zoology

2002

ABSTRACT

LANDSCAPE FORAGING ECOLOGY OF GIANT HONEY BEES, APIS DORSATA,
IN AN INDIAN FOREST

By

Puja Batra

I conducted a study of the spatial and temporal foraging dynamics and pollen
foraging of giant honey bees, Apis dorsata, in Kamataka, India. Through observations of
the honey bee dance language over the course of two flowering seasons at several nesting
aggregations I inferred that bees forage at maximum distances of over nine km from the
nest, thus covering an area of over 250 kmz. Ninety five percent of their flights occur at
distances within 2.7 km from the nest. They did not exhibit predictable expansions or
contractions of flight range according to the type of forest they were nesting in, year in
which data were collected, or week in the flowering season. Instead, flight range
variation was due to week* site differences, and week* colony (site) differences. This
suggests that colonies adjust their flight distance according to local fluctuations in
resources, and shift their foraging locations such that they do not overlap with
neighboring colonies.

I examined the feces of colonies over the course of two flowering seasons to
discover which plants A. dorsata utilizes as its pollen resources, and also to examine
whether they exhibit preferences for certain plants. I used data on the relative frequency
of the major tree species to quantify whether bees used pollens in proportion to their
occurrence in the environment. Across several sites and weeks, bees overutilized pollen

of Catunaregam spinosa (Rubiaceae) relative to its abundance in the forest, overused the

 

relatively rare genus Schleichera (Sapindaceae), and underused the dominant genus
Terminalia (Combretaceae). When phenological variation in plants was taken into
account, bees still exhibited a preference for the same taxa. These results suggest that
Schleichera and Catunaregam may be important for maintaining the population of A.
dorsata, which in turn provides pollination to many other plant taxa.

I used Geographic Information Systems (GIS) to examine how the spatial and
temporal heterogeneity in food plant distribution influence bees’ flight range and pollen
usage. Stepwise and multiple regression revealed that the flight range of A. dorsata is
negatively correlated with Terminalia flowering availability. Though they do not overuse
it for pollen, bees’ apparent tracking of Terminalia may be due to the importance of this
genus as a nectar source. Pollen preferences remained largely similar to the results
above, although some site-specific patterns of preferences for pollens of different plants
emerged when spatio-temporal heterogeneity was accounted for. These results
demonstrate the power of using 618 methods alongside field observations to understand
spatio-temporal variation in ecological systems.

Finally, I quantified visitation rates of A. dorsata to flowers of C. spinosa at
distances within two km from the nest aggregation, and at distances beyond two km to
test the hypothesis that trees within the typical foraging range of bee nest sites experience
greater visitation than trees outside the foraging range, ultimately structuring the bee
plant community around nest sites. There were no differences among visitation rates to
the two distances, but a trend in the expected direction suggests that a more detailed study
may be a promising avenue of research. This research sheds new light on the ecology of

a crucial pollinator, and will be useful for conservation corridor planning in Asia.

Copyright by
PUJA BATRA

2002

ACKNOWLEDGMENTS

This dissertation is the collective work and cumulative result of years of encouragement,
assistance, advice, and efforts of many people and institutions. First and foremost, my
parents, Usha and Narinder Batra, have all my life been a source of inspiration,
encouragement and support for my efforts and escapades, and this dissertation was no
exception. I thank them from the bottom of my heart. My thanks also to Leena, my

sister, for her positivity and faith in me.

My advisor Dr. Fred Dyer changed the course of history for me, a newly orphaned and
floundering grad student, searching about for an advisor and project, when he told me
about the bees and some of his ideas about studying them. Over the years I have learned
much from Fred about the value of different approaches to scientific inquiry, attention to
detail, and taking risks in the name of discovery, among other things. His knowledge,
ideas, enthusiasm, and vision are pervasive throughout this dissertation, and have made it

a work of which I am truly proud.

Dr. Kama] Bawa has been a mentor and friend for the last decade from whom I have
learned much about science, conservation, and the meaning of grace, friendship, and
support. Kama] has come through with institutional assistance, financial support,
intellectual challenge, scientific advice, ethics, humor, and kind words on every occasion
possible. His unrelenting support saw me through many moments of discouragement and

self-doubt, and I hope that I have done justice to his belief in my abilities.

It has been a great privilege to have had the opportunity to work with several people in
the field: M. Ketha Gowda, who provided me with impeccable and tireless assistance in
the field at all hours of the day and night, and made sure I stayed safe and happy; Jade
Swamy, a very skilled driver and gifted storyteller whose memory and insights provided
much engaging conversation and entertainment; Sidda, who worked with me as a driver
in my first field season in BRT, and whose patience, concern, and indulgence saw me
through many rough moments. They all taught me Kannada, tried to teach me how to
climb trees to get away from charging wildlife, and caused me to think and learn about
the living human history of the forest where I was working. I express my sincere
gratitude to these three people for their friendship and caring, and for keeping me
smiling, and to many other friends from the Soliga and ER. Hills communities that I
made while in BRT. From them I began to understand how big the world is, and yet how

small it is.

Many institutions in India provided assistance which allowed me to conduct my research.
First and foremost, I express my deepest gratitude to the Ashoka Trust for Research in
Ecology and the Environment (ATREE) for their continued assistance of many kinds.
ATREE provided me with housing, computers, work space, permits, seminars, contacts,
advocacy, and the list goes on. To Dr. K.N. Ganeshaiah, Dr. R. Uma Shaanker, and all
the past and present staff of ATREE, particularly Mr. N. Ramesh, I owe an

immeasurable debt of appreciation for logistical, intellectual, and moral support.

vi

I thank also Vivekananda Girijana Kalyana Kendra (VGKK) in B.R.Hills for help with
many aspects of my field work. I am most grateful for the advocacy and encouragement
of Dr. H. Sudarshan, Mallesh, and everyone else at VGKK. The Kamataka Forest
Department was an enthusiastic supporter of this work throughout, and I greatly
appreciate the unimpeded access I had to the forest and to the expertise of KFD officials.
Mr. Vijay Kumar Gogi, then the Deputy Conservator of Forests in ER. Hills was
especially helpful, as were all the forest department officials with whom I came in

contact.

The French Institute of Pondicherry, under the directorship of Dr. F. Houllier, and later
Dr. D. DePommier, generously allowed me to use their excellent palynology facilities
under the leadership of Dr. Raymonde Bonnefille. While there, I could not have
succeeded without the expertise of Mr. S. Prasad, palynologist extraordinaire, Mr.
Sabaraj, who made acetolysis a breeze, and my friends Doris Barboni, Mohan Seetharam,
Beatrice Moppert, and Thomas Bouix. Thanks to the Center for Ecological Studies,
Indian Institute of Science, and especially Dr. R. Gadagkar, for assistance of many kinds
during my stays in India, and to Mr. Utkarsh Ghate, Ms. Moushumi Sen Sarma, and
others at 1.1. So. for their help and interest in my work. Thanks also to Mr. D. B.
Mahindre and his colleagues at the Central Bee Research and Training Institute in Pune.
Thanks to my cousins Dilip Batra, and Madhu Sood and their families for putting me up

in Bangalore on a number of occassions.

vii

The residents and staff of the Biligiri field station from the period between 1996-1999
tolerated my demanding work hours and American sensitivities, and all that they entailed.
I especially thank Mr. R. Siddappa Setty, for his great generosity, for teaching me the ins
and outs of life in BRT, and for making me part of the family there. I remember
posthumously my friend Sidda, for his caring and good natured spirit. Rajanna, Mada,
Mahadeva, and many others took care of me at the field station. My dear friends
Sumathi and Sridhar provided me with kombucha, many cups of tea and dinners,
laughter, warmth, and conversation which I will continue to recall. Special thanks to R.
Ganesan, an extraordinary botanist who has freely shared his knowledge with me on
many occasions. Thanks to the following people for additional field assistance: Dase
Gowda and S. Kethe Gowda helped with vegetation transect work, and identified
hundreds of trees. Kumbha helped with gathering phenology data, and several other
people accompanied me in the field at various times. The honey collectors from Hosa
Podu, Muttuga Gadde, Bangale Podu, Sige Betta Podu, Kanneri Colony, and Keredimba
Podu allowed me to tag along on honey collecting expeditions, helped me collect pollen
combs from colonies, and always made sure I had my fill of forest honey. Some of them
also indulged my crazy fantasies of capturing and settling a bee colony. Though the bees

outsmarted me every time, I am grateful for the assistance of these people.

In the US, there are many people and institutions whose support was integral to the
completion of this work. I gratefully acknowledge the generous financial support during
my graduate career for stipend, research, and travel support from a number of programs

at Michigan State University: the Department of Zoology, the Graduate Program in

viii

Ecology, Evolutionary Biology, and Behavior, the Graduate School, the Office of Urban
Affairs, Kellogg Biological Station’s Research Training Grant and Long Term Ecological
Research programs, and the International Studies Program. Additionally, I was very
fortunate to receive extramural funding from the Garden Clubs of America;
Conservation, Food, and Health Foundation; and the National Geographic Society

Committee for Research and Exploration.

Many thanks to my dissertation committee, who greatly improved this work and whose
criticisms were always constructive: Drs. Guy Bush, Bryan Epperson, Catherine Lindell,
and Pete Murphy. I wish to acknowledge also my first advisor at MSU, Dr. Steve
Tonsor, who provided my first and perhaps last foray into theoretical evolutionary
biology, and from whom I learned a lot in my first few months at MSU. Thanks to Dr.
Andy Jarosz whose moral support went a long way early on. I greatly appreciate the
input and encouragement of past and present members of the Dyer lab and the constant
and reliable assistance of the past and present members of Zoology office staff, especially

Tracey Bamer, Lisa Craft, Chris Keyes, and Judy Pardee.

Finishing graduate school would not have been possible for me without my family and
friends, and the wisdom, humor, sanity, challenge, and perspective they have brought to
my life: Beth Capaldi for bee lessons, and all kinds of genuine caring; Erin Boydston for
dissertation advice, GIS help, listening , and for knowing enough to get me back to TX;
Lisa Horth for being there at any time; Gray Stirling for lessons of honesty and respect;
Mike Ryan, for being a true friend and letting me be one too; Paco Moore, my first lab

mate, for not always saying what I wanted to hear, but always being reassuring; and my

ix

other first lab mate Jeff White, and his family, for reminding me of the multitude of
things that can make people happy; Martha Tomecek and Michel Cavigelli and Katia for
Thanksgiving dinner and gardening; Walter Pett, for Bar-B-Que in Battle Creek; The
Wedgies-- Kali Majumdar and Brad Ennisch, Christie Boger, Laurie Boger, Jo Lesser,
Kevin Nicholoff, for mornings at Beaner’s and for keeping me social; Todd Ross, for
getting me to laugh at myself; Paula Lane, for some important “Come to Jesus” talks and
David Stocking, for good wine and conversation; Corine Vriesendorp, for debunking the
bread myth; Aditi Sinha for hours spent sitting under the tree; Nitin Rai for hours of
loitering time at Koshy’s; Heather Eisthen, for indulging my complaining; Marianne
Simmons; for life on the farm down in Drippin’; Dan Gohl, my dissertation babysitter
and friend from way way back, for getting me to sit down and write; Drs. Gustavo
Fonseca and Tom Lacher at Conservation International, for providing me with the space,
time, and threat of continual embarrassment needed to finish this up, and for giving me a
job before I did; D.H., who in some way started this journey so many years ago with
stories of the magical forest; and finally, my extended family in India and in Texas, for

helping me keep some perspective on who I am, and what’s important.

TABLE OF CONTENTS

List of Tables .................................................................................................................... xiii
List of Figures .................................................................................................................. xiv
Chapter 1: Introduction ....................................................................................................... 1
Why study honey bees in tropical Asia? ..................................................................... 1

The study species: Giant honey bees ........................................................................... 4

The study site: Ecological and human landscape ........................................................ 7
Chapter 2: Dynamics of flight range across spatial and temporal scales .......................... 16
Introduction ................................................................................................................... 16
Distance dialect curve ............................................................................................... 20
Forage map data collection ....................................................................................... 22
Forage map construction ........................................................................................... 24
Statistical analysis ..................................................................................................... 26
Results ........................................................................................................................... 27
Distance dialect curve ............................................................................................... 27
Forage maps .............................................................................................................. 27
Flight range distribution and foraging area ............................................................... 28
Analysis of variance .................................................................................................. 28
Discussion ..................................................................................................................... 29
Chapter 3: Pollen diet composition and food plant preferences ........................................ 49
Introduction ................................................................................................................... 49
Methods ......................................................................................................................... 54
Fecal sampling ........................................................................................................... 54
Vegetation sampling plots ......................................................................................... 55
Flowering Phenology ................................................................................................ 56
Palynology ................................................................................................................. 57
Statistical analysis ..................................................................................................... 59
Results ........................................................................................................................... 60
Discussion ..................................................................................................................... 64
Chapter 4: Spatio-temporal dynamics of foraging ............................................................ 87
Introduction ................................................................................................................... 87
Methods ......................................................................................................................... 89
Spatio-temporal correlates of flight distance ............................................................ 90
Spatio-temporal variation in food plant preference ................................................... 94

xi

Results ........................................................................................................................... 95

Correlates of flight distance ...................................................................................... 95
Food plant preference ................................................................................................ 96
Discussion ..................................................................................................................... 97
Chapter 5: Nesting aggregations, pollination, and forest structure ................................. 107
Introduction ................................................................................................................. 107
Methods ....................................................................................................................... 1 10
Field methods .......................................................................................................... 110
Statistical analysis ................................................................................................... 11 1
Results ......................................................................................................................... 1 12
Discussion ................................................................................................................... 1 12
Appendix I ................................................................................................................... 117
Appendix II ................................................................................................................. 119
Appendix III ................................................................................................................ 120
Appendix IV ................................................................................................................ 122
Appendix V ................................................................................................................. 123
References ................................................................................................................... 124

xii

LIST OF TABLES

Table 2.1. Analysis of variance results. The dependent variable was normalized using a
natural log transformation. y = ln flight distance. Random variables are year, site (year),
colony (site, year), week* site (year) and week*colony (site,year) .......................... 36

Table 2.2. MANOVA results. Dependent variables are the x and y coordinates of
foraging location. For each site, the Pillai’s Trace F ratio is reported with p-value in
parentheses ........................................................................................... 37

Table 3.1. Shannon-Weiner indices for pollen diet diversity .................................. 76

Table 3.2. Composition of forest tree community, overall bee pollen diet, and individual
plant genus contributions to total G-score for sites pooled across weeks. Percent of a
site’s fecal pollens is given in parentheses. Total G-scores for each site are given. G-tests
compare each site’s values, except “other” category, to the composition of percent bee
plants ................................................................................................. 77

Table 3.3. Partial G-scores for 1998 pollen diet composition compared to forest
composition with plant taxa weighted by flowering status per week ......................... 81

Table 4.1. Results of stepwise regression for BK 98 site. These three out of the original
six taxa emerged as being the most valuable predictors of flight range. Here, “sig prob”
is the theoretical significance probability, because in stepwise regression it cannot be
determined absolutely. Cp is Mallow’s criterion, and “p” is the number of parameters.
Based on the results here, Canthium was removed from the model before performing

multiple regression. (See text for details.) ...................................................... 103

Table 4.2. Results of stepwise regression for KA 98 site. These three out of the original
six taxa emerged as being the most valuable predictors of flight range. Here, “si g prob”
is the theoretical significance probability, because in stepwise regression it cannot be
determined absolutely. Cp is Mallow’s criterion, and “p” is the number of parameters.
Based on the results here, Schleichera was removed from the model before performing

multiple regression. (See text for details.) ..................................................... 103
Table 4.3. Results of multiple regression for sites BK 98 and KA 98 .................... 104
Table 4.4. G-tests for KA 98 site and BK 98 site comparing spatio-temporally explicit
forest floral composition to pollen diet composition. DF = 5, G critical = 11.07 (p<.05);
G critical = 20.515 (p<.001) ..................................................................... 105

Table 5.1. Results of Wilcoxon rank sums test of visitation rate to flowers of
C. spinosa ............................................................................................ 116

xiii

LIST OF FIGURES

Figure 1.1. a. Apis dorsata nest b. Apis dorsata nest aggregation of over 80
colonies ............................................................................................... 14

Figure 1.2. Map of India and landscape map of Biligiri Rangaswamy Temple (BRT)
Wildlife Sanctuary, Karnataka, India. Map of BRT: Ramesh, B.R., S.Menon, and
K.Bawa, 1998. Some legend entries from the latter have been changed to be consistent
with wording in the text. Red circles mark locations of study areas with site names on
left side of map ...................................................................................... 15

Figure 2.1. Dance language. The vertical angle relative to the upward direction at which
a dancer runs (right panel) corresponds to the horizontal angle between the solar azimuth,
the patch of forage, and the nest. She repeats this waggle run several times, by stopping,
returning to her original position and starting again. Distance is encoded by the duration
of the waggle run .................................................................................... 38

Figure 2.2. Distance dialect curve for the waggle dances of Apis dorsata .................. 39

Figure 2.3. Basavanakadu (BK 97) site forage maps throughout 1997 flowering season.
Symbols indicate different colonies. Circles are placed at 1 km increments ................ 40

Figure 2.4. Beduguli (BG 97) site forage maps throughout 1997 flowering season.
Symbols indicate different colonies. Circles are placed at 1 km increments ................ 41

Figure 2.5. Doddesampige (DS 97) site forage maps throughout 1997 flowering season.
Symbols indicate different colonies. Circles are placed at 1 km increments ................ 42

Figure 2.6. Kamari (KA 97) site forage maps throughout 1997 flowering season.
Symbols indicate different colonies. Circles are placed at 1 km increments ................ 43

Figure 2.7. Basavanakadu (BK 98) site forage maps throughout 1998 flowering season.
Symbols indicate different colonies. Circles are placed at 1 km increments ................ 44

Figure 2.8. Kamari (KA 98) site forage maps throughout 1998 flowering season.
Symbols indicate different colonies. Circles are placed at 1 km increments ................ 45

Figure 2.9. Sige Gudi (SG 98) site forage maps throughout 1998 flowering season.
Symbols indicate different colonies. Circles are placed at 1 km increments ................ 46

Figure 2.10. a. 1997 flight range distribution combined over two habitat types and all

colonies. Inset shows flight range in evergreen (upper) and deciduous (lower) habitats.
b. 1998 flight range distribution ................................................................... 47

xiv

Figure 2.11. Flight range distribution 1997 and 1998 combined. 50th, 75th, and 90th
quantiles are indicated .............................................................................. 48

Figure 3.1. Saturation curves for randomly selected pollen samples. Sampling effort
of 300 grains captured most of the diversity in pollen taxa .................................... 79

Figure 3.2. Light micrographs of pollen taxa which were common in fecal samples. ....80

Figure 3.3. Contribution of each pollen taxon to the total pooled across sites and
weeks. Numbers above bars indicate percent of total, y-axis values indicate cumulative
percent of total ....................................................................................... 81

Figure 3.4. Frequencies of pollen types found in feces sampled at each site, pooled
across weeks. First two bars respectively show frequencies of bee pollen plants in the
entire BRT sanctuary overall, and frequencies of those plants relative to other bee

plant taxa .............................................................................................. 82

Figure 3.5. Weekly fecal samples’ pollen composition. A. BK 97 site B. KA 97 site
C. BG 97 site D. DS 97 site ...................................................................... 83

Figure 3.6. Weekly fecal samples’ pollen composition. A. BK 98 site B. KA 98 site
C. SG 98 site ......................................................................................... 84

Figure 3.7. Comparison of flowering phenology weighted by relative abundance of
major bee plants in dry deciduous forest. Numbers on y-axis represent the mean
flowering status of all individuals on the transect per species where flowering values

fall between 0 and 3 (see text for details.) ........................................................ 85
Figure 3.8. C. spinosa male flower. Note pollen on the central pseudo-stigma ............ 86
Figure 3.9. Apis dorsata foraging on male Catunaregam spinosa flowers. Note pollen

deposited on the head .............................................................................. 86
Fig 3.10 Terminalia crenulata flowers ........................................................... 86

Figure 4.1. Example of weeks 1-6 of forage maps for KA 98 and BK 98 overlain onto
weekly interpolated flowering index surfaces for Catunaregam spinosa. Darker shades of
gray indicate increasing pollen availability. Equally spaced dots are two kilometers apart,
and denote the vegetation sampling grid. Different symbols are foraging points for
different colonies. The upper aggregation is KA, and the lower is BK ................... 106

XV

CHAPTER 1: INTRODUCTION

WHY STUDY HONEY BEES IN TROPICAL ASIA?

A major challenge in understanding the maintenance of biological diversity in tropical
forests is to comprehend the role that is played by animals in the pollination of forest
plants. In the Neotropics, an estimated 98% of forest tree species are dependent on
animal mutualists for pollination, with bees representing the largest fraction of that
fauna (Bawa 1990). The behavior of pollinators has a major influence on
interconnections among organisms in these complex and threatened ecosystems. For
example, distances traveled by pollinators will determine distance of pollen flow, and
thus can profoundly influence the genetic structure of a plant population. At the
community level, because the majority of tropical pollinators are generalists and not
specialists on specific plant species (Roubik 1993), seasonal differences in food plants
utilized by one pollinator species may create linkages between different plant Species.
Therefore, population changes in one food plant may have cascading effects on the

forest community via its effects on a shared pollinator.

Most studies of tropical plants and their pollinators have been conducted in the New
World; however, the bee faunas of the tropical regions differ markedly. Whereas
pollination in the Neotropics is mediated by a highly diverse bee fauna, in Asian tropical
forests pollination is presumably dominated by the two to four sympatric species of
eusocial honey bees (genus Apis) in any given locale (Ruttner 1988; Roubik 1990).

Perhaps because behavioral specializations of Apis confer superior competitive ability,

the Asian tropics are relatively depauperate in solitary bees when compared to the New
World, in which there are no native species of Apis (Roubik 1990). Thus, pollination
ecology, and therefore population and community ecology, of forest plants in the Asian
tropics may substantially differ from the situation in the Neotropics, simply due to
differences in behavioral attributes of pollinators, such as whether they have a colonial
or a solitary lifestyle, and their nesting requirements, foraging ranges, and degree of

dietary generalism.

An understanding of pollination by Apis will have ramifications for forest management,
conservation and reforestation plans. Generalizations made solely on the basis of
Neotropical studies are likely to be inappropriate in Asian forests. We currently know
very little about the actual role of honeybees in the reproduction of Asian tropical plants.
In order to make predictions about the impact of major habitat fragmentation or
alteration on loss of biodiversity and regeneration, it is crucial to have baseline data
from a relatively contiguous forest. In particular, it is important to have data on the
dynamics of flight range and foraging area, and the consequent distances of potential
pollination by Apis throughout the flowering season, and on their use of food plants over
the course of the flowering season. How those two attributes interact as resource levels
vary over time will in large part determine the ability of Apis to absorb larger scale
habitat changes. I report here my findings on a Study of the pollen foraging ecology of

giant honey bees, Apis dorsata Fabricius in a tropical forest of southern India.

Various aspects of giant honey bee ecology and behavior have been studied, including
their mating behavior (Koeniger et al. 1990); seasonal movements (Koeniger and
Koeniger 1980; Venkatesh and Reddy 1989; Dyer and Seeley 1994); nocturnal foraging
behavior (Dyer 1985); ecophysiology (Dyer and Seeley 1987); colony defense behavior
(Seeley et a1. 1982; Kastberger et a1. 1996; Kastberger et al. 1998); agricultural
importance (Abrol and Kapil 1996); beekeeping (Nguyen et a1. 1997) #49; nest site
preferences (Starr et a1. 1987). However, the pollination ecology of these bees has
scarcely been studied in non-agricultural systems. Although visitation by bees does not
ensure pollination, an understanding of their patterns of resource usage provides the
foundation for explicit studies of the bees’ role in pollen transfer. If it is indeed true that
Apis dorsata are pivotal pollinators in these forests, identification of major components
of their pollen diet may identify plant species which serve as "keystone" links by
maintaining honeybee populations at sizes large enough to serve the diverse forest

community as a whole.

Honeybees are ideal organisms for studying foraging behaviors because a great deal of
information can be obtained at the nest. Their unique system of communication via the
dance language is easily interpreted and provides the observer with information about
the distance and direction of food sources exploited by the colony (Frisch 1967). I have
constructed daily "forage maps" (Visscher and Seeley 1982) from hourly dance
observations to identify the locations and number of patches being visited by a colony.
In addition, I identified pollens utilized by the bees by sampling feces under the nests on

a weekly basis. When taken over the course of entire flowering seasons, these data

reflect the seasonal changes in foraging range and preferences of several colonies in
different areas of the habitat. When the data were used in combination with the
extensive maps and data compiled on the area's flora by local collaborators, I was able to
address a number of questions. First, I could determine the typical foraging ranges of
giant honey bees, and thus, not only the scale of possible pollen mediated gene flow, but
also the amount of forest area that is necessary to maintain that bee population size.
Secondly, I was able to identify what Apis dorsata is using as its important pollen food
plants, thus which plant species they are potentially pollinating, and also which plant
species are crucial in maintaining the bee population. Third, I could examine how the
variation in the distribution and availability of their heavily utilized resources influences

their foraging behavior.

The specific aims of my research were to conduct a study of the foraging distances of
Apis dorsata across different spatial and temporal scales, a topic I address in chapter 2.
In chapter 3, I elucidate the major pollen components of the bees’ diet relative to what is
available in the environment. In chapter 4 I use Geographic Information Systems (GIS)
to explore the bees’ foraging dynamics in the context of the spatial and temporal
heterogeneity in the distribution of their resources, and finally, in chapter 5 I begin to
address how the above findings may have interesting ramifications for forest community

SII'UCIUI'C.

THE STUDY SPECIES: GIANT HONEY BEES
Apis dorsata Fabricius, commonly referred to as the “rock bee” or “giant honey bee,” is

a large open-nesting species which hangs a single large comb from high tree limbs and

rock cliffs, and lives in colonies of 40,000 to 60,000 bees (figure 1.1). The species
ranges throughout tropical and parts of subtropical Asia and is believed to be among the
most important pollinators of trees in areas where they occur. Moreover, they are
important pollinators of many economically important fruit, oilseed, and fiber crops
(Sidhu and Singh 1962; Sihag 1986; Verma 1987; Crane 1991; Abrol and Kapil 1996;

Sinha and Atwal 1996).

Like the other eight described species of honey bees, A. dorsata is a eusocial species in
which there is only one reproductive female, commonly referred to as the queen. All
other female bees in the colony are workers that perform tasks needed for colony
maintenance, such as feeding and grooming the queen and larvae, cleaning the empty
comb cells of debris and parasites, thermoregulating and guarding the nest, and foraging
for and storing food (Winston 1987). Honey bees maintain perennial colonies, and are
largely considered to have a generalist diet in that many different plants can be exploited

for nectar and pollen.

As with all honey bees, colonies reproduce by fission after the parent colony has reached
a critical size. This phenomenon is referred to as swarming, and is not to be confused
with the dramatic colony migrations that are a unique aspect of the natural history of
Apis dorsata and its Himalayan sister species, A. laboriosa. The migrations, in which
colonies presumably track resources according to the rainy/dry season cycle, are not
well understood, but they may occur over a distance of 100 km or more (Koeniger and

Koeniger 1980). Another fascinating aspect of the biology of giant honey bees in some

parts of their range including India, is their large nesting aggregations of up to one
hundred colonies on a single tree or cliff face (figure 1.1b). The implications of the
yearly migrations and nesting aggregations on individual colony foraging behaviors and
pollination remain unexplored and fascinating questions. Before such questions can be
investigated however, a clear understanding of relationships between Apis dorsata and

its forest food plants is necessary.

Past studies on the foraging range of pollinating insects have relied on inference of
potential flight range by displacement and return of marked insects released at various
distances from their nests (e. g., J anzen 1971; Roubik and Aluja 1983). Such studies
estimate an upper bound for foraging range, but do not help to resolve the question of
how the insects are typically foraging in the environment throughout the course of the
flowering season. Honey bees are ideal for study of this question, as an enormous
amount of information about their resource use can be obtained by behavioral
observations at the nest itself. Various aspects of pollinator foraging behavior, here most
importantly the distance and direction of a foraging location, are readily obtained
through observations of activity at the nest (Visscher and Seeley 1982). In the following
chapter I rely heavily on the techniques and insights from the classic work of Karl von
Frisch on the unique communication system of Apis, known as the dance language

(Frisch 1967), to infer the foraging locations of giant honey bees.

THE STUDY SITE: ECOLOGICAL AND HUMAN LANDSCAPE

Along the western coast of peninsular India runs the Western Ghats, a hill range that has
been designated as one of the global “hotspots” of biodiversity, due to its high levels of
plant endernism as well as the high degree of threat it faces from human activities
(Myers et a1. 2000). Along the eastern coast of the peninsula runs another hill range, the
Eastern Ghats, a much drier region due to its position in the rain shadow of the Western
Ghats. I conducted this work in the Biligiri Rangaswamy Temple Wildlife Sanctuary
(BRT), located between 11° 40'-12° 09' N and 77° 05'-77° 15' E in the Biligiri Rangan
(BR) Hills of Kamataka, India (figure 1.2). The BR Hills are a low hill range that forms
a saddle between Eastern Ghats and Western Ghats. Although they contain floristic
elements of both regions, and form a stepping stone between different floristic regions of
Asia (Ramesh 1989), their floral affinities more closely overlap with the Western Ghats,
especially in the dry deciduous forest, the predominant vegetation type (Murali et a1.
1996). For this reason BR Hills are often considered the easternmost spur of the

Western Ghats, and mark the eastern range edge of many Western Ghats species.

BR Hills experiences both of the monsoons of southern India, and thus has two rainy
seasons in addition to “summer showers,” convectional storms which occur prior to the
onset of the Indian southwest monsoon. The southwest monsoon is responsible for most
of the rainfall in the region and spans the period from June through September.
Following that, BR Hills also experiences the retreating northeast monsoon, primarily
caused by cyclonic activity off the Bay of Bengal, which hits the regions during the

period from October through December (Ramesh 1989). In the deciduous forest, leaf

shed and new leaf flush occurs during the dry season from February to April. This is
also the main flowering season for trees, and the season during which Apis dorsata

migrates into and forages in the area.

Biligiri Rangaswamy Temple Wildlife Sanctuary (BRT), a protected area covering 540
km2 within the BR Hills region, reflects much of the heterogeneity of habitats and
history of the Western Ghats (figure 1.2). Among the predominant vegetation types is
the low elevation (700-900 In) scrub forest, dominated by low, dense undergrowth and a
canopy height of 10 m or less. The scrub forest is characterized by relatively low
average annual rainfall (750 i 130 mm), and mainly occurs on the periphery of the
sanctuary. It is typified by thorny vegetation in the families Mimosoideae, Rubiaceae,
Rhamnaceae, and Euphorbiaceae (Shankar et a1. 1998). Apis dorsata can sometimes be
found nesting on the large trees that occur along riparian zones, but does not commonly

nest in this forest type (pers. obs.).

The second of the major habitat types in BRT is the dry deciduous forest, occurring at
elevations between 1000-1400m (Ramesh et a1. 1998). This type is dominated by
Anogeissus and Terminalia, both in plant family Combretaceae, as well as Grewia,
Dalbergia, Pterocarpus, and other genera. The canopy structure is relatively open, and
includes trees of 20 m or more in height. This forest type occurs across a variety of
rainfall regimes, and hosts several large aggregations of giant honey bees. The majority

of my research was conducted on colonies nesting in the dry deciduous forest.

A third major vegetation type is evergreen forest which occurs mainly in narrow bands
and patches in the higher elevations of BRT (1200-1400 In). The average annual rainfall
in this type is between 1400-1800 mm depending on the hill chain and slope upon which
the patch resides. Evergreen forest here includes areas with a canopy height of more
than 20 In. The evergreen forests are very similar in physiognomy and composition to
the riparian forests that occur throughout BRT, dominated by Elaeocarpus, Canarium,
and Michaelia (Ramesh 1989). Apis dorsata colonies can often be found in the tall

evergreen forests, although large aggregations are uncommon.

Another category of evergreen forest occurs above 1400 m elevation. The stunted shola
forests are high elevation patches of evergreen forest which occur in hydrogeological
pockets of moisture surrounded by grasslands. Their physiognomy differs markedly
from that of evergreen forest, with a canopy of around 15 m maximum, and relatively
little understory. The shola forest tree community composition is dominated by the
family Lauraceae (Ramesh 1989). Although bees are seen foraging on plants in sholas,

they are rarely, if ever, seen nesting in these patches.

Amidst the matrix of these main types of habitats are many patches of smaller native
habitat types such as the high elevation grasslands, tree savanna, riparian zones, and
others (figure 1.2). Besides the complex floral communities that it harbors, BRT
Wildlife Sanctuary is home to healthy populations of a great number of the charismatic

Indian megafauna, including tigers, leopards, elephants, langurs, macaques, the elk—like

sambar, spotted deer, barking deer, gaur, sloth bear, wild dogs, wild boar, giant

squirrels, and many others.

The human landscape in the area is as heterogeneous as its habitats, reflecting several
centuries of varied types of human use. The Soliga people are the traditional inhabitants
of the BR Hills, and although many have moved away to cities and towns nearby, about
5000 Soligas live within the boundaries of BRT and depend on the forest for fuelwood,
fodder, and non-timber forest products (Shankar et all998). Originally they were
shifting agriculturists whose land use practices involved clearing small patches of land
by setting ground fires, farming it for 8-10 years, and then leaving it fallow for 50-60
years (Rudrappa 1996). They actively managed the forest as well as the lands they
cultivated, most directly by rotational burning of small patches. They set fires in order
to control the thorny undergrowth which impeded movement through the forest, and to
control the build up of fuel on the forest floor such that highly destructive fires could not
occur. Rudrappa (1996) reports that their management of the forest through fire selected
for the growth of certain useful plants, and created grazing patches which attracted
species of herbivores that they hunted. The forests here, as perhaps most of Asia’s
forests, are products of centuries and generations of management by humans whose
absolute dependence on the products of their ecosystem required an intimate familiarity
and coexistence. Although major traditional land use practices have changed, Soli gas
living in BRT continue to depend on the wealth of the forest for food, medicine, and

spirituality.

10

As in much of Western Ghats, BR Hills underwent commercial logging by the British
colonialists, which then continued into post-colonial era. By 1987, all tree felling was
banned and the area was designated as the Biligiri Rangaswamy Temple Wildlife
Sanctuary, to be administered by the Kamataka State Forest Department. Most of the
people living within the boundaries were resettled to an area in the northern part of the
sanctuary. The Soligas continue to have usufruct rights to harvest non-timber forest
products (NTFP) such as fruits, lichen, medicinal plants and honey; however, they do
not have hunting or fishing rights, nor do they practice Shifting agriculture anymore, and
at present almost all families subsist at least in part on the cash economy that sales of

forest products and temporary employment brings them.

Among the most Significant NTFP’s that are collected for both personal use and for sale
is honey from Apis dorsata. Honey collection by Soligas is a major seasonal
undertaking, and a dramatic scene to witness. Collection from A. dorsata occurs at night
at heights of up to 40m or more; thus, it is a vocation practiced only by a few highly
skilled people, and honey collecting season is typically opened with a prayer ceremony
by the collectors. Although collection requires destruction of the comb, it does not
typically result in destruction of the colony; thus, it is not clear what, if any impact the
collection of honey has on the population of Apis dorsata. Honey collection is a practice
that has occurred sustainably for several thousands of years in Asia (Crane 1999);
however, present day changes in economic imperatives are altering the traditionally

sustainable harvesting practices of many forest products around the world. Studies are

11

ongoing in BRT as to whether the honey collection practices and harvest levels will

ensure the persistence of the resource in perpetuity.

Because of the presence in this area of a 500 year old Hindu temple which once
belonged to the Prince of Mysore, and which is now heavily visited by worshippers from
surrounding regions, the approximately 4 km2 area in which the indigenous people were
resettled is not technically part of the wildlife sanctuary (figure 1.2). Instead it is
designated as revenue land in which people can settle, purchase land, grow crops, etc.

In addition to the Soliga settlements in this area, there exists a hospital and school run by
a local NGO working with the Soligas called the Vivekananda Girijana Kalyana Kendra
(VGKK), the temple and its surrounding administrative bodies and visitor facilities,
forest department housing and outpost, a place for religious study called an ashram, a
few small public works offices which are or were in the past involved in government
development schemes with the local people, and a research field station operated by
Ashoka Trust for Research in Ecology and the Environment (ATREE) from which I

conducted this work.

In addition to Soliga settlements in the above described area, a few Soliga settlements
are scattered in some areas inside the actual sanctuary. Here, houses are typically made
of thatch, and the people cultivate small kitchen gardens that include bananas, tubers,
and gourds. In addition to these settlements, there are other human influences inside the
sanctuary boundaries. Much of the original range of evergreen forest was replaced by

coffee plantations by British colonialists in the early 20m century, and these still exist as

12

functioning plantations today. Tree plantations of Eucalyptus and Teak were planted in a
number of small isolated patches by the forest department, and although they are no
longer maintained, they persist as discrete types in the landscape and have not yet been

overgrown by native vegetation.

Amidst this varied biotic and human landscape, a study of honey bees has even greater
relevance. Our scientific knowledge of rock bees is somewhat scant, and human
knowledge of it is rapidly disappearing as the wisdom accumulated by indigenous Asian
cultures transforms to keep pace with the demands of the modern world. A forest such
as BRT that has been inhabited and managed by humans for millennia represents the
situation that is typical of forests throughout South Asia (Gadgil 1992). It is not a
pristine habitat, but one in which the forces of human alteration have acted and continue
to act at a pace much slower than the changes presently occurring around it. People
throughout Asia depend on A. dorsata directly, for the products they derive from bees,
and indirectly, through their pollination services to forests and croplands. Thus, I
believe that research which sheds light on aspects on the ecological role of Apis dorsata
and pollination in the Asian tropics could have potential widespread value in many areas
of Asia in which similar networks of dependency exist among people, honeybees, and

forests.

13

 

Figure 1.1. a. Apis dorsata nest b. Apis dorsata nest aggregation of over 80 colonies.

due mo 86 t2 co 8:3: 8%

55 82m 35% Mo 28:82 SEE 8—86 com 4x8 05 E with; 53» 203.28
on 8 “Sewage :03 032 5:2 2: Bob 8.55 ccowB oEom .wafl 63am;—

vcm .=o=02.m ..M.m .fiofiwm ”ham Co 93.4 SUE .axSaEaM .5385 8:33
Che 29:08 agammmfim 535 Co 92: 2389:: can «:2: Co as: .NA Eswfi

fl
.5. v o
328 3032009
1..
nuang ”A . “we—m om
sac—=38 09F i ‘ . . .

            

      

5:553 02,—.
3.28 :3._wco>o-_Eom
328 nEom ,,
528 559% E

as was
628 Woe—whom “
50525901
283 ootoo I

<

Z

820 E88?

  
     

 

seem
as we

 

 

 

a; 8qu eoamfim «

15

CHAPTER 2: DYNAMICS OF FLIGHT RANGE ACROSS SPATIAL AND

TEMPORAL SCALES

INTRODUCTION

Insect pollinators have recently become focal points in applied ecology for their role as
crucial links in the long-term integrity of ecosystems (Buchmann and Nabhan 1996;
Allen-Wardell et al. 1998; Kearns et al. 1998). Consequently, a number of research
efforts are underway to assess the effects on the pollination mutualism of habitat
changes such as fragmentation. These implicitly spatial problems are somewhat limited
by our lack of information regarding the foraging distance of pollinating insects. The
foraging ranges of insects are especially of interest since insects are presumed to be
limited in their ability to effectively cross-pollinate as the spacing between conspecific
plants increases. The consequences of this spatial limitation on insect foraging have
implications for fruit and seed set (Nason and Hamrick 1997; Nason et al. 1998; Steffan-
Dewenter and Tscharntke 1999), levels of inbreeding depression and seed viability (Hall
et al. 1996), population genetic structure (Murawski and Bawa 1994; Dayanandan et al.
1999), and plant community composition (Nason and Hamrick 1997), all via their

effects on pollinator foraging.

In contrast to the Neotropics where honey bees are introduced, Asian tropical forests
host from one to three sympatric and native species of Apis in any given area (Ruttner
1988). In Asia, where the bee fauna is less diverse than in the Neotropics, and where the
native fauna and flora have evolved alongside Apis, it has been hypothesized that the

highly eusocial bees have largely “pre-empted” the insect pollinator niche (Michener

l6

1979; Roubik 1990), in part due to the unrivaled efficiency with which they recruit
nestmates to resources over long distances. Thus, although habitat fragmentation has
negative effects on plant populations in some Neotropical plants (Aizen and Feinsinger
1994), such effects may be mitigated in Asia by the recruitment system and relatively
long foraging distances of honey bees. Consequently, these forests may be somewhat
“immune” to the effects of spatial disruption of the habitat as they affect pollinator
foraging. There have been findings in support of this idea from studies in the
Neotropics: studies of fruit set, seed set, and even genetic variation has shown that the
introduced European honey bee, Apis mellifera scutellata is in fact now assuming the
role of an dominant tree pollinator in fragmented or highly disturbed habitats, in some
cases where the native pollinator fauna has dwindled (Aizen and Feinsinger 1994; Dick
2001). This suggests that the genus Apis may indeed possess unique traits that allow it to
exploit resources in a way that is less limited by spatial configuration of habitat.
However, before being able make predictions about the impact of large scale habitat
alteration on loss of biodiversity and habitat regeneration, it is crucial to have baseline
data from a relatively unfragmented forest: the dynamics of flight range and foraging
area, and thus distances of potential pollination throughout the flowering season must be
documented. These parameters may vary in response to resource fluctuations, so
estimates made over an extended time period will allow for a distribution of flight

ranges reflective of the variation in natural systems.

Due to the difficulties of gathering such information, there are very few estimates of the

foraging ranges of insects, despite their crucial importance in all terrestrial ecosystems.

l7

Past estimates of pollinator flight range have relied on indirect methods to estimate the
upper bound of potential flight range (Janzen 1971; Roubik and Aluja 1983; Dramstad
1996; Saville et a1. 1997; Roubik 2000) due to the impossibility of tracking individuals

as they perform foraging flights over several kilometers (but see Osborne et al. 1999).

Honey bees (genus Apis) present a unique opportunity to investigate flight range due to
their communication system known as the “dance language” (Frisch 1967). Foragers
communicate to nestmates the distance and direction of a food source via means of the
“waggle dance” which occurs on the nest surface. Across the entire genus Apis, the
dance language works in the following way. A forager that finds a profitable source of
forage returns to the nest and informs her nestmates of the location of the patch by
running in a straight line on the vertical surface of the nest comb, while rapidly waggling
her abdomen back and forth. The angle relative to the upward direction at which she
runs corresponds to the horizontal angle between the solar azimuth, the patch of forage,
and the nest (figure 2.1). She waggles for some period of time, usually on the order of a
few seconds, then returns to her original position, repeats the run, and may continue to
repeat it several more times. The distance signal is encoded by the duration of the
waggle run such that the duration of each run translates into distance flown. The
relationship between waggle run duration and distance flown is referred to as a “dialect”
because, as with human linguistic dialects, it varies across geographically isolated
populations (Frisch 1967). In Apis, the distance dialect also varies across species
(Lindauer 1957; Dyer and Seeley 1991), and can be calibrated by researchers for any

given species or population. Thus, by observing dances one can use the bees’ own

18

communication signals to infer direction and distance flown without the need to track

individual bees as they fly.

This method has been used by past researchers examining foraging organization and
strategies of Apis mellifera in both native (Schneider 1989; Schneider and McNally
1992a) and non-native environments (Visscher and Seeley 1982; Waddington et al.
1994). Although there are estimates of the foraging range of Apis mellifera from the
above studies, other species of Asian Apis (Punchihewa et al. 1985; Dyer and Seeley
1991), some stingless bees (Roubik and Aluja 1983), and some bumblebees (Bombus
spp.) (Saville et al.1997; Osborne et al.1999), until now there have been no estimates for
any pollinating insect across various spatial and temporal scales, and none that
accounted for intercolonial variation by examining several colonies over the same

extended time period.

The giant honey bee, Apis dorsata, is ubiquitous throughout tropical Asia, and like all
other honey bees, possesses the ability to communicate via dance language. A. dorsata
nests hang in the open from tree limbs or rock cliffs, often in large aggregations
(Deodikar et al. 1977) (figure 1.1b). These aggregations occur at sites which are
recolonized by migrating colonies of A. dorsata year after year. Nests consist of a single
comb covered by a “curtain” of bees (Ruttner 1988) and waggle dances occur on the
surface of the curtain, visible to the naked eye or through a spotting scope (Dyer and

Seeley 1991). I observed the dances of several colonies in different nesting sites over the

19

course of the flowering seasons of 1997 and 1998 in the Biligiri Rangaswamy Temple

(BRT) Wildlife Sanctuary in Kamataka, India.

In this study, I used the dance language Signals to infer locations of foraging patches
throughout the entire flowering season for two consecutive years in order to examine
flight range distributions across a range of spatial and temporal scales. An estimate of
flight range and its variation will allow me to estimate the area of a given forest type
which is needed to support colonies in nesting aggregations of A. dorsata, and in turn
how much forest area a nesting aggregation provides pollinator services to. I expected
that different habitat types, nricrohabitat characteristics associated with the nesting site,
and colony differences to contribute to variation in flight range. Furthermore, I expected
to find that flight range contracts and expands in response to resource availability when

examined chronologically across the season as the flowering season progresses.

METHODS

DISTANCE DIALECT CURVE

The translation of circuit duration—distance to flight distance was made by obtaining a
“dialect curve” for the population using standard methods (Lindauer 1957; Frisch 1967;
Dyer and Seeley 1991). Data for dialect curve calibration were collected in Bangalore,
India (12.58 N, 77.35 E) during October 1998. We located an A. dorsata nest on a
partially constructed building with a fairly long stretch of open, flat land in front of it,
ideal for this type of data collection. The nest was on the second story of the building,

and thus was within the normal range for nesting height of A. dorsata. Since it was

20

constructed in a window space, it was possible to approach from inside the building in
order to “bait” the bees. On the first day of data collection and every morning
subsequently, we baited the nest with concentrated sugar solution scented with orange
essence. This technique of spraying the nest with the same solution used in the training

feeders induces the bees to search for that scent elsewhere.

We trained bees from the above colony to forage at a feeder with scented. sugar solution
by initially placing it close to the nest. Once a few bees had found it and recruitment of
nestmates was in progress, the feeder was progressively moved further away from the
building, and we marked recruits with acrylic paints as they fed. By markng each bee
with a unique color combination on the thorax and/or abdomen, we ensured that dance
measurements would only be taken from foragers seen at the feeder, and we also were

able to avoid resampling the same individuals when measuring their dances.

After marking several recruits, the feeder was moved to a distance of 100 m from the
nest and we began data collection. One person was stationed at the feeder, marking new
recruits and noting the identities of feeder foragers. One person was stationed near the
nest, measuring the dance angle and waggle circuit duration of foragers returning from
the feeding station as described below. We communicated with each other via two-way
radios to confirm identities of bees seen at the feeder and whose waggle dances were

therefore to be timed.

21

Observations were made by watching each marked dancer bee through a Swift 15-60x
spotting scope. To measure the distance signal in the dances, we timed the entire
waggle circuit duration, as opposed to only waggle run duration. By timing a minimum
of three continuous circuits, average circuit duration could be obtained while minimizing
the error associated with starting and stopping the stopwatch, thus improving precision
as well as accuracy in the distance estimate. Waggle circuit duration is often used as a
surrogate for waggle run duration because the two measures are highly correlated

(Frisch 1967).

After at least 20 different bees’ dances were measured, the feeder was again moved out
in gradual steps to the next 100 m increment, and the process was repeated. Over the
course of two weeks, we trained the bees to a distance of 1100 In, beyond which it was
impossible to entice them to the feeder. Sample sizes associated with the 1000 In and
1100 In training distances are less than 20, as it appears that fewer bees were motivated

to recruit to the sugar solution at distances so far from the nest.

I ran a least squares linear regression of circuit duration vs. feeder distance to describe
the waggle circuit duration-distance relationship. I used this regression equation to infer

distances flown by dancers whose dances were observed for forage mapping.

FORAGE MAP DATA COLLECTION

Forage map data were collected in BR Hills during the dry seasons of 1997 and 1998. In

1997, [observed the foraging activities of A. dorsata in both deciduous forest and

22

evergreen forest, the two habitat types in which they are known to occur predictably
every year in B R Hills. In each habitat, two nest aggregations of >20 colonies were
chosen on the basis of their accessibility. The nesting aggregations are a convenient
spatial unit in which to choose replicates because all colonies in an aggregation
experience the same surrounding environment. At each aggregation (or “site”) I chose
three or four representative colonies from which to collect data throughout the season.
The selection of colonies at a site was largely determined by the accessibility of a clear
line of sight perpendicular to the planar surface of the comb, and a ground surface level
enough to support the spotting scope at a distance of 10-30 meters from the nest. In

1998, for logistical reasons I restricted the study to three deciduous forest sites.

Forage maps were constructed by observations of the dance language of the bees,
measuring waggle angle and waggle circuit duration. On each focal colony, I randomly
chose dancer bees for observation. I watched each bee with one eye through the spotting
scope, and looked at a carpenter’s protractor (an instrument with a mechanical plumb
line) while holding it against the scope with the other eye. Binocular vision creates the
appearance of the bee dancing along the edge of the protractor, and thus, I was able to
measure the vertical angle of the waggle dance. Waggle duration was timed using the
methods described in the dialect calibration section of this chapter. Other necessary
pieces of information needed were those that affect the position of the solar azimuth: the
date, and exact time (HI-IMM) at which the observations were being made, longitude,
and latitude of the site. The latter two were taken using a Magellan Trailblazer GPS

unit.

23

Each site was visited for one full day per week. During each observation day, each
colony was observed for twelve nrinutes every hour starting at approximately 0600 or
with the first sign of dance activity after 0600. Observations continued until the time at
mid-day when flight and dance activity by the bees stopped, usually around 1200. We
then resumed dance observations from approximately 1600 until 1800, at which point
the bees were often still dancing, but light levels became too low to see them through the
spotting scope. During 1997, the Beduguli site was observed every week from 1600-
1800 on one day, and from 0600-1200 the following day. The end of the observation
season was when colonies started migrating and/or honey harvesting (and therefore
colony comb destruction) began. In 1998, observations at Sige Gudi site ended earlier in
the season than those at the other two sites because two of the three colonies under
observation absconded for unknown reasons in mid-season. Appendix 1 summarizes the
site characteristics, dates of observation, number of colonies and weekly sample sizes of

dances measured.

FORAGE MAP CONSTRUCTION

I constructed forage maps by first plugging the average circuit duration for each dance
observed into the regression obtained from the dialect calibration to obtain distance
flown by each dancer. Second I translated the dance angle into the bee’s flight angle
using a Microsoft Excel macro programmed for calculation of the sun’s azimuth. Based
on site latitude and longitude, date, and time of data collection, the program calculates

the position of the azimuth with the following equation:

24

Azimuth = arccos [ (sin D- cos Z sin L) / sin Z cos L]
D = solar declination
L = latitude
Z = zenith distance = arccos ( sin D sin L + cos L cos D cos H)
H = hour angle, i.e., time in hours relative to local noon * I 7°/hr’
Using the position of azimuth, the program translates dance angle into the bee’s actual

flight direction relative to geographic north.

The flight direction and distance for each dance were converted into (x,y) coordinates
using the following set of formulas:

Flight angle radians = Radians ( 90-compass flight directionz).

x = cos radians * flight distance in meters

y = sin radians "' flight distance in meters.
The (x,y) coordinates were plotted relative to the origin (0,0) which indicates the

honeybee nest site aggregation.

 

' 17 degrees per hour is the rate of the sun’s movement at the latitude of the study site.

In a compass angle, 0° occurs in the forward (or upward) position, but in a mathematical angle it occurs
in the right hand horizontal position. Thus, the compass angle first had to be transformed into a
mathematical angle using the formula (90-flight angle).

25

STATISTICAL ANALYSIS

In order to determine whether flight range varied between years or according to time in
the flowering season, and whether various levels of spatial differences (habitats, sites
within a habitat, colonies within a site) accounted for variation in flight range, flight
distances were compared by performing a repeated measures nested analysis of variance
(ANOVA). Using SAS 8.0 software to run the general linear models procedure, model
factors were defined as year, site (year), colony (site, year), week (year), week* site
(year) and week*colony (site, year). I designated year, site, colony (site, year), week*
site (year) and week*colony (site, year) as random variables. Initially, 1997 data were
analyzed separately with habitat type as the highest level in the nested hierarchy to
determine whether colonies nesting in evergreen vs. deciduous habitats varied in flight
range. Since there was no significant difference due to habitats (p = .8041) (figure 2.10
inset), the 1997 data were pooled with the data from 1998 in which I studied nests only
in deciduous habitat. Sites BK and KA were used in both years, but they were
considered in the ANOVA to be separate, hence the site term is nested within year.
Flight range was non-normally distributed, therefore all ANOVA’s were performed on
natural log transformed values. The data were still non-normally distributed, but
ANOVA with large sample sizes is considered to be robust to non-normality (Sokal and

Rohlf 1981).

To distinguish whether colonies forage in a manner that results in segregation of
foraging locations from week to week, a two-way multivariate analysis of variance

(MANOVA) was performed for each site data set on the (x, y) coordinates of each

26

foraging location as they varied with colony, week, and colony*week. Using JMP 3.2
statistical software (SAS Institute 1999), the response design used was the identity
matrix, utilized in order to keep the integrity of each variable separate instead of
summarizing the two dependent variables. I report F ratios and p-values from Pillai’s
Trace test since this is the test considered to be the most robust to unequal sample sizes

(T abachnick and Fidel] 1983).
RESULTS

DISTANCE DIALECT CURVE

The calibration curve for the distance dialect reaches to 1100 m from the nest, with a
clear linear fit of circuit duration vs. distance (y = 1.39 + 0.0030 x, 1'2 = .86) (figure 1.2).
I used this equation in forage map construction to calculate the distance flown by
dancers observed at BRT. For dances with circuit durations greater than that observed
for the longest training distance (1100 m), I assumed that the slope of the dialect curve
could be extrapolated to greater flight distances. This assumption has been supported

for A. mellifera (Frisch 1967).

FORAGE MAPS

Forage maps for all sites on a weekly basis are given in figures 2.3-2.9. An examination
of the forage maps alone suggests certain patterns: flight range is largely concentrated
within a few kilometers from the nest, but there are some occasional longer distance
flights; distance does not appear to contract and expand in a linear chronological

progression; any given colony does not restrict its entire season’s flight activity to a

27

certain distance or a certain direction, but instead seems to shift its use of the landscape
on a weekly basis; and colonies appear to be using patches in roughly different places.

These patterns are examined statistically in greater detail below.

FLIGHT RANGE DISTRIBUTION AND FORAGING AREA

Figure 2.10a inset illustrates the similarity of flight range in the two habitat types
examined in 1997. Figures 2.10 and 2.11 illustrate the overall flight ranges of A.
dorsata, in 1997 and 1998 separately and combined, respectively. Results of the two
years combined indicate that 90% of foraging locations occur within 2278.8 In from the
nest (figure 2.11), corresponding to a circular area of 16.3 kmz. The results also indicate,
however, that bees occasionally advertised a location that was close to ten kilometers
away (figure 2.11); thus the maximum circular foraging area covered by any colony was

289.2 kmz. Images in this dissertation are presented in color.

ANALYSIS or VARIANCE

ANOVA results (table 2.1) reveal that flight range did not vary temporally, either
between years, nor between weeks within a year. A significant portion of the variation
was, however, explained by the interaction between week and site, and by the interaction
between week and colony. The latter suggests that there may be some separation of
foraging locations among colonies in the same aggregations, a point which is further
investigated below by a MANOVA on the coordinates of foraging locations. Site and

colony main effects were significant, but confounded by their significant interaction

28

terms with week. It appears then that the above foraging range distribution does not vary
across different temporal scales or according to habitat characteristics, and is
representative of a typical Apis dorsata colony’s foraging movements over the course of

an entire flowering season.

MULTIVARIATE ANALYSIS or VARIANCE

MANOVA results for each site confirmed that colonies are not foraging in the same
patches (table 2.2). In only one site (BG 97) were there significant colony main effects;
that is, in six of the seven sites, colonies did not “specialize” on a particular segment of
the landscape throughout the season. In most sites there was a significant effect of week
on location of foraging patch, that is, the flowering patches used shifted significantly
between weeks. In all sites, there was a significant effect of colony“ week interaction,
that is, in any given week, colonies foraged in different locations relative to each other,
confirming the ANOVA result that colonies within sites varied relative to each other in

the distances they foraged among weeks.

ISCUSSI N

The forage maps in figures 2.2-2.9 illustrate the areas of terrain that are covered over the
course of a flowering season by an aggregation of Apis dorsata. Colonies send foragers
out over a maximum area of close to 300 kmz, a figure that rivals and even exceeds
home range estimates made in the same or similar habitats for several large mammals

(Batra et a], in prep).

29

Despite the enormous maximal range, however, most flights were concentrated within
two kilometers from the nest, as was also found in Apis mellifera in both native and non-
native habitats (Visscher and Seeley 1982; Schneider 1989). Furthermore, the maximum
A. dorsata flight distance was approximately the same as that of A. mellifera, but the
90% quantile for A. mellifera was much greater (6 — 6.5 km), indicating that the flight
range distribution of A. dorsata shown in fig 2.11 is less evenly distributed, that is, far
more skewed than that of A. mellifera. A similarly skewed flight range distribution was
found by Dyer and Seeley (1991) for a single colony of A. dorsata in Thailand,
suggesting that my conclusions may indeed be applicable throughout tropical Asia

where A. dorsata occurs.

The results of the statistical analyses suggest a picture of flight range that is fairly stable
when viewed over large Spatial and temporal scales, but one that undergoes constant
dynamic shifts at smaller spatio-tempora] scales. As evidenced by the AN OVA, neither
year nor week, the two temporal factors, had significant effects on distance flown in
foraging flights by A. dorsata. Flight range data for two consecutive years did not show
any differences in overall distribution, suggesting that there may be little inter-annual
variation, and that A. dorsata may be constrained in its flight distance by factors that
have little to do with large scale environmental fluctuations that occur between years.
Ideally, such findings would be confirmed over several more years, and thus span a

greater range of climatic or other conditions that may vary among years.

30

Flight distance also did not vary across shorter time periods. During any week in the
season, the entire distribution of flight range was exhibited when all colonies were
examined together; that is, flight range does not undergo predictable expansions and
contractions which respond directly to progression of the flowering season. Thus, the
hypothesis that flight range would respond uniformly across all colonies to changes in
floral resource availability is not validated. It is possible that although flowering density
increased and then decreased over the season, colonies’ energetic needs changed in
tandem, such that the simple equation of being able to fly shorter distances due to an
abundance of resources could not compensate for the increase in numbers of brood and
adults as reproductive swarming time approached. Instead, an interaction between
changing resources and changing needs resulted in no net change in flight distance over

the weeks of the study.

Results of the spatial influences on flight range reveal that on the large scale level of
“habitat type” flight range does not vary, as evidenced by the AN OVA on 1997 data.
Similarly, neither site differences nor colony differences explain variation in flight
range. Instead, the significant week*site and week*colony interaction terms suggest that
flight range is far more dependent on individual site and colony conditions as they
change from week to week. The weekly site differences most likely reflect local
availability of flowering patches influenced by rrricrohabitat variation, and are highly

site-specific. I address this issue further in chapter 4.

31

The statistically significant week*colony term suggests that foraging range expansions
and contractions occur on a more colony specific basis over short temporal scales. A
similar finding was reported by Waddington et a] (1994), who found the colony* day
interaction to be highly significant in a study of two A. mellifera colonies placed side by
side and studied in two habitats for four days each. Waddington et a] suggested that
different nutritional demands might result in different foraging patches being preferred.
Flight range may be influenced not only by a colony’s discovery of resources, but also
by “colony state,” that is, its availability of workers, its need for incoming pollen and/or
nectar as demanded by the adult to brood ratio and potential for swarm production
(Schneider and McNally 1992a, 1992b), and factors Such as disease, parasite load, and
predation losses by wasps or birds—all dynamic influences that may interact and vary

between colonies and between weeks.

Estimates made by Visscher and Seeley (1982) of flight range for Apis mellifera in a
temperate forest showed a flight range distribution in which 90% of all flights occurred
within a 6600 meter distance, clearly exceeding the 90% quantile of 2280 m that I
observed for A. dorsata. Their estimate, however, was based on intensive sampling of
only one colony spread out over a three month period, but it did not cover the early and
late parts of the season, nor did it not account for intercolonial or interannual variation.
If the statistical results that I report here for Apis dorsata are applicable to A. mellifera
as well, Visscher and Seeley’s estimate may nonetheless be an accurate indicator of the
region’s A. mellifera population on the whole, since I found no interannual,

intercolonial, or weekly differences. However, in a study of A. mellifera scutellata in

32

southern Africa by Schneider and McN ally (1992a, b), there did appear to be differences
based on seasonal availability of food, with flight range showing strong (although not
statistically analyzed) increases as the dearth season approached. It remains unclear as
to what extent the tropical race of A. mellifera studied by Schneider and McNally and
the European race studied by Visscher and Seeley are comparable, nor do we know to
what extent the results I report for A. dorsata may be generalized to other species of

honey bees.

A further implication of the significant week* colony interaction is that during any given
observation period, colonies were not foraging at the same distances from the nest site.
The suggestion that distances and hence foraging patch locations were different is
confirmed by the consistent results of MANOVAS performed on the (x,y) coordinates of
all inferred foraging locations. The coordinates are a combination of the distance and
direction of a foraging patch. 1f colonies nesting at the same site are utilizing distinct
patches, the MANOVA results should show a significant effect of week*colony on (x,y)
coordinate combinations. This is indeed the case in all seven sites, confirming the
results of the ANOVA which examined distance alone. I observed only a small
proportion of the colonies nesting in each site, so it is possible that the particular
colonies observed incidentally happened to be visiting locations widely separated from
each other. However, this “sampling artifact” scenario is unlikely to be an adequate
explanation for my results given that all seven sites showed the same result quite
strongly. Additionally, since there was a significant main effect of colony in only one of

the seven sites, there is no evidence that colonies generally separate their foraging

33

locations by “specializing” on a certain segment of the surrounding environment and

staying within it throughout the season.

A similar result implying spatial separation of foraging effort was found by Waddington,
et a]. (1994) in a study of Apis mellifera in suburban habitats in which two colonies were
placed side by side and foraging locations mapped. Although they examined only flight
distance, the divergence in the colonies’ distance distributions was clear enough to lead
them to conclude that colonies were not utilizing the same patches simultaneously. They
point to the possibility of differences in initial discovery that are then reinforced by
recruitment and eventual segregation between colonies. Precedence at a patch, they
point out, also may result in decline of the patch’s resources, causing foragers from other
colonies to find it less attractive. Waddington et al. also consider that colonies may

choose different patches based on different nutritional needs.

Another possible mechanism for patch segregation may involve interactions at the patch
itself using chemical cues. Colony specific odors at floral resources, in addition to
attracting nestmates looking for the patch (Winston 1987), might additionally serve as a
deterrent to scouts from other colonies. Once a newcomer “recognizes” that the patch is
already being exploited and thus depleted, the patch may become less attractive as a
potential food source for their own colony. The use of pheromones for avoidance
competition is likely to be less costly than interference competition because it would
prevent the cost to both colonies of losing foragers and expending energy involved in

fighting among their workers. Such a mechanism may be important in allowing A.

34

dorsata colonies to minimize competition at resources while coexisting in aggregations

that sometimes exceed 100 nests, all of which forage over the same expanse of forest.

Apis dorsata’s ability to forage over enormous areas of terrain, an overall flight range
distribution that typically stays within a few kilometers of the nest, and short term
micro-segregation of habitat among neighboring colonies, when taken together, suggest
several implications for the pollination of plants in the vicinity of nest aggregations.
Areas surrounding the nest sites are well covered by foraging pollinators coming from
one colony or another, although the occasional long distance flight may indeed be
important in long distance pollen flow and maintenance of genetic diversity in plant
populations. In addition to providing a glimpse at how the ubiquitous giant honey bee
interacts with the landscape, foraging area estimates provide a spatial baseline from
which to formulate hypotheses regarding pollen mediated gene flow and population
genetic structure of plants pollinated by A. dorsata, effects of habitat fragmentation on
genetic, population, and community level properties, and finally, the evolution of plant-

pollinator relationships in Asian forests.

35

Table 2.1. Analysis of variance results. The dependent variable was normalized using a
natural log transformation. y = 1n flight distance. Random variables are year, site (year),
colony (site, year), week* site (year) and week*colony (site, year).

 

 

 

 

 

 

 

 

 

 

 

 

 

SOURCE OF VARIATION DF MS F p

year 1 6.27 .51 .5068

site (year) 5 14.44 4.17 .0056 **
colony (site, year) 18 2.58 2.87 .0004 ***
week (year) 12 2.83 1.25 .2975
week*site (year) 28 2.55 2.56 .0004 ***
week*colony (site, year) 84 1.04 2.06 .0004 ***
residual 2372

 

36

 

Table 2.2. MANOVA results. Dependent variables are the x and y coordinates of
foraging location. For each site, the Pillai’s Trace F ratio is reported with p-Value in

 

 

 

 

parentheses.
BK 97 KA 97 DS 97 BC 97 BK 98 KA 98 SG 98
colony 2.1350 .3685 1.0011 4.4560 2.1392 .2103 .9904
(.0747) (.8312) (.3686) (.0126) (.0745) (.8105) (.4129)
week 5. 8435 8.3514 2.5784 16.8849 9.7952 4.8423 1.5970
(<.0001) (<.0001) (.0365) (<.0001) (<.0001) (<.0001) (.1748)
colony* 2.3771 2.5730 2.1594 1.8464 3.3532 2.4285 3.9738
wk (<.0001) (<.0001) (.0036) (.0005) (<.0001) (.0010) (<.0001)

 

 

 

 

 

 

 

 

 

37

 

:2 2mm“? 05 we .5583 05 .3 6250:... w_ 8555

.ana wfitfim “Es :oEmom fifiwto 5:8 mEEBo: $5306 .3 .88: .838 :3 2mm?» £5 83:2

cam do: 2: 98 63:8 we :83 05 .5585“ :38 2: 5253 2mg figure: 05 8 megmmobg 20:3
EmE 2:: 80:3 a £023 a 558% 985:: 05 9 03:22 2mg :89? 23. .omaamg— 855 Ad oSwE

 

   

 

 

 

 

 

 

 

38

.8323» «Ev. mo 80:3 2mm?» 05 :8 2:3 Boo—Se 8ng .N.N Eswm

c5 8:35 BE

Geo ocv can o
_ . . . .

P b Li D F P

co- cco_ cow
. . .

r F r

 

(39s) uorrernp rtnorrg

 

 

 

 

39

£58805 :5— : “a Coos: 2a 3.86 .moEoBo E8056 ofiofifi flonfiam
.538 wctoaoc 52 30:38:: 85: 09:8 86 A3 Ma seaxa§>amam .mN Bswfi

w :83 e :83 z 0 EB

   

w x3?

 

N x33 fl x003

 

40

v 0.003

 

9:050:05
Ex fl 3 @002: 0.5 00.85 .m0E200 E2830 80065 0.898% .5008
@530: 3a Sop—mach: 30:. 0&88 0% A3 we znwseom .vN 0.5mm

w 0.003 Z 0 x00?

 

m 0.33

 
 

m 0.003

 

L 0.83

 

41

w 0.003

 

£580.85 Ex _ an @0003 20 00.86 02:28 Bobtfi 03065 £353
£8000 wct030c 33 Seaway—5 308 0w88 0% A3 we $380000qu .m.~ 0.5me

u

 

m 0.003 N 0.003

 

m 0.003

 

_ 0.83

 

42

w 2003

 

05080.55 :5. 2 :0 0000—: 0:0 00.020 02:28 0:22:20 0.02:5
238% :00000 w:t030c 33 Seaway—5 0:2: 0w2£ 020 A3 <5 580M 6d Eswfi

Z-

5 0.83 0 0.83

 

2 0.003

 

43

08080.82 :2 2 :0 @002: 0:0 02020 02:28 0:002:20 20202 238%
.8080 w:t0>»oc wag 523585 09:: 0228 020 30 Me sugggmam fin 0.53m

 
 

m 0.83 N 0.83 : v.83

 

°§2§§§§§ HE

160 Fm Ohm (m)

   

T1111

°§§§§§§§ é.§§§§é§§§§§§

Flight Distance (m)

 

°°§§§ §§§§§§§§§§§§ 333233

Flight Distance (m)

Figure 2.10. a. 1997 flight range distribution combined over two habitat
types and all colonies. Inset shows flight range in evergreen (upper) and
deciduous (lower) habitats.b. 1998 flight range distribution

47

Frequency

50% = 1189.3 m

    
  

75% = 1713.3 m

90% = 2278.8 m

 

Max: 9594.6 m

0 0 000000 0 0000
§§§8§888888§8§888§§§
v-v-NNMMVVW'OwDI‘NQD 09

Flight Distance (m)

Figure 2.11. Flight range distribution 1997 and 1998 combined. 50th,
75th, and 90th quantiles are indicated.

48

CHAPTER 3: POLLEN DIET COMPOSITION AND FOOD PLANT PREFERENCES

INTRODUCTION

Apis dorsata, like other honey bees, is assumed to be a generalist forager. Its large body
size, the need to sustain large, perennial colonies, and its ability to occupy a range of
different habitat types over a broad geographic range all imply an ability to consume
nectar and pollen from a wide diversity of food plant species. Understanding the patterns
of food plant utilization by this important pollinator is critical for evaluating its role in
the ecology of Asian forest communities. Recent studies in non-agricultural habitats
have generally supported the assumption that A. dorsata exploits a wide diversity of
plant species (Kiew 1997; Devy 1998; Momose et al. 1998), but it has proven difficult to
obtain a comprehensive picture of the pattern of exploitation of different plant species
over time in a given habitat. In this chapter, I address this question through analyses of
pollens found in the feces deposited in the vicinity of nesting aggregations in the BR

Hills

Even if a species can be characterized as a generalist, foragers in a single population
may not forage on all plants available in a given area, and may not forage equally across
all the plants that it does use. Such apparent “preferences” may emerge as a by-product
of the sequential exploitation of plants according to their phenology and their flowering
density at any given time in the season (Kiew 1997). However, true preferences may
also exist depending on the economic trade-off experienced by the forager between the

nutritional content of various food plants and the energy expended on searching and

49

foraging for them. Such trade-offs are known to be important in A. mellifera colony
level foraging strategies (Visscher and Seeley 1982; Seeley et a1. 1991). There have
been many studies which have experimentally attempted to discern the decision making
criteria and sensory biases of bees for particular flower colors (Daumer 1958; Chittka et
al. 1993; Giurfa et al. 1995), shapes (Frisch 1914; Free 1970; Wehner 1971), floral
display arrays (Pyke 1978; Pleasants and Zimmerman 1979; Zimmerman 1981), scents
(Frisch 1919), energetic trade-offs (Heinrich 1979), and so on. It is an extremely
complex task however, to determine how those factors interact in a natural setting, with
all of its complexity, to result in biased resource exploitation. In fact, simply discerning
if a preference exists has scarcely been done in a forested setting even without
considering the multiple mechanisms which act to produce it. A dietary “preference” in
nature can only be discerned if a food plant is used disproportionately more than ‘
expected when compared to its availability. Here I will explore this question as it relates
to A. dorsata pollen foraging in BRT, and its possible implications for the forest

community.

The importance of knowing the components of an animal’s diet depends on the role that
such consumption patterns play in structuring an ecosystem. Bees are by far the most
important of all insect taxa involved in the pollination mutualism between insects and
flowering plants (Bawa 1990), and the process of pollen transfer among conspecifrc
plants is dependent on bees’ consumption of floral resources. The diet of bees therefore
provides a partial window into knowing which plants may depend on them for

pollination, and also provides a list of the plant species which maintain the bee

50

08080.88 82 2 8 2.002: 0.8 002820 02:28 800022220 202208
288% 800000 82.20322 :02 Sonwsobz 2208 09802 220 223 <v2v 28832 .m.~ 0.8mm

Z

 

m 02003 N 02003 2 02003

 

 

 

45

 

0
500
1000
1500
2000
2500

§§§§§§§§§

Flight Distance (m)

7500
8000
8500
9000
9500
10000

 

°§§§§§§

§§§§§§§§§§§§§§

Flight Distance (m)
Figure 2.10. a. 1997 flight range distribution combined over two habitat

types and all colonies. Inset shows flight range in evergreen (upper) and
deciduous (lower) habitats.b. 1998 flight range distribution

47

50% = 1189.3 m

  
     
     

75% = 1713.3 m

01

Frequency
N
O

90% = 2278.8 m

24.
01

Max: 9594.6 m

l

.5
O

0 00000000800000 0
00000000 00080 0
00800000000“) "3800

PFNN MVVIDIOQCDBNQQ 030
‘_

Flight Distance (m)

Figure 2.11. Flight range distribution 1997 and 1998 combined. 50th,
75th, and 90th quantiles are indicated.

48

CHAPTER 3: POLLEN DIET COMPOSITION AND FOOD PLANT PREFERENCES

INTRODUCTION

Apis dorsata, like other honey bees, is assumed to be a generalist forager. Its large body
size, the need to sustain large, perennial colonies, and its ability to occupy a range of
different habitat types over a broad geographic range all imply an ability to consume
nectar and pollen from a wide diversity of food plant species. Understanding the patterns
of food plant utilization by this important pollinator is critical for evaluating its role in
the ecology of Asian forest communities. Recent studies in non-agricultural habitats
have generally supported the assumption that A. dorsata exploits a wide diversity of
plant species (Kiew 1997; Devy 1998; Momose et al. 1998), but it has proven difficult to
obtain a comprehensive picture of the pattern of exploitation of different plant species
over time in a given habitat. In this chapter, I address this question through analyses of
pollens found in the feces deposited in the vicinity of nesting aggregations in the BR

Hills

Even if a species can be characterized as a generalist, foragers in a single population
may not forage on all plants available in a given area, and may not forage equally across
all the plants that it does use. Such apparent “preferences” may emerge as a by-product
of the sequential exploitation of plants according to their phenology and their flowering
density at any given time in the season (Kiew 1997). However, true preferences may
also exist depending on the economic trade-off experienced by the forager between the

nutritional content of various food plants and the energy expended on searching and

49

foraging for them. Such trade-offs are known to be important in A. mellifera colony
level foraging strategies (Visscher and Seeley 1982; Seeley et al. 1991). There have
been many studies which have experimentally attempted to discern the decision making
criteria and sensory biases of bees for particular flower colors (Daumer 1958; Chittka et
al. 1993; Giurfa et al. 1995), shapes (Frisch 1914; Free 1970; Wehner 1971), floral
display arrays (Pyke 1978; Pleasants and Zimmerman 1979; Zimmerman 1981), scents
(Frisch 1919), energetic trade-offs (Heinrich 1979), and so on. It is an extremely
complex task however, to determine how those factors interact in a natural setting, with
all of its complexity, to result in biased resource exploitation. In fact, simply discerning
if a preference exists has scarcely been done in a forested setting even without
considering the multiple mechanisms which act to produce it. A dietary “preference” in
nature can only be discerned if a food plant is used disproportionately more than ‘
expected when compared to its availability. Here I will explore this question as it relates
to A. dorsata pollen foraging in BRT, and its possible implications for the forest

community.

The importance of knowing the components of an animal’s diet depends on the role that
such consumption patterns play in structuring an ecosystem. Bees are by far the most
important of all insect taxa involved in the pollination mutualism between insects and
flowering plants (Bawa 1990), and the process of pollen transfer among conspecifrc
plants is dependent on bees’ consumption of floral resources. The diet of bees therefore
provides a partial window into knowing which plants may depend on them for

pollination, and also provides a list of the plant species which maintain the bee

50

population. Furthermore, indirect linkages between plant species which all rely on the
same generalist pollinator at different times of the season may be crucial in maintaining
the ecological integrity of the plant community as a whole. For example, if true
preferences for food plants exist, they may provide necessary resources to a population
of pollinators which also provides pollination services to several other “less preferred”
plant species. Such a preferred plant species would be considered a “keystone” resource
in that its ecological importance in the overall community is disproportionately large
relative to its abundance (Power et al. 1996). Characterizing the diet of A. dorsata, a
ubiquitous generalist pollinator which dominates the bee fauna in Asian tropical forests,
will help to identify elements of the plant community upon which regeneration and

structure of Asian tropical forests may rest.

Visitation and foraging by bees on a flower does not necessarily guarantee pollen
transfer, seed set or actual plant reproductive success (W aser 1983). However,
compiling lists of plant taxa visited is a necessary first step in identifying the possible
species which may depend on A. dorsata for pollination. Indeed, such lists are
especially useful in tropical forests where plant species richness can be staggeringly
high. There is a long tradition of identifying of a forager’s food plants based solely on
the suites of floral traits known as “pollination syndromes.” However, several studies
have shown us that floral morphology does not tell the entire story of the plant-pollinator
relationship, and they may in fact leave out important plant species (W aser et al. 1996;

Johnson and Steiner 2000). Thus, a comprehensive examination of the entire diet of a

51

putative pollinator is the only way to bypass these assumptions and begin to get an

accurate picture of how plants interact with their pollinators.

Apis dorsata’s utilization of food plants outside agricultural areas is known for two
forest areas through observations of visitation to flowers. In Lambir Hills, Indonesia,
researchers have found that giant honey bees forage on, and possibly pollinate about
30% of the species of canopy trees observed during mast flowering periods (S. Sakai,
pers. comm). In a wet evergreen forest in the Western Ghats of India, Devy (1998)
found that A. dorsata visited about 20% of the tree flora, and actually provided
pollination services to around 15%. These studies were conducted by observation of
visitation to flowers of selected tree species. Other researchers (e.g., Jhansi et al. 1991;
Kiew 1997) have examined the honey and pollen stores from honey bee combs after the
combs were cut for honey harvest, but these late season analyses may exclude those taxa
whose pollen was collected and fully consumed early in the flowering season, and thus
may not give an accurate picture of relative proportions of the food types collected by
the bees. Virtually all bees rely on both pollen and nectar from flowers, and the process
of gathering either may result in fertilization of the flower. While it would be ideal to
know which plants comprise the entire set of food plants, pollen provides protein needed
for colony growth and reproduction and is a far more tractable resource to investigate

than nectar.

Apis dorsata presents a unique opportunity to study the pollen diet of a population of
bees in an unbiased and comprehensive manner. Giant honey bees’ perform “cleansing

flights,” in which most or all individuals fly off the comb and defecate en masse only a

52

few meters from the nest, a behavior that may be adaptive for efficiently dissipating heat
from the colony (Mardan and Kevan 1989). This behavior, further amplified by giant
honey bees’ large colony sizes and nest aggregations, allows for investigation of pollen
foraging activities because the feces can be collected very efficiently from the forest
floor and consist primarily of the exine shells of consumed pollen grains. These outer
shells do not break down in the digestive tract of honey bees, and their shape, size, and
microscopic sculpturing are diagnostic characters which can be traced to family, genus,
and sometimes species of the plant from which they came (Faegri and Iverson 1964).
Analyzing the diet by analyzing the feces relies on information taken directly from the
bees, and not on the subjective observer-biased process of choosing species of plants at
which to quantify bee visitation, or on the overgeneralizations of the pollination
syndrome concept. In this chapter I report the first detailed characterization of the
pollen diet of A. dorsata from a forest habitat, and provide a picture of weekly changes

in the composition of the diet throughout the flowering season.

After characterizing the diet contents, the only way an actual preference can be
discerned is by comparing composition of the diet to the relative availability of those
resources in the environment. In order to get good estimates of relative abundance of
the tree species in BRT, I rely on data from collaborators at Ashoka Trust for Research
in Ecology and Environment (ATREE) which characterizes the tree community
composition in the major forest types of BRT. They uniformly sampled trees from a
total of 125 plots, totaling five hectares overall. The community structure revealed by

these plots overall closely matches the structure of other sampling plots in BRT whose

53

community was characterized for different studies (e. g., Murali et al. 1996), thus appears
to capture the major elements of the tree community quite well. Taking relative
abundance of a floral resource as the null expectation, I examine here, using data from
1997 and 1998, whether A. dorsata collects pollen types in proportion to their
availability or whether they exhibit preferences by foraging disproportionately more on
some plant species and less on others. The availability of a floral resource depends not
only on the relative abundance of that species in the forest, but also on its flowering
status at any given time in the flowering season. Weekly flowering phenology data
taken during the 1998 season, combined with the above data on relative abundance of
tree species in one forest type, provides for an even more accurate assessment of the
relative availability of food plants utilized by A. dorsata. I incorporate temporal
variation in floral resources in asking the question of whether A. dorsata utilizes trees in

a manner that suggests foraging preferences.

METHODS

FECAL SAMPLING

During the flowering seasons of 1997 and 1998, I collected samples of honey bee feces
by taking advantage of their mass defecation flights, which occur within a few meters
from the nest (Mardan and Kevan 1989). Approximately once a week, I laid out ten
sheets of plastic or newspaper underneath a nesting aggregation. Appendix II lists the
dates of sampling at each site, and the corresponding week numbers with which I refer
to them. The locations of the sampling sheets were most often arbitrarily chosen, but in

some areas the terrain excluded certain spots (for example, many of the feces under the

54

cliff colonies actually landed on the cliffside itself.) The sheets remained in place for one
week and on occasion more than one week, and then were changed. They accumulated
yellow spots of fecal matter which I scraped from the paper and later processed for light

microscopy as described below.

The samples may be considered random samples of the diet of all colonies in that
particular nest aggregation, but the method does not guarantee equal representation of
the feces of all colonies. This sampling method does not standardize the number of
grains that landed on each sheet of plastic or paper; similarly, it is impossible to know
from how many colonies the feces had come, as the defecation may occur up to 20
meters from the nest, and sometimes even farther (pers. obs.) Furthermore, in some
weeks the sampling sheets may have accumulated more feces simply due to more
defecation by the colonies at that aggregation, an activity that may be correlated with
ambient temperature (Mardan and Kevan 1989). Some locations in some weeks were
not sampled due to removal or destruction of the plastic or newspaper sheets by
passersby, rain, or wildlife. In spite of these difficulties, there is no reason to suspect
that the proportional representation of specific pollens should be skewed towards one

plant taxon or another.

VEGETATION SAMPLING PLOTS

Data on the abundance and distribution of the major tree species were made available to
me through collaboration with ATREE. The data were collected by ATREE researchers

by first subdividing the entire BRT Wildlife Sanctuary into 125 2 x 2 km grid squares.

55

All trees above 10 cm diameter at breast height (DBH), i.e., 1.3 m above ground, which
occurred along 80m x 5m line transects running through the center of each grid square
were identified and DBH measured. Sampling was spread uniformly across BRT. Thus,
the data provide a good measure of the community composition of the major vegetation

types for all but the rarest species of trees.

FLOWERING PHENOLOGY

In 1998 from 12 March - 7 May I recorded weekly (sometimes biweekly) assessments of
flowering phenology of the major species of deciduous forest trees using a transect of
previously identified and tagged adult trees. Appendix 11 gives the sampling dates and
the week numbers to which they correspond, and Appendix 111 lists tree species on the
transect. The flowering status of each tree was defined according to the following
categories: “bud” when only flower buds were present on the tree, “early flowering”
when the number of buds exceeded the number of open flowers, “peak flowering” when
the number of open flowers exceeded the number of buds or old flowers, “late
flowering” when the number of old flowers and/or fruits exceeded newly opened
flowers, and “fruit” when only fruits but no flowers were present. The categories were
given numerical values as relative quantitative measures of pollen availability, or
standing crop. The relative numerical values assigned to different floral stages are 1 for
bud, 2 for early flowering, 3 for peak flowering, 2 for late flowering, and 0 for fruit.
The bud stage still offers visual and! or olfactory cues for bees to forage on the plant,
and mature buds at the time of measurement may well have opened later that day or in

the evening. For these reasons, trees in this stage were not given a score of “0”. The

56

values assigned to each flowering category are relative measures only, and do not
account for any differences that exist between numbers of flowers per tree or volume of

pollen produced per flower.

PALYNOLOGY

The fecal samples were analyzed for pollen types in the palynology laboratory of the
French Institute of Pondicherry in Pondicherry, India. This lab is well equipped with
microscopes and wet lab facilities, and with an extensive reference collection of Western
Ghats pollens. I prepared the samples by using the standard acetolysis technique, which
extracts the cell contents, leaving the hollow pollen exine intact (Keams and Inouye
1993). Contents of each sample were first thoroughly crushed through a wide mesh
sieve to remove all debris. They were then washed with glacial acetic acid in order to
remove all water from the sample, and centrifuged for approximately three minutes at
2000 revolutions per minute (rpm). A 9:1 mixture of acetic anhydride: concentrated
sulfuric acid was prepared by slowly adding the sulfuric acid to the acetic anhydride.
Approximately five ml of this solution was added to the sample test tubes, and the test
tubes were put in a hot water bath and maintained at a temperature of 85 to 90 C° for
five to ten minutes. Samples were stirred during this time, taking care to keep stirring
rods separate so as not to mix sample contents. The tubes were removed from the water
bath and the reaction stopped by adding glacial acetic acid to the test tubes. After
centrifuging again for 3-5 minutes, and removing the liquid, samples were washed with
distilled water, centrifuged, and drained 3 to 5 times to remove all traces of acid in the
test tubes. After determining with pH paper that samples were no longer acidic, test

tube contents were mixed with 50% glycerine, stirred and then transferred with a pipette

57

to a vial for storage. Slides were prepared by transferring a drop of the sample
suspended in 50% glycerine onto a slide, covering it with a coverslip and sealing the slip
edges with paraffin wax. The glycerine underneath the slip allows the examiner to
rotate a grain, if desired for better viewing, by pressing gently on the cover slip with the

point of a pencil.

In total there were 43 samples of fecal pollens. Detectability of pollens used by insects
is equally high for all taxa, as the size ranges generally above 10 microns. I identified
three hundred grains on each slide by arbitrarily choosing (x,y) coordinates at which to
start a sampling line and traveling along the vertical direction. I repeated the process as
many times as necessary to count three hundred grains. The number of grains sampled
was determined by plotting saturation curves (cumulative number of pollen types vs.
number of grains sampled) for several samples from different locations and dates as I
counted. Three hundred grains adequately captured the diversity of the samples and the
point at which sampling effort no longer revealed many new types (figure 3.1). Pollen
types were identified with assistance of Mr. S. Prasad, slides from the reference
collection, and several illustrated books (Huang 1972; Vasanthy 1976; Nayar 1990;

Tissot et al. 1994).

Due to our knowledge of the flora of BRT, many grains were identified to species level,
and most were identified to genus or probable genus. Palynologists usually can make no
distinction between the families Combretaceae and Melastomataceae, and identify them

only as “Comb/Mel ”type (Anupama et al. 1999). In my samples, however, I identified

58

such pollens as the Combretaceae genus Terminalia due to the dominance of this family
in BRT, the relative rarity of Melastomataceae in all of BRT (ATREE, unpublished
data), and due to the fact that the only other genus in Combretaceae that occurs in BRT,
namely Anogeissus, does not flower simultaneously with the species of Terminalia

which flower during the peak of bee season (pers. obs; ATREE, unpublished data.)

STATISTICAL ANALYSIS

Any taxon in the pollen samples that did not amount to at least two percent of the overall
total for a site combined across weeks was put into the category “other”. Each taxon
that constituted two percent or more of the total for a site was left in its own category,
and all analyses were done using these categories. Hereafter, taxa are also sometimes

referred to as genera.

A Shannon-Weiner diversity index for pollen diet was calculated for each site combined
across weeks (table 3.1) using the equation: H' = p,- 2‘. ln p,-, where p,- is the proportion of

the ith pollen taxon (May 1975).

In order to test the null hypothesis that bees forage in an opportunistic manner and that
their pollen diets conform to an ideal free distribution, several G-tests were employed.
First, 1997 and 1998 data were tested keeping sites separate and pooling pollen
frequencies across weeks. I compared the composition of fecal pollens to the overall
forest composition based on the dominance values of those taxa across all forest types

combined. Dominance values were generated by dividing the total number of stems in

59

each genus of interest occurring across all the sampling plots, by the total number of

stems sampled (n = 2042).

Second, I conducted a G-test to compare the distribution of fecal pollens to the expected
dominance values. In this test, I used the dominance values that would result if the
forest were composed only of the plants utilized by bees, that is, each genus’ dominance
relative to only the other bee plants. The total number of stems sampled in those genera
(nfi79) does not account for anything that fell into the category “other,” thus the
“other” category is not included in the G-test. This tests, within the few taxa that bees

do use, whether they forage in an unbiased manner.

Phenological data from 1998 was incorporated into a third G-test by averaging the
weekly phenology score for each genus of interest on the transect per week, and using
that average phenological score to weight the relative abundance values of each genus of
interest. The G-test then compared the weighted frequencies of flower types to diet
composition. The expected weighted frequencies were standardized to one, so as to be

proportions.

mum

The overall results of the analyses of fecal pollens point to strong foraging preferences
exhibited by A. dorsata, despite the plasticity that generalism confers on them as a
species. Colonies use a small fraction of the available species richness, and even within

those taxa that they use, they heavily overutilize only a few tree Species for the bulk of

60

their pollen diet. Shannon-Weiner indices calculated for dietary diversity (table 3.1) do
not reveal any broad patterns of diversity with respect to habitat type. One reason for
this may be simply due to the differences in number of weeks sampled. However, even
when the weeks sampled are equalized such that the same four weeks of all pollen
samples in 1997 only are included in the analysis, the result does not change. Thus, it
does not appear that the predominant habitat surrounding the nest site determines dietary
breadth. When the same two sites (BK and KA) were sampled across two years, 1998
shows a higher diversity of dietary pollens in both sites, although this result may be
confounded with the variation in number of weeks sampled and/or dates of the weeks

sampled.

Figure 3.2 is a light micrograph showing some of the more common pollen taxa which I
found in the samples of bee feces. Figure 3.3 illustrates that the pollen diet was heavily
skewed toward a few taxa by illustrating the relative contributions of each to the
cumulative percent total. Figure 3.4 illustrates the breakdown among taxa of pollen
composition found at each site pooled across weeks, and figures 3.5 and 3.6 further
dissect the components of the diet and show the weekly pollen composition for each site.
The composition of the pollen diet overall was restricted to 10 genera, excluding pollen
types which occurred at a frequency of less than 2% of the total. These ten genera
comprise a maximum of 18 species of the total 216 species of trees that were
documented in the vegetation plots. Appendix IV lists all species in BRT that occur in
these ten genera, and thus are potentially food plant species of A. dorsata. Appendix V

lists the 27 pollen genera that were found at extremely low frequencies in fecal samples,

61

and were put into the category of “other”. In addition to the taxa listed in Appendix V,
there were several instances in which one or a few grains were simply classified
according to an unidentified morphotype. None of these unidentified types were major

contributors to the sample.

The vegetation plots give a conservative estimate of the number of tree species since the
rarest tree species are not likely to have been captured in the sampling. A. dorsata is
only utilizing a maximum of 8% (that is 18 out of 216) of the tree species richness of
BRT for the bulk of its pollen diet. The samples also contained pollen from Eucalyptus,
coffee (Cofiea arabica) and an unidentifed type either in the Combretaceae or the
Melastomataceae (referred to as “Comb/Mel 2” in figure 3.6), which differed strongly
from that of Terminalia. These three taxa were each found in only one site’s samples,
and for a brief period in the season. Eucalyptus occurred in the BG 97 site, which is a
coffee estate site surrounded by evergreen forest. The estates contain stands of
Eucalyptus (pers. obs.), an exotic to India planted as a fast growing shade tree.
Although these three taxa comprised more than 2% of the total pollen from that site,
they are excluded from the G-test analyses for the following reasons. First, the density
of neither Eucalyptus nor Coffea can be quantified since they were not found in the
vegetation sampling plots, and because the sampling plots did not include the coffee
estates where both of these occur. Second, the Combretaceae/ Melastomataceae cannot
be identified even to family, and therefore its dominance cannot be extracted from the

vegetation plot data.

62

Among the ten genera that contributed to the bees’ pollen diet, representation of pollen
is heavily skewed toward six genera (table 3.2, figures 3.3 and 3.4). Catunaregam and
Terminalia were the only genera found in feces at all sites. Grewia and Syzygium were
at all but one site, and Schleichera and Canthium were at all but two sites. Relative to
their abundance in the forest, use of these plant taxa was highly disproportionate.
Terminalia is in the family Combretaceae, which dominates the dry deciduous forest
(Ramesh 1989), and has four member species in BRT. Of these four, there are three
species of Terminalia which flowered during the time span of fecal sampling, and thus
could have accounted for their presence in the fecal samples. These three species, T.
bellerica, T. crenulata, and T. chebula comprised over seven percent of the total stems
in BRT across all sampling plots, and over 31% of all bee plants in the plots.
Terminalia pollen in the feces accounted for between 3.5% and 24.62% of the fecal
samples, pooled across weeks for each site. Thus, it was consistently lower than the
expected value of 31%. Catunaregam accounted for almost 13% of all bee plants in the
forest, and except in one site, its representation in the bees diet was consistently much
higher than expected, with a range of occurrence in the fecal pollens from 11.02% and
85.60%. Of the other commonly utilized taxa, the Schleichera taxon (which may be
either S. oleosa or the closely related Dimocarpus longan) was relatively rare in the
forest, accounting for only 1.25 % of all bee plants. However, in the sites where it was
used by bees, it was overutilized relative to its abundance in the forest, ranging from
6.2% to 26.85% of fecal pollen samples at different sites. The G-scores for these data,

totaled across weeks but keeping sites separate, all were highly significant, indicating

63

that pollen resources were not used in proportion to their occurrence in the forest, even

when compared to distribution of bee plants only (table 3.2, fig 3.4.).

Similar conclusions arise from analyses of data which used more precise estimates of
relative flowering availability by incorporating phenological variation. G-tests for 1998
pooled data across sites but kept weeks separate, and thus examined whether bees
foraged in proportion to the actual floral occurrence, as determined by flowering status
in any given week, weighted by the abundance of that plant in the dry deciduous forest
relative to other bee plants. The results, shown in table 3.3 are consistent with above G-
tests showing that taxa of pollen were used disproportionate to their occurrence, with
Catunaregam and Schleichera contributing heavily to the positive G-scores, indicating
their overuse, and Syzygium also contributing mostly large positive values. Negative
contributions to G-scores, i.e., those taxa that were underused relative to expectations,
were Terminalia and Grewia, although their negative values were not as large or

consistent as they were in the previous G-tests.

Figure 3.7 shows the phenological status weighted by relative abundance of those
species which belonged to bee plant genera, and which occurred in the dry deciduous
forest. Three species emerge from this picture as having long flowering periods:
Catunaregam spinosa, Grewia hirsuta, and Terminalia crenulata. These three also have
high flowering indices relative to the other bee plants. Of the three, Catunaregam
spinosa has the lowest floral index at any given time, but is also the most overused taxon

of all the bee plants. Images in this dissertation are presented in color.

DISCUSSION

Apis dorsata, while known to be a species capable of exploiting a great diversity of food
plants, appears to utilize a very restricted set of plant taxa for most of its protein diet.
The findings of Devy (1998) and those of Momose, et al (1998) whose observations of
floral visitation led them to conclude that Apis dorsata uses relatively small fraction of
the available flora, are confirmed and extended here. Their studies, which did not
distinguish between nectar and pollen food plants, found that approximately 20% of the
tree species they observed were being used by A. dorsata as food plants. Kiew (1997)
also looked at pollen and honey in combs at different times of season. My approach,
while more effective in identifying all the pollen sources being used without having to
choose which trees to watch for potential activity, could not identify nectar sources. The
pollen diet of the population of A. dorsata in BRT showed that they utilized less than
10% of the tree species available, and that within that group of utilized plants, they

relied primarily on three genera, comprising a maximum of six species (figure 3.4).

The most common pollen by far in the samples belonged to the genus Catunaregam.
This genus is monospecific in BRT (R. Ganesan, unpublished data), and thus all pollen
belonging to this genus came from a single species, C. spinosa. This species has been
formerly classified as Randia dumetorium and Gardenia spinosa, and has recently been
combined with what was previously classified as Randia brandisii (R. Ganesan, pers.
comm). Catunaregam spinosa is a small, thorny understory tree which appears to be

dioeceous (K. Bawa, pers. comm; P. Batra, unpublished data). Flowers are

65

actinomorphic, and male flowers are approximately 1 cm in diameter (fig. 3.8, 3.9).
Both male and female flowers produce nectar, which is also collected by A. dorsata
(pers. obs.), and both male and female flowers are white when new and begin to turn
yellowish after one day. Interestingly, its floral morphology describes a typical “moth
syndrome” (Faegri and van der Pijl 1979)(K.S. Bawa pers. comm). Indeed its second
most frequent visitor after A. dorsata is a diurnal sphingid moth (P. Batra, unpublished
data), thus confirming recent arguments that generalizations based on criteria associated
with classical pollination syndromes may not be useful in identifying the true food
sources of a forager, nor the true pollinators of a plant (W aser, Chittka, Price, Williams
and Ollerton 1996), (Johnson and Steiner 2000). The male flowers open in the early
morning, between 0400 and 0500, and the females open approximately two hours later
(P. Batra, unpublished data). This is a common phenomenon in dioecious plants, and
helps to ensure that visitation to males, and thus pollen collection, occurs before
visitation to females (K.S. Bawa, pers.comm.). Male flowers open with their pollen
actually having already dehisced from the anthers before opening and having been

deposited onto a central pseudo-stigma (figure 3.8).

A somewhat unusual feature of the relationship between C. spinosa and A. dorsata is the
manner in which bees collect pollen from the anthers. The bees’ usual method on most
plants is to cover the body hairs with pollen by walking around among the anthers, comb
their bodies of the pollen, and then pack it into balls which they then carry in their hind
leg corbiculae (Winston 1987). By contrast, all foraging on C. spinosa was seemingly

for nectar, but except in the very early morning when foraging first began, bees were

66

almost always seen with pollen balls 1-2 mm in size stuck on their heads (fig 3.9). The
pseudo-stigma from male flowers appeared to deposit pollen in a location on the bees
that would greatly facilitate pollen deposition onto stigmas of female flowers by bees
foraging for nectar. Observations of visitation to female flowers confirmed that the
heads with pollen balls made strong contact with the stigma (pers. obs.). Balls of
deposited pollen, presumably from C. spinosa, could be seen very often through the
spotting scope while measuring waggle dances, confirming that bees did not remove
them before returning to the nest. An interesting question that arises from this
phenomenon is whether the floral resource was evaluated by dancers for its nectar value
or for its pollen value. Even if it was the nectar that was being advertised by bees,
clearly the pollen is a critical resource, as we can see from its abundance in the feces.
Indeed one might speculate that one of the advantages behind C. spinosa’s preferential
appearance in the diet could be the ease and passive efficiency with which bees collect

it.

In some weeks, C. spinosa pollen comprised more than 80% of the pollen in the fecal
samples, although the rank importance of this tree in the forest is sixth overall. C.
spinosa accounts for 3% of all the trees in the forest, and almost 13% of all bee plants
(table 3.2). The importance of this species to A. dorsata in BRT is disproportionately
large relative to its occurrence in the forest and also relative to its occurrence even when
only bee food plants are considered. In fact it is C. spinosa’s frequency in the pollen
samples that appears to be the most influential on the overall G-scores in both the site by

site comparison (table 3.2) as well as the weekly weighted comparisons of 1998 (table

67

3.3). In all instances, this species contributes an extremely large positive number to the
overall G-score, indicating its consistent overuse relative to expectation. Given that they
are dioeceous, the population size of trees from which honey bees may collect pollen
will actually be lower than that reflected in the plot surveys since the plot surveys
counted both male and female trees. This means that its true expected values for the G-
tests should be even lower than the values reported here, making the overabundance and

ubiquity of C. spinosa in the fecal samples even more indicative of a preference.

Appanah (1981) found that A. dorsata in dense forest are canopy foragers, cueing in on
colors, large floral displays, and utilizing species which have brief flowering periods.
Similar to my findings, Devy (1998), in her study on the pollination biology of trees
visited by Apis cerana and A. dorsata in an evergreen forest in India, found that A.
dorsata used flowers from canopy trees, but that the flowers did not necessarily conform
to morphological “syndromes” expected to be used by honey bees. She also found that
the bees relied very heavily on only two species, and in contrast to the findings of
Appanah (1981) found that the most important of them, Palaquium ellipticum had the
longest flowering period of any of the trees utilized. Additionally, the second most
important species, Elaeocarpus munronii had a flowering period that did not overlap in
time with that of P. ellipticum, and in fact acted to extend the duration of A. dorsata’s
stay in the forest before its migration. Both of the two tree species flowered annually
and produced high volumes of nectar, unlike many of the supra-annual flowering species
that were used less heavily by the bees. Devy found further that honey bees’ migration

into the study area always coincided with the onset of P. ellipticum flowering. She

68

concluded from all of these observations that A. dorsata utilized those species whose
reliability was high, both within and between years. Like P. ellipticum, Catunaregam
spinosa has a flowering period that lasts almost throughout the dry season (fig 3.7); thus
its reliability as a food source may make up for its lack of dominance in the forest, and
for the fact that its stature as an understory tree means it may not be easily visible to
scouting bees despite the relatively open canopy. Furthermore, since passive collection
of C. spinosa pollen seems to be an added benefit of active nectar collection as
explained previously, it may actually be a preference for the nectar from this tree that is

the most important factor in explaining its overuse.

It does not appear that the abundance of a floral type in the forest over an extended
portion of the flowering season is enough to explain its abundance in the bees’ diet. The
genus Grewia is actually slightly more abundant in the forests of BRT than C. spinosa
(fig 3.4) and exhibits a flowering season that is also extended over several weeks. In
fact, its flowering status is higher at any given time than in that of C. spinosa (fig 3.7).
Its pollen was indeed used by A. dorsata, but was consistently underutilized relative to
expectations (table 3.2, 3.3). This is a surprising result, but one perhaps explained in
part by the fact that Grewia flowers are orange, and thus may not produce a display that
attracts A. dorsata more than other flowers that are in the preferred color spectrum of
bees (Barth 1991). Bees can and do learn to associate rewards with colors other than
those that stimulate their peak wavelength receptors, but may not do so readily if floral
resources that fall within the peak wavelengths are concurrently available; that is, bees’

ability to learn depends also on the changing ecological context (Barth 1991).

69

Similar to Grewia, upon further examination of figure 3.7, it is clear that Terminalia
crenulata also experiences a long flowering peak and yet its abundance in the pollen diet
is surprisingly low. In fact, Terminalia is the most common genus in the forests of BRT,
with the three species accounted for in this study comprising close to ten percent of all
adult stems, and 30% of all bee-utilized plants (fig 3.2). Its utilization, however, is
consistently lower than expected based on its frequency (table 3.2). The flowers of the
genus are small and arranged in spikes which occur together in clusters (fig 3.10), and
the trees often exhibit large floral displays. Such characteristics are amenable to Apis
foraging behavior in which they may move between flowers rapidly. Indeed the genus
appears to be extremely important as a source of nectar (M. Kethegowda, pers comm;
pers obs.) With respect to pollen foraging however, there is some evidence to suggest
that although T erminalia flowers appear to be hermaphroditic, they may be cryptically
monoecious or even dioeceous (A. Sinha, unpublished data.) If this is the case, the
number of individual Tenninalia stems producing pollen may be much smaller than
reflected in the plot surveys, and the hence the expected values reported in the G-tests

may be gross overestimates.

Schleichera oleosa (and its close relative Dimocarpus longan, whose pollen is

indistinguishable) might be considered a keystone resource, in that it is an early resource
(fig 3.2) utilized by immigrating A. dorsata colonies. It provides pollen protein at a time
in the season when flowering is not yet abundant and is needed by colonies settling in to

build their combs and start brood production. Indeed, it fits the qualitative definition of

70

a “keystone species” in that it is a relatively rare species whose utility is disproportionate
to its occurrence (Power et al. 1996). Given the generalist diet of the species as a whole,
it is not likely that A. dorsata in BRT is obligately reliant on any one resource. The sites
in which Schleichera pollen was never found (DS 97 and BK 97) also correspond to
sites where the first week was not sampled. However, in both of those cases, week two
was sampled, which is the week during which Schleichera peaked in the other two
samples that year. Despite the fact that Schleichera forest types occurred within the
areas of DS and BK sites, the possibility remains that the flowering times in those areas
may have exhibited enough variation from the other two areas for the pollen not to have
shown up in the samples that began in week 2. It may also be the case that there simply

is spatial heterogeneity in reliance on various pollen types.

The results here do not account for differences inherent in amount of pollen per flower,
or the variation between species in the number of flowers per tree. 1 account for
temporal variation in flowering by using phenological variation to weight abundance,
but here I do not deal with any spatial variation in availability. Thus, all species are
assumed to occur evenly across the forest. The next chapter will account for both spatial

and temporal variation with the use of GIS.

Although giant honey bees occur across a wide geographic range, their food plants vary
across that range in presence or frequency, and bees’ use of these food plants may vary
in different areas. For example P. ellipticum, which was of primary importance in

Devy’s study area in a wet evergreen forest southwest of BRT (Devy 1998), does occur

71

in low densities in the evergreen forests of BRT (Ramesh 1989), but its pollen did not
appear in the fecal samples. Kiew (1997) examined the pollen stored in combs of A.
dorsata in Malaysia and found a dominance of Eugenia, which is synonymous with
Syzygium, a taxon that was commonly found in my samples as well. Jhansi et al (1991)
found many of the same genera I found in their analysis of honey in another part of
southern India, although they did not report finding of Catunaregam. It may simply not
occur in their study area. However, a heavy contributor to the pollen found in the
honey was Cassia. This is a taxon which occurs in BRT, but requires sonication, or high
frequency “buzzing,” for pollen removal. Honey bees are incapable of wing buzzing at
frequencies high enough to dislodge the pollen, thus they probably only collect the
pollen that remains after sonication and pollen foraging by Xylocopa (K.S. Murali, pers.
comm). Jhansi et al’s findings of it may simply reflect bees collection of nectar from
Cassia. In summary, there do appear to be some taxa that are used across different parts
of the geographic range of giant honey bees, as well as others that may not be of equal

importance in different areas.

Based on the results observed here, it appears that A. dorsata does exhibit preferences
for certain pollen resources, and that these food plants may be disproportionately
important in maintaining the population size of honey bees in BRT such that they also
forage on, and pollinate, several other species. Since nectar sources were not accounted
for in my study, there remains another set of food plants whose pollination also may
depend on A. dorsata; thus the indirect links between plants in the forest become even

more complex. If Schleichera oleosa can be considered a keystone resource, that implies

72

that more common taxa like T erminalia spp. are actually indirectly dependent on it via
its maintenance of the common pollinator pool. Murali and Sukumar (1994) found in a
study in a forest very similar in composition and structure to BRT, that rarer tree species
tended to flower earlier than more common ones. They postulated that this may be an
adaptive strategy for avoiding competition for pollinators. Phenological divergence as a
mechanism for avoiding competition for pollinators has been hypothesized by a number
of authors (e.g., Frankie et al. 1973; Gentry 1974). Especially in ecosystems such as
BRT where species share a pool of generalist pollinators, such a strategy may indeed
prevent the diluting effects on pollination that flowering by the more dominant species
may have. Additionally, such a phenological schedule may act to extend the season
used by the bees, although it is still unknown as to what resources they rely on prior to
their immigration to BRT. If the relatively rare species S. oleosa were suddenly absent
from the forest, the question then would be whether A. dorsata would time its inward
migration to be later in the season when other flowers are available, or instead just
exercise a “generalist” strategy of relying on something else that may occur in other
strata, be of less nutritional value, or be less conspicuous. Many of the other tree species
that flower simultaneously with Schleichera produce large, red flowers (ATREE,
unpublished data; pers. obs.) mostly used by birds (Murali and Sukumar 1994) and may
actually be visible to bees (Chittka et al. 1993) but perhaps not highly stimulating or
preferable (Giurfa et al. 1995). Bees, however can learn to associate rewards with colors
(Dukas and Real 1991; Waser et al. 1996), thus adding another level of complexity to

their ability to adapt in ecological time.

73

Catunaregam spinosa also may be considered a keystone resource in BRT due to its
heavy usage by giant honey bees. However, A. dorsata clearly can and does utilize
other species and their preferences may not be based on nutritional preferences but on a
complex combination of sensory and reliability (risk-aversion) factors (Barth 1991).
These factors may result in lower prioritization, but not exclusion, of less useful food
plants, e.g., Grewia. Hence, despite the complex links, forests such as this with large
populations of generalist pollinators may be largely buffered from cascading extinctions
that could result from the breakdown of the pollination mutualism. In order to
distinguish whether the diet of bees in BRT is flexible enough to accommodate dramatic
changes in forest structure, long term documentation of changes in bee diet
corresponding to interannual variation in flowering phenology would need to be done
across multiple sites which have some overlap in their plant composition. Measures of
actual colony growth and reproduction also would be needed in order to examine the

true adaptive nature of dietary flexibility.

Intriguing mysteries remain regarding A. dorsata’s ability to respond opportunistically
to sudden and aseasonal bursts of forage. In some instances, such as the dramatic and
supra-annual general flowering periods in the Dipterocarp forests of Southeast Asia, bee
colonies appear to respond immediately to the increase in forage by a drastic increase in
inward migrating colonies (Sakai et al. 1999). It is these species that they most often
provide pollination to in these forests and not the more reliable but less abundant ones.
Similarly, anecdotal reports from India state that the supra-annual flowering of

Strobilanthes spp. (Acanthaceae) results in a much larger number of colonies than in

74

non-flowering years despite the fact that it occurs in the herbaceous stratum of the
forest. Somewhat in contrast to Devy’s finding (1998) that A. dorsata utilizes reliable
and relatively constant sources of forage over more ephemeral or acyclical ones, Sakai et
a1 (1999) reason that it is unpredictability of resources that allows highly social,
generalist bees to dominate in Asian forests as opposed to the more specialized bee
pollination guilds present in the Neotropics. Perhaps both are true. The generalist
ability of A. dorsata combined with its migratory ability may make it a forager and
pollinator that depends on constant reliable resources and persists in the forests in
relatively low numbers most of the time; however, it may also be an effective pollinator
of less reliable species which produce massive flowering bursts, unpredictable in time

and space, due to its ability to track resources and respond to them immediately.

Although honey bees are generalists and are not thought to have undergone strict
coevolution with a restricted set of plants, they may exhibit tendencies to specialize
locally. The ecological consequences of such facultative preferences may have
important implications for how Old World tropical forests may be structured, while also
buffering against the loss of the pollinator community during periods of rapid change.
By examining how this generalist forager and important pollinator may link its food
plants together into a complex web of indirect mutualism, we can begin to generate
hypotheses about how behaviors such as foraging preferences scale up levels of

biological organization to community and ecosystem properties.

75

Table 3.1. Shannon-Weiner indices for pollen diet diversitya

 

 

 

 

 

 

 

 

 

 

 

 

SITE AND YEAR HABITAT H' (N WEEKS) H' (4 WEEKS
STANDARDIZED)

DS 97 Evergreen 1.40 (7) 0.99

BG 97 Evergreen 1.71 (6) 1.43

BK97 Deciduous 0.56 (4) 0.56

KA 97 Deciduous 1.31 (8) 1.20

BK 98 Deciduous 1.58 (5) --

KA 98 Deciduous 1.69 (9) --

SG 98 Deciduous 1.86 (5) --

 

76

 

 

 

 

 

 

     

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

— 325922 0000
0:862: «eachfimcn iaaaedhuu .Iiemfihefl 9.09333 euhmlhéwfln Quinta: 0.58-0
28.8 200.0 5.0: 23.: 20.0 80.3 20.: 2.0.8 550
22.0
8.05 20038.5 28033.5. 8.80 28.0; 288° 00.2- 20%... m: use
2.8.: : 28.3 28.30 2808
80 $.83 8.000008 60.83200 2.02 800900.80 8.82 8.2 30 582356
28.8 80.8 5.8 2H3; 60.8
28.8 o 0000- 289 o 8021 0000- 8.2 3.2- 80 02 $5580
000: 28.8 20008 2088
00.30 28.90 8.08 28.30808 3008 80:: 28.80 03 and 2082030
8.8 23.8 28.8 ans
0% 28.80 0000- 50328 202002.27 BR- 80? 2.2 80 £523.
and 200.8
E 0- 28.8 o 28.8 o 28.8 o 28.8 o 40.2- 289 o M: .4 00.0 352583
28.2.8 gm: W00: 50.9
3:- 2.2.2- 50:802-23580: 2000:008- 802- 8.8- an: 2.2- 88.58
29.00 fines and 80.8
:00- 3.033.5- 05.5- 093802. 2.8- 2.8- 28.80 3.: 9.0 586
28.8
28.8 o 28.8 o 8.9. 28.9 o 28.8 o 28.8 o 28.8 o 03 43 SESSSF
28.9 2030 and
28.8 o :00- 289 o 3.8. 800. 28.8 o 28.8 o 8.: 80 sates:
0.2.25: amaze-.2
0 80: Re: 003. 55. 00.: 80.0 mm: a. z. a.» 0:200

 

 

 

 

082m 03 00 20 822202288 05 2 .Domoaao .2050: 2.2088 0023 0.220 .2000 08.288
0202-0 89% 0.8 220 :80 :02 00800.0 280.2. 0805:0822 :2 :0>2w 02 0:02—22 .82 0.0220 0 he 2:00:02 0200? 0080: 2.28.2 0220
.52 280-0 282 2 82:52:88 03:0w 80.: 20223282 20:: .222. :02222 03 22826 528.888 008 00:2 20 8222092800 .N.m 030.2.

77

Table 3.3. Partial G-scores for 1998 pollen diet composition compared to forest
composition with plant taxa weighted by flowering status per week.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

EOLLEN WEEK 1 WEEK2 WEEK3 WEEK4 WEEK 4.1 WEEKS
YPE

lDalbergia -2797 4.50 -9.72 -734 -315 0.00]
Forewia -952 42.88 -87.36 -81.48 -79.66 -138.36|
Terminalia -144.18 -19.10 -67.92 50.96 -23.50 —70.82
lagerstromia -0.0002 -0.0002 -12.10 -19.40 -18.40 7.41
S zygium -300 -8.70 76.47 383.90 667.27 571.57
Schleichera 1011.19 639.78 1000.63 561.69 198.41 114.85
[Catunaregam 762.80 427.64 947.04 380.79 107.80 574.58
E‘anmium 68.90 39.68 12.99 4.73 -350 0.00.
[cs-score 3316.43 2063.81 3720.04 2547.6966 1690.5083 2118.4531'
(*ttp<.001) *1“! Illa"! ”It Mali *1“: "It

 

78

.053 5:3 5 .3836 05

mo 52: 02:an 0.5an 8m 00 E50 wE—afiam @0383 5:8 0800—00 hinged». 08 0033 :ouguam ._.m 05mm

 

 

 

 

 

 

8.08% a 8:8 .6 868:2
8m 8m om 8N ca 9: 8 o
_ _ _ _ _ _ _ _ O

0

n

0

$68 I m.

m.

335 11111 m

n

3850 .- I I m

m

938 I l m.

w

58mm“ - - I m

m

0-55 I . l 2
03.4.2 I

 

 

 

79

.8388 30$ E .5888 803 5033 053 5:8 mo mama-"meow: Emfi Nam Bswfi

0.303030% 3335005
_

 

 

 
 

 

 

 

 

 

 

 

 

 

 

 

 

. \EaEauaW

1 3.3.26

 

 

80

.39 00
E0800 030.0050 000062 002? 008-» 4300 .00 E300 000065 was 0>0nu 0009602
00—003 000 00:0 000000 00—000 300 0:0 00 :03 00:00 £000 00 02:55:00 .m.m Ezwfi

0053 00:?”

m m.-
0 m M. m. m
m. w. ... m. 0.

11183018110183
81111111111101
amqowlqos
111018-1253

 

 

 

 

O

m... D

m %
“Io

. . mm H
:N 00 Ulc—

Tom
”L; m
- w
n m.
HOV w
m w...
low
m m
Fee a
n w
”12
wow
W00
.OA:

NNAV

81

.33:

0:2: 00: 5:8 8 03:20: 3:2: 085 00 86:05:20 0:: £805 530:8 9mm 05:0
05 E 3:2: 5:800: :0 86:05:20 30% 33:00:00: 0:0: 03: HEE 00—003 380:
00.00: .030 :80 H: @0388 0000: E 6:38 0090 5:0: :0 00:0:0zw8m 4mm Eamfi

 

 

 

   
    

 

3 n. 3
m 3 N. S 3 I. W 0
w m a m. w m. m m w a
J m m w u w m. I i _ m _ I.
_ _
i _ . .q ._ . 3:
08m!

88 D
80: n
85. D
35. I
03.: fl
5.: a

3:2: 00:
:05: 2 02.20..

 

 

82

aIdwes JO waxed

 

 

 

 

use =38 I

 

 

0:0 3 ma .Q 0:0

no Om .U 0:0 3 <M .m 0:0 3 Mm .< :028950 :26: .3383. :88 A2003 .m.m 0::wE

E:u0..::3a0n gain—=05:

:05: I 80:0_0_:0mI £30..va
~05 I 5330mm Im=auaooa0§ I
535.30 I mans—ausm I Suki—.5 I

 

 

.00:
a h E. e m w m u _ n b .
y...
-2
1.
16V”
3
w
.Sm
m.
,8
.9
r2:
. F . m 0 n u _ a h e
Fl'rlflr L I. u _ -
.3
m
.3...“
m
ism
2.
.m

 

e:

[not wand

 

 

you

I
e
V

mo: was: ad

I
9
”

 

. ee—

83

.3: 3 cm .0 a: ma 5: .m 28. ma Mm .< 82888 8:8 .8368. .80: 583 .3 250::

e —.m

 

 

 

 

 

 

hvn—ué I
:0ch I

N ESE—:00 I
sausage n...
0.30.5 I

 

30.3.0: I

 

 

.02.

 

 

 

 

e
v
1810:)uaoJad

. e9

:5

 

r en

.02.
u m 3 v n a —

b b

 

 

 

 

 

 

 

 

 

 

 

 

 

 

-.
.3
-:
.s

.U
8

0 m e m 0 _
-5 1°
.8
:3.
.8

.<

 

mo) nomad

mo: maxed

84

25% a: :3 80 m 05

c 50.50: :0: 0020> wEBBoc 80:3 380:0 :0: 88:05 05 :o $006365 :0 .8 03000
3:030: :02: 05 0:008:00 0:8-» :0 0:09:52 000.5: 0000200: D: 5 3:2: 00:
8.8:: :0 00:00:30 03:20: 3 ©0330; awe—0:2: 3:36: :0 gnaw—0:800 5m 0.5mm:

 

803008 £00050...
§EE§§O
ggfigfisgfiv
£23 093...:
8:05: £80603

figagifia
E§§§8§§§0
cagflfifiafi
flfifiiaéfig

£5008? 55336
«5.88.. «3.3.5

 

+

 

E§3§i85nlll

 

 

 

8;.
-20 0
L3 s
12.0
12.0
1:0
:20
-20
r20
10;

 

ad

 

 

85

 

.0332: 0:03:03 00:35:50,: ofim wE

 

.000: 05 :0
00:60:00 :0:0: 0002 0:030: 300E.” Eumfigscb
£05 :0 950:0: 80:00 0.30. .0.m 0:3,:

 

.0Ewnm
-0008: 30:00 05 :0 :0:0: 0:02
:030: 20:. 5.03% U .w.m 0:0me

86

CHAPTER 4: SPATIO-TEMPORAL DYNAMICS OF FORAGING

INTRODUCTION

There is abundant evidence that the spatial distribution of resources should affect animal
movements, and that such movements in turn affect ecological landscape patterns. Some
of the more important patterns affected by animal movements include the distribution of
herbivorous insects across the range of host plants (T urchin 1991), habitat patch selection
(Ims 1995) and coevolution of geographically structured populations (Thompson 1997),
interactions between species in a guild (Danielson 1991), conservation corridor
effectiveness and spread of disturbance (Lima and Zollner 1996), and the plant-pollinator
relationships (Bronstein 1995). Theoretical and empirical work on insect movements and
their ecological consequences has mainly addressed patch selection by herbivorous
insects (e.g., Kareiva 1982; Turchin 1991). Far less work has focused on the landscape-
scale ecological consequences of the movement patterns of pollinators, whose
relationship to plants is of a very different nature than that of herbivores. Bronstein
(1995) argues that the patchy and ephemeral nature of floral resources and the importance
of pollinators in structuring plant communities should have great influence at the
landscape scale on the persistence of different types of pollinator mutualisms. Her call
for attention to the “plant-pollinator landscape” provides a sketch of the types of traits in
both plant and animal mutualist that will determine whether the mutualism persists over
time, and if so, the processes that will influence it. Critical to the study of the plant-
pollinator landscape are the phenological variation of food plant populations across space
and time, the foraging and searching behaviors of the pollinating animals, and the relative

spatial scales of these two processes.

87

The movement patterns of central place pollinators, such as honey bees, are certain to
have spatial consequences that differ greatly from the movements of widely dispersing
insects. In fact, it may be true that central place foraging makes the study of
heterogeneity of resources and the correlated movement patterns of insects more tractable
over large spatial scales than dispersal types of movements because central place foraging

provides a natural center from which to search for pattern.

While a voluminous body of literature exists on foraging behavior and decision making
by animals at small spatial scales (e.g., Pyke 1978; Schmid-Hempel 1984; Stephens and
Krebs 1986), it is extremely difficult to scale these results up to large spatial scales and
complex foraging environments (Bronstein 1995; Lima and Zollner 1996). One difficulty
is simply to observe large scale foraging patterns, but I have shown in an earlier chapter
that honey bees afford unique opportunities to gain insights into landscape-scale foraging
patterns through observations at their nests. Another difficulty is to integrate information
on foraging behavior with data on spatial and temporal variation in food availability.
However, this second difficulty can be met through the use of spatially explicit databases

and modeling tools available through Geographic Information Systems (GIS).

In chapter two, I used foraging locations inferred from bee dances to examine overall
patterns of flight distance over time in the flowering season. The conclusions which
emerged from that set of analyses suggested that besides dynamics between colonies,

foraging distances might be mediated by different levels of resources that were not

88

 

detectable by the broad scale surrogate of “week in flowering season”; that is, changes in
flight range when compared across weeks may be related to differences in site-specific
resource levels. In chapter three, I found preferences for certain pollen-providing food
plants both with and without temporally explicit differences in floral availability. In this
chapter, I draw upon the forage maps, fecal pollen composition, and phenological

variation from two 1998 deciduous forest sites. Combined with baseline distribution

— ”1.“ '.Le 1

maps of important bee plants, I use an integrated GIS environment to examine whether

colony foraging distance and selectivity in pollen foraging varies as a function of the

spatio-temporal variation in site-specific resource availability of the bees’ preferred plant

 

as” “ .

taxa.

METHODS

Geographic Information Systems (GIS) provide a powerful tool for ecological analysis by
integrating several spatially referenced data sets onto the same map. By overlaying
different layers of data, for example, bee movements and plant distribution, it is possible
to visually examine and illustrate spatial correlations between different factors. If
numerical values such as flight distance and flowering intensity are associated with each
point on the map, the numbers of the data sets that correspond to a given area or point on

the map can pulled out and subjected to statistical analysis.

89

SPATIO-TEMPORAL CORRELATES OF FLIGHT DISTANCE

The following analyses rely on a combination of data gathered using methods presented
in earlier chapters. These include the forage maps (Chapter 2), phenology measurements
(Chapter 3), and pollen diet composition (Chapter 3). Since phenology was measured
only in 1998 and can be expected to vary between years, forage maps from 1998 only
were used. Data sets used for analyses in this chapter are the forage maps from sites KA
98 and BK 98, 1998 phenology numerical weights as described in the previous chapter,
and the frequencies of six genera in the vegetation plot sampling done by ATREE. In
each of the 125 2 x 2 km sampling grid cells, the frequencies of stems >10 cm DBH of
Canthium (summed across C. dicoccum and C. parwflora), Catunaregam spinosa,
Grewia (summed across G. hirsuta and G. tiliaefolia), Schleichera (summed across S.
oleosa and Dimocarpus longan), Syzygium (summed across S. cuminii and S.
malabaricum), and Tenninalia (summed across T. bellerica, T. chebula, and T.
crenulata) were multiplied by the average phenology index for each of the six weeks of
phenology sampling. This resulted in a weekly matrix of relative flowering availability

per cell per genus.

The vegetation plot sampling design, as described in chapter 3, was at the scale of 2 x 2
km; however honey bee flights mostly occur at a radius less than 2 km (see chapter 1).
Therefore, I wanted to know the relative flowering status of each taxon at a finer spatial
scale than that available from the plot sampling data. By using the Geographical
Information System (GIS) software Arc View 3.2 and Spatial Analyst 1.2 (Environmental

Systems Research Institute 1996)., I interpolated values between the sampling plots in the

90

center of each grid cell at a spatial resolution of l x 1 km. Arc View has two different
computational functions available in the “interpolate surface” option. These functions
interpolate between data points in different ways. One method uses the regularized spline
function fitted with neighboring points. This method allows for smoothing of the
interpolated surface, which is achieved by minimizing the curvatures between the input
points (Environmental Systems Research Institute 1996). The alternative method is the
inverse distance weighted interpolation, in which the values at points between the input
points are computed as a function of the distances separating the input points. The spline
function was preferable in this analysis because it allows for values to exceed or fall
below the input values, as opposed to returning an interpolated surface whose values
always fall between the input minimum and maximum. Since the values in the grids are
relative flowering indices, and not absolute numbers of trees or flowers, it was more
useful to have values that were not constrained be the maximum and minimum of the
other cells. Thus, some values in the grids are negative values, meaning only that they
are expected to have relatively less flowers of a particular taxon than other cells with less
negative values. Figure 4.1 shows an example for Catunaregam of the interpolated

weekly surface, as well as the overlain forage maps, explained further below.

One weakness of the approach of using relative and not absolute phenology measures is
that it relies on the assumption that all tree taxa are equal in the size of their flowering
display and rewards per flower. In nature, this is not likely to be the case, but is an issue
which requires a large amount of data on the relative amounts of pollen produced per

anther, number of anthers per flower, and number of flowers per tree for each taxon.

91

Such data sets are not available for the flora of BRT, nor are they available in a
comprehensive fashion for any other bee flora. The approach I have used does not
address the fine scale differences in levels of pollen availability, but addresses the issue at
a larger spatial scale, that of floral patches produced by the distribution of trees and
relative visitation and foraging at those patches. Furthermore, from a plant’s perspective,
the relative amount of visitation by a generalist pollinator is what matters more for the

process of pollination than the energetic efficiency of pollen foraging.

A second assumption that the following analyses rests on is that there is no variation
across space in flowering phenology. The phenology data were taken on a transect that
ran through one area of BRT; thus by using those ranks for any given week in two
different sites in BRT, I assume that there is there is little spatial heterogeneity between
sites, or between the sites and the transect with respect to flowering status. Personal
observations confirm that this assumption is generally valid in the dry deciduous forest;

however, no data exist regarding actual levels of variation.

1 selected forage maps for BK 98 and KA 98 by week and overlaid them onto the
corresponding week’s interpolated surface for each of the six plant taxa. I then extracted
flowering index values from the surface at the intersection of bee foraging location.
Finally, I averaged the flight distances per colony per week at each site, and also
averaged the corresponding flowering indices for each plant taxon. Thus, the resulting
numbers were average flight distance flown by a colony for each week, and the flowering

availability averaged across each of the foraging points for that colony. Colony level

92

differences in flight range appear from the analyses in chapter 2 to be due to indirect
interactions between colonies by which they avoid using the same patches as other
colonies, thus resulting in flight ranges that shifted relative to each other. Since the
analyses here are not concerned with the differences between colonies, summarizing at
the colony level removes variation that is attributable to those differences, and leaves in

the variation associated with the level of overall site.

To address the question of whether the six preferred plant taxa had any influence on
flight range, I performed a stepwise regression with flight range as the dependent
variable, and flowering status of each taxon as the six predictor variables. The stepwise
procedure was done using JMP software (SAS Institute 1999)] as a “mixed” stepwise,
that is, the iterations of computing the best model by adding and removing x’s were done
in both the forward and backward directions using a value of p = .25 as the entry and
removal criterion. A p-value greater than .25 resulted in a variable being removed from
the model, and a p-value less than .25 resulted in a variable being added to the model. A
stepwise regression is most often used when the number of x variables needs to be
reduced to only those that explain the majority of the variation in an independent variable

(Sokal and Rohlf 1981).

I used the Mallow’s Cp criterion of model selection to determine the cutoff number of
variables. This approach finds the point at which Cp approaches p, where p is the number
of model parameters (SAS Institute 1999). I used this method of model selection instead

of simply using all the variables returned by the stepwise regression because in order to

93

perform a multiple regression using the x variables “chosen” by the stepwise, there
should be roughly a 5 to 1 ratio of sample size to number of x variables (T abachnick and
Fidel] 1983). The sample size in the KA 98 site was N = 17, and thus not large enough to
accommodate four predictor variables. The Mallow’s Cp criterion actually resulted in
limiting the number of x’s to three, and I used the same criterion in the BK 98 (N = 23)
for the sake of methodological consistency. Once the stepwise regression eliminated the
plant taxa which did not explain a significant portion of the variation in flight range, I ran
a multiple regression of flight range on the flowering indices of the remaining taxa to

determine the predictive value that those taxa have on flight distance.

SPATIO-TEMPORAL VARIATION IN FOOD PLANT PREFERENCE

In the previous chapter I tested whether the representation of pollens in the bees’ diet was
associated with floral availability in the environment, as determined from the phenology
data. Here I expand on this question by explicitly integrating spatial and temporal
variation in floral availability in the GIS environment, and then asking whether the bees’
diet varied in association with changes in local density of particular floral species. As in

the previous chapter, the null hypothesis is that bees have no food plant preference.

To carry out this analysis, I generated an average flowering index for each week that took
into account the densities of each plant taxon only in the actual places where the bees
were flying, as opposed to assuming homogeneity in the distribution of plant taxa across
all bee nest sites. To do this, I averaged, for each site, the interpolated flowering index

scores for each taxon across all foraging locations per week. The proportion of the total

94

sum of these indices per week per site generated each taxon’s expected value for a G test.
In this test, the null expectation was that pollen use was proportional to floral availability.
In the BK 98 site I added 1 to each index value in order to eliminate several negative
values. This was necessary because G tests use a natural logarithmic function in the
computation of G-scores. The observed value for each taxon was its number of pollen

grains in each site’s weekly sample.

RESULTS

Spatial and temporal variation in flowering status appeared to be linked to variation in
flight distance, a pattern that was not detected in the previous coarse grained analysis
(chapter 2). Results of the pollen diet analyses here largely confirmed the pollen
preferences found in the chapter 3, with some new site-specific patterns emerging as
well. I discuss each of these analyses in turn. Images in this dissertation are presented in

color.

CORRELATES or FLIGHT DISTANCE

The stepwise regressions of flight distance on six different taxa originally returned four
variables as explaining a significant portion of the variation at both sites (table 4.1, 4.2).
Using Mallow’s criterion, I determined that each model should restrict the number of
variables to the top three. In both sites, Terminalia and Grewia were chosen by the
stepwise procedure; additionally, the BK 98 site model included Schleichera and the KA

98 site model included Syzygium.

95

The multiple regression yielded results which showed consistency across sites, as well as
site-specific divergence (table 4.3). T erminalia emerged as the most important predictor
of flight range, with its flowering availability having a highly statistically significant on,
negatively correlated with, flight distance in both sites. Other taxa did not show
consistent effects across the two sites, and thus may be indicative of locally specific
phenomena. Although Grewia was signifith in both sites’ models, it was positively
related to flight distance in BK 98, but negatively related in KA 98. In BK 98,
Schleichera, the third taxon chosen by the stepwise regression, was not statistically
significant in the full model, but in KA 98, a third taxon Syzygium, was statistically

significant and negatively correlated to flight range.

Foon PLANT PREFERENCE

With respect to pollen diet composition, the results were largely similar to those found
when diet was examined without accounting for spatial variation in resource abundance,
but also yielded some new insights (table 4.4). Catunaregam in BK 98 was overused
relative to expectations, but was not a particularly heavy contributor to the partial G-
scores in KA 98. Instead in the KA 98 site Syzygium appeared to be more overused in
weeks when Catunaregam’s importance was low. Terminalia was mostly underused,
although its G-values do not differ strongly from the null expectation of zero, thus
indicating that it was used according to expectations. Schleichera was overused in both
sites early in the season, used according to expectations late in the season in KA and for
most of the season in BK, indicated by its small positive and small negative G-scores. It

was never grossly underused. At its peak time of availability, it is overused, and it at

96

other times it is used according to expectations. However, since the flight range data
analyses do not indicate that Schleichera was a correlate of flight range, it appears that
the apparent preference for Schleichera is mediated by its availability and is not
something for which the bees will expand their flight range in order to seek out when its

availability drops.

DISCUSSION

By incorporating spatial and temporal variation into the analyses of flight distance and
pollen preference, the complexity of the relationship between Apis dorsata and its forest
food plants begins to emerge. Indeed the heterogeneity in the distribution of bee plants
across the forest and through time does appear to influence flight range of colonies as

well as result in site-specific patterns of floral utilization.

When results of the spatially explicit pollen preference analyses are compared to those
that did not incorporate spatial variation in resource levels, it becomes clear that
preferences for taxa do exist, but they are diffuse and mediated by relative availabilities
of a few selected food plants. For example, the strong preference for Catunaregam
shown in the previous chapter is somewhat diluted once the pollen distribution is made
spatially explicit. Feces at the KA 98 site in fact showed little deviation from expected
values for Catunaregam but instead showed an overuse of Syzygium (table 4.4). While
Syzygium was a highly preferred taxon in the weekly 1998 analysis of chapter 3 which
used phenological variation and pooled across sites (table 3.3), it was not overused in the

BK 98 site when analysis pooled across weeks. That is, at no level of pooling was

97

Syzygium preferred in BK. Therefore, the BK results across the board are different from
KA results with respect to Syzygium. Such site-specific preferences may be due to initial
discovery and learning of different food sources, differences in tree size and thus floral
displays and crown attractiveness of certain tree taxa over others in some environments,
or other factors that are particular to the local environment beyond differences in relative

abundance.

The site-specific pollen preference results are confirmed independently when the flight
range predictors are examined. At KA 98, Syzygium floral availability had a strong
influence on flight distance. The relationship was negative, suggesting that when
Syzygium flowers were readily available, bees did not travel as far as when Syzygium
flowers were rare. At times when Syzygium was rare in the forest, bees flew further from
the nest. This was not the casein the BK 98 site, which is also consistent with its lack of

importance in the pollen diet at that location.

Catunaregam was a preferred plant in BK 98 and mildly so in KA 98 (table 4.4).
Somewhat surprisingly, however, its floral availability was not a predictor of flight range
variation; that is, bees are not tracking the resource, despite its overall importance in the
diet. Even if they were more reliant on Catunaregam as a nectar source and not a pollen
source, as postulated in the previous chapter, the floral index would remain the same and

any tracking behavior should be apparent.

98

Schleichera was among the four chosen factors in the stepwise regressions, indicating its
marginal importance, but was eliminated from the model as the least important of the
four in KA 98. It was not eliminated from the BK 98 model, but did not turn out to be a
strong predictor of flight range according to the multiple regression. Thus, from a
combined interpretation of both the fecal pollen results and the flight distance results, it
appears that Schleichera did appear to be used preferentially when it was available, but
bees did not track it when its peak flowering declined. Schleichera is one of the first
trees to flower as the immigrating colonies come in to BRT (chapter 3; figure 3.6). Thus,
it is an important early resource, perhaps only because there is less to choose from early
in the season. Once other pollen sources begin to flower, giant honey bees no longer seek
out Schleichera. Thus as a relatively rare plant in the forest, its importance lies in the
tinting of its floral display, and perhaps less in its total contribution as a nutritional or

energetic resource.

Grewia was a pollen resource that was used in proportion to its occurrence in both
locations. Although Grewia was among the strongest of the six predictors of variation in
flight range at the two sites, its relationship to flight range differed in sign across

locations; thus, its influence in flight range may reflect site-specific differences.

The most surprising result of the combined spatio-temporal analyses was the unintuitive
result of Terminalia. As a pollen resource, it was most often strongly underused (table
4.4). This is consistent with the results from chapter 3 where Terminalia had a negative

G-score in almost all cases (table 3.2, 3.3). In the multiple regressions, however,

99

Terminalia was a statistically significant correlate of flight distance, showing a negative
relationship to flight distance. That is, when Terminalia flowering index was low, flight
range was high, suggesting that bees are foraging close to the nest when this resource is
frequently encountered, and flying further when it is not readily available in the vicinity.
This is somewhat puzzling given that they did not exhibit any preference for its pollen;
however, bees are known through much anecdotal evidence and observation (A. Sinha,
unpublished data; Soligas, pers. comm; pers. obs.) to rely strongly on Terminalia for
nectar. Most of the Terminalia flowers in the forest belong to the species T. crenulata
(ATREE, unpublished data), and the peak in these flowers occurs in the latter part of the
flowering season (fig 3.7). The bees may be tracking this nectar source in preparation for
their long migratory flights, which may be as far as 100 km or more (Koeniger and
Koeniger 1980). Pollen is crucial for brood production, which necessarily stops prior to
migration; thus, it is possible that giant honey bees temporally shift the preference of
food category according to season, and are willing to fly further in search of this food

resource than they appear to be for any of the major pollen resources.

The complex picture that this plant-pollinator landscape analysis reveals poses some
interesting questions for further exploration. One question which may be particularly
valuable in understanding how the phenomenon of generalist foraging impacts the forest
is whether the same taxa will emerge as preferred food plants if the study were replicated
across years. The site-specific preference for Syzygium in KA 98, for example, may be a
phenomenon particular to the conditions of the year it which the data were collected.

Given that the signal was the same in both the pollen analysis and the foraging range

100

analysis, the question of its generality across years is an intriguing one. One mechanism
by which the site-specific patterns may emerge could be that bees learn to find profitable
resources, and do not learn new ones until the old ones begin to wane (Barth 1991). This
bias in searching is reinforced by odor cues at the nest, visual cues in the form of floral
display, and several other factors that may be due to differences in initial conditions
between sites. However, these initial conditions may not vary over years since floral
display is related to crown sizes of trees, and learning of a floral source will depend on
the timing of its first flowering relative to other resources, which is roughly the same
every year. In combination with the overarching preferences for Catunaregam, and
influence of Terminalia on foraging distance, the site-specific results suggest that there
may be local phenomena occurring that, if they hold up across years, could have
important influences on forest community structure and seed production via differential

patches of pollination associated with certain locations.

Another avenue for study is whether the relationships between plant taxa based on the
timing of their flowering and relative importance in the bees’ diet is a general
phenomenon in different parts of the range of giant honey bees. Is it often the case, for
example, that early season resources such as Schleichera are relatively rare in the forest,
and overused only for a short duration despite the low level presence of other taxa at the
same time? Is the relationship between giant honey bees and Grewia an opportunistic
one such that the correlations of opposite sign seen in the two locations actually do
indicate the flexibility of their foraging movements? Are there plant taxa in other parts of

tropical Asia which are not used as pollen resources but seem to drive foraging distances?

101

Why is Catunaregam not an influence in predicting the flight range, given the

overabundance of its pollen in the bees’ diet?

Spatio-temporal heterogeneity of resources is the rule in ecological systems, not the
exception, although the problem of understanding the large scale patterns and processes
associated with such heterogeneity has remained largely intractable until recently.
However, the use of new, high technology tools such as GIS, along with old, low
technology tools, such as observations of bee behavior, flowering phenology, and
vegetation transects holds great promise in uncovering new patterns in the ecology of
plant-animal relationships. Although many questions remain incompletely answered in
the analyses presented here, they have provided an unprecedented time series picture of
the landscape scale correlations between trees and their pollinators. Equally importantly,
I believe they have demonstrated the power of such approaches to detect otherwise

unobservable phenomena.

102

 

Table 4.1. Results of stepwise regression for BK 98 site. These three out of the original
six taxa emerged as being the most valuable predictors of flight range. Here, “sig prob”
is the theoretical significance probability, because in stepwise regression it cannot be
determined absolutely. Cp is Mallow’s criterion, and “p” is the number of parameters.
Based on the results here, Canthium was removed from the model before performing
multiple regression. (See text for details.)

 

 

 

 

 

 

 

 

 

PARAMETER “SIG PROB” SEQ ss R2 le p
Terminalia .0008 16372070 .4221 6.3239 2
Grewia .1172 2648686 .4904 5.3314 3
Schleichera .1 172 2453654 .5537 4.5591 4
Canthium .1697 1767033 .5992 4.5627 5

 

 

 

 

Table 4.2. Results of stepwise regression for KA 98 site. These three out of the
original six taxa emerged as being the most valuable predictors of flight range. Here,
“sig prob” is the theoretical significance probability, because in stepwise regression it
cannot be determined absolutely. Cp is Mallow’s criterion, and “p” is the number of
parameters. Based on the results here, Schleichera was removed from the model before
performing multiple regression. (See text for details.)

 

 

 

 

 

 

 

 

 

 

PARAMETER “SIG PROB” SEQ SS R Cp p
Terminalia .0067 9448563 .3967 12.873 2
Syzygium .2213 1506029 .4599 12.161 3
Grewia .0056 5 89394.6 .7074 3.5488 4
Schleichera .2132 87724.71 .7442 3.9692 5

 

 

103

 

 

 

 

 

 

 

: mm 2
:88. u 3 8.8 88063 6:: 02.; e 60.2.: 0:05:50:
A035 3 030..-..—

38. n e an: a: We 80 :3 <0: samba: \ ea 0:: 68086360
630. n :V 00.2 Gwvoflnc 3.0 A030.» .5 050...: 030:0
K. mm. N: .0062

808. u 3 02: €883 £0 03.: e 600..-: .6062

 

00.080 6:62.850:
+ 80.023 5500:.
+ 80300.0 6.565 + :03. n :

3&3 WV 0:055:05
+ 80.833 5000:0300. +
$0.38 SEED + :03: n a

550.50 :0_wm0._w0m

 

 

wag

 

w: v:—

 

 

 

00 <0: 05 Ma 0:: 80... 5: 80802 2:035 .8 838: .2. 2.5

104

Table 4.4. G-tests for KA 98 site and BK 98 site comparing spatio-temporally explicit
forest floral composition to pollen diet composition. DF = 5, G critical = 11.07 (p<.05);
G critical = 20.515 (p<.001).

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

BK 98

[POLLEN TYPE WEEK 1 WEEK2 WEEK3 WEEK4 WEEK 5|
Eatunaregam 45.64 217.24 105.13 47.75 90.43|
[Canthium -3045 -102 -2309 -1494 0|

Erewia -9.12 —4.78 -2523 -30.68 -13.23|

ISyzygium .435 -500 78.05 47.31 51.15]
ISchleichera 413.58 -19.72 -7.03 -1293 -5.96I
ITerminalia -15732 -62.61 -79.28 -229 -1424]

lG-score 515.97 248.23 97.10 -120.80 11.71|

KA 98 -
IBOLLEN TYPE WEEK 1 WEEKz WEEK3 WEEK4 WEEKS WEEK6|
rCamnaregam 166.94 -14.1 129.82 -2373 36.99 -2473
Eanthium 20.68 0.64 22.65 -121 0 -0.08]
IGrewia 0 0 4.30 -1075 -26.74 -19.27|
[Syzygium 0 -1.32 8.26 155.05 88.77 125.95]
[Schleichera 403.21 3653.78 520.84 384.58 54.45 26.60]
[Terminalia -56.44 -15.84 -5377 34.45 -15.60 43.33
[G-score 1068.76 7246.3 1247.01 1076.76 275.74 303.54l

 

 

 

 

 

 

 

105

.Mm m_ .532 05 28 Jam

fl couawawma ~25: och .8528 2:5me .5“ 35a wimp—8 2a £38.? 220:5 firm wE—qfiam
sonSowo> 05 30:8 93 figs 820523 95 0.3 88 38% bis—om .bzfiazga 5:8 madcap—oi
8365 5% .8 83% Exam— daofimu Euwwggéuu .58. 8823 x09: war—030: BEBEBE
>283 9:0 Euros ma Mm “Ea ma <M .88. am:— ow88 mo 0; 333 no oEEaxm .30 0.5me

 

 

 

OOOOOOOOOO

IIIIII
......

..........

“at.“
.mmmmm.
m
6.."
”HHHHHH” .... K
{if

 

 

106

CHAPTER 5: NESTING AGGREGATIONS, POLLINATION, AND FOREST

STRUCTURE

INTRODUCTION

The field of ecology continues to address the question of what role individual species
play in overall ecosystem function, an approach that will be increasingly relevant in
developing new approaches to habitat and biodiversity conservation (Balvanera et al.
2001). It is clear that some species are particularly influential in structuring aspects of
their environment beyond their role in the trophic web. Such species have been called
“ecosystem engineers” (Lawton 1994) whose presence provide structures (such as beaver
dams) that affect the rest of the biotic community. In this chapter I begin to explore a
hypothesis built on the results of the previous analyses which examines the role of Apis

dorsata in engineering overall forest structure of the areas in which it forages.

Apis dorsata colonies nest as singletons scattered throughout the forest, but in many areas
of its range, giant honey bees also aggregate their nests in groups that may number 100 or
even more colonies on a single tree or rock cliff (figure 1.1b) (Morse 1969; Deodikar,
1977). Colony size is on average 50,000 individuals, and an estimated 20% of workers
are at any given time engaged as foragers (Morse 1969). A. mellifera foragers make an
average of 10-15 trips per day, though the number may range up to 150 (Winston 1987)
and although it is likely that A. dorsata foragers make fewer trips per day (Dyer and
Seeley 1991) individuals clearly can and do go on multiple foraging bouts per individual

when presented with sugar water feeders (see chapter 2). Thus, a conservative estimate

107

of five trips per forager per day results in approximately 50,000 foraging trips per day
from each colony. An aggregation of 100 colonies therefore results in millions of
potential pollination events every day, all radiating out from a single point in the forest.
Such an aggregation locale then could be considered an “epicenter” of pollination.
Moreover, these aggregation locations are known to recur from year to year. That is,
despite their migration during the rainy season, giant honey bee colonies always
recolonize the same locations in large numbers. In BRT, the trees on which there were
nest aggregations were 20 m or higher, and presumably quite old. The cliff faces have
potentially been recolonized for hundreds or thousands of bee generations. This nesting
behavior, due to the sheer numbers of pollinators it places in a single location, combined
with the recurrence of colonization by large numbers of colonies in a single site may have
some extremely important implications for forest spatial structure, both at the community

composition level as well as population genetic level.

The forage mapping analyses showed that 90% of the time, bees forage within a distance
of about 2 km from the nest. Preferred plants such as C. spinosa that occur in a forest
with radius of 2 km centered around the nest aggregation should receive higher numbers
of floral visitors than regions that are at further distances from the central nest
aggregation. If visitation rates are directly related to fruit and seed production, the patch
centered around the nest site aggregation with radius of 2 km could be a source patch for
fruits of preferred bee plants. Since fruit dispersal is assumed to be leptokurtic with
distance from parent tree (Howe 1990), over several generations of dispersal and

regeneration close to fruit bearing trees, one might hypothesize the development of

108

“contours” of abundance of C. spinosa, centered around the stable honey bee aggregation

sites.

There are several assumptions inherent in the above scenario, none of which have been
tested for Catunaregam spinosa, or for any other species. First, it assumes that fruit set
in C. spinosa is pollination limited as opposed to being limited by other factors such as
resource levels. This assumption can only be tested experimentally over a number of
years; however, it may not be inconceivable that in dioeceous species, in which typically
the number of female flowers is far lower than the number of male flowers (Bawa and
Opler 1975), pollination may be limited by visitation rates to the females since the floral
display is much smaller and less conspicuous. The hypothesis here is that the bees’
foraging range and nesting behavior will result in a landscape mosaic of pollinator

limitation, and that this will result in variation in fruit production across the landscape.

A second assumption is that the scale of fruit dispersal will occur over a spatial scale that
does not encompass the foraging region of another nest site aggregation. The fleshy
fruits of this tree are eaten by large vertebrates, especially the forest bison, or gaur (Bos
gaurus) (M. Ketha, pers comm.). It is not known whether the fruits need to pass through
a mammalian digestive tract for germination to occur or whether those that fall from the
parent tree also can germinate. However, gaur range over distances of 2-3 miles a day
(Schaller 1967), a scale of movements that could enable dispersal of seeds across the

boundaries of the foraging ranges of adjacent bee aggregations (pers obs). Detailed

109

studies on dispersal by gaur would be crucial in setting expectations for spatial structure

of this plant species.

Before conjecturing that the spatial distribution of bee plants is controlled indirectly by
the foraging limitations of bees themselves, the first step is to quantify whether honey bee
and other insect visitation rates to trees within a 2 km radius from the nest sites differs
from visitation rates to trees farther away. By measuring visitation by other species of
Apis and other insect taxa, I begin to explore also whether there are differential zones of
competition for floral resources such that in areas that may experience relatively lower

densities of A. dorsata foragers, other insects are more frequent floral visitors.

METHODS

FIELD METHODS

Ninety percent of all A. dorsata foraging flights occur within 2.2 km from the nest site
(chapter 2). Therefore, I located Catunaregam spinosa male individuals in bud stage in
locations within two kilometers from any nest aggregation, here called “proximal”sites,
and trees that were beyond two kilometers distance from all nest aggregations, called
“distant” sites. There may have been, however, single colonies nesting close to focal trees
at either proximal or distant sites. The probablity of such singletons is assumed to be
equal for any area of the forest. In total, there were eight trees in each distance category.
At each site I chose two trees in approximately the same stage of flowering, with

approximately the same crown size. The purpose of this was to minimize differences in

110

floral display between replicates, though because it was done by visual estimation only, it

not strictly standardize the patch size.

At each tree, my field assistant and I chose three branches with a varying number of
flowers on them, and counted the number of flowers. For five minutes every half hour,
we watched each branch of flowers and tallied the number of visits (floral landings by the
same or different individuals) by A. dorsata, A. cerana, A. florea, Trigona, lepidopterans,
and “other” insects. As the period during 1100-1500 has very little insect activity, we

counted visits from 0600 until 1100 and again from 1500-1800.

STATISTICAL ANALYSIS

There was an abundance of zeros in all data sets, which is problematic for statistical
analysis. In order to try and reduce the number of zeros, I pooled all visits per day to each
branch, and then computed the number average of visits per minute per flower on the
branch, or visitation rate per flower. This resulted in each branch being treated separately
as a replicate unit, so there were 24 units per distance category. I did this for all
categories of visitors. The data were still quite non-normally distributed; therefore, I
analyzed them using a Wilcoxon’s rank sums test, which is the non-parametric equivalent
of a t-test. The independent variable here was distance from an aggregation (distant vs.

proximal), and the dependent variable was visitation rate per flower.

lll

RESULTS

The results of two-way unpaired comparisons between honey bee and other insect
visitation to proximal sites vs. distant sites did not yield any conclusive results. All
groups of insects showed higher visitation to the proximal sites (table 5.1). When
visitation by A. dorsata to distant and proximal sites were compared, the result of the
non-parametric test was not statistically significant at the p<.05 level. The p-value of 0.08
suggests that there may be a biologically significant trend worth exploring further. A.
cerana showed virtually no difference in visitation rates to areas that were at different
distances from A. dorsata nest sites, whereas A. florea seemed to visit those areas near to
nest aggregations at higher rates per flower. All other insects also visited the proximal
sites more often than the distant sites. Thus, it appears that other insects, including the

other two species of Apis, do not avoid areas that may be dominated by A. dorsata.

DISCUSSION

Visitation by giant honey bees was not Significantly higher in areas of the forest that were
within a two kilometer foraging distance from nest aggregations than in areas further than
two kilometers. Interestingly, the response was similar across all insects except for A.

cerana in which visitation rate was basically equal for the two distance categories.

The rates of visitation were quite low in all cases, thus the statistical results are difficult
to interpret with much confidence. There were many counts of zero in the data. This
poses a challenge for statistical analysis, a challenge that is further compounded by the

fact that some of the zeros are hypothesized to have biological meaning; that is, at the

112

distant site, I hypothesized that there would be more zeros than at proximal sites. There
were five proximal site branches out of 24 where the average visitation rate was zero and
only six at distant sites where it was zero. Thus it seems there is not much difference in
zero visitation rates between the two sites. Other insects had even lower visitation rates
to flowers of C. spinosa than A. dorsata. It was not the case that in areas which were
further from giant honey bee nest aggregations, other insects utilized the resource more
heavily. From this result I draw two conclusions: First, A. dorsata is the most frequent
visitor to C. spinosa male flowers, and second, if indeed there is variation in visitation
rates at different distances from the aggregation, it does not result in a mosaic of different

mutualists that may be avoiding competition.

The fact that visitation rates for proximal and distant flowers verged on being
significantly different in the hypothesized direction for A. dorsata suggests that this
hypothesis is worthy of further exploration. The fact that all taxa exhibited higher rates
to the proximal sites may mean that those trees chosen were simply more attractive to all
foraging insects. In order to fully address the question, the study would need a larger
sample size, would perhaps need to incorporate more than two distance categories, and
would need to equalize the floral display in a systematic manner, either statistically or by
removal of flowers from the focal and neighboring trees such that all sites were of equal
attractiveness. It would also need to stratify the sample sizes at various distances, thus
controlling for the fact that at longer radial distances, the area searched by the bees is
larger and thus any given tree within that radius has a lower probability of being visited.

This effect alone could account for the marginally significant results I obtained in the

113

Wilcoxon test. Furthermore, in light of the findings in chapter 2 that colonies do not seem
to share patches, a better design would incorporate many patches of C. spinosa near the
same aggregation being observed simultaneously, as opposed to this study, which made
simultaneous observations only two trees at a time, and those trees were close enough to
each other that they may be considered the same patch. Discovery of a patch would in
this case be dominated by one colony, regardless of its distance from the nest. The
improved design of incorporating several patches also may correct for any singleton
colonies that may have been nesting near the focal trees and whose visitation may have

swamped out any effects of distance from large aggregations.

Another approach to examining whether C. spinosa is Structured around nest site
aggregations would be to use GIS to map the fine grained distribution of adult trees
around several nest site aggregations up to distances that exceed the foraging range of the
bees, and compare them with the distribution of C. spinosa trees around singleton A.
dorsata nests. These singletons occur randomly throughout the forest at sites which are
not recolonized from year to year (pers. obs.) and therefore would not result in a contour-

like organization of trees around them.

The phenomenon which this chapter hypothesizes could have interesting ramifications for
the population genetic structure of selected tree Species. Gene flow largely within a
confined area that has a natural center, the nest site aggregation, may result in a mosaic
landscape that can actually be observed as patches of allelic diversity, and one in which

plant neighborhood size may be predicted by honey bee flight range.

114

 

Tr .
. .‘u

 

 

The implications of such a phenomenon would also have implications for conservation
applications and landscape management plans in the Asian tropics. While it is clear that
such aggregations of pollinators should be included in preserved habitat regardless, if
they actually do act to structure the forest, the bees may be truly considered both an
“umbrella” and keystone species (Simberloff 1998) whose nest sites and foraging range
delineate natural units of community and genetic spatial organization. The inclusion of a
number of these units might be a strategy for preserving a desired level of forest
diversity, and would also result in protecting areas in a manner that does not disrupt
existing movement corridors of some elements of forest community, including the large

frugivorous mammals which act as dispersers.

115

"' ‘~ month.
L

‘ Mgr...» 4—-. _.__._ —

 

 

Table 5.1. Results of Wilcoxon rank sums test of visitation rate to flowers of C. spinosa.

 

 

 

 

 

A. dorsata A. cerana A. florea other
Proximal .018958 .000186 .000395 .002430
(visits/flower/minute)
Distant .009560 .000075 .000000 .001 128
(visits/flower/minute)
Wilcoxon .0870 .9828 .0410 .0620
p-value

 

 

 

 

 

 

116

 

 

 

 

 

 

 

 

 

 

_ _ «N 2 _ 2 2 v 8828 8-2 ”88
o. N o. NN : N ”-22.2522 88m”:

song-RE

_ w o 8 _ _ 2 ON N 88828888
N o 8 _ 8 : 2 _ 8 we
8-82-8 <2 8 8< :.2 8 8< 2-8 8185,. 8 82 ”N8 8 82 8-2 8-82-2 8838
b a 2 _N 8N NN 2N v 8828 8-8 ”88
m 2 8 mN 8 a 2 N. N m 8.2 i .z 8.3:
8 MN 2 mN mm 8N M: 2 N 8888
S m 2 _ 8 we

8-822 8-853 8852 8-85: 8-85m 8-82-8N 8-82-2 N 8-82-: 885838:
83:28 nm-m~ ”8.5

2 2 m 8 2 ”N N_ _ m m N188 .z 8.8”:
MN 2 8 mm 9. NN 8N : N 88288 to
: 2 8 2 2 cm 2 8 _ 8 Me
8-82-8 8-858 8-852 8-85N 8-852 8-82-8 8-82-2 8-82-2 35—2288:
8828 2-8 828:8

2 2 2 Nm NN 2N 8m : m m 8.8”: .2 2.80:
2 2 _ _ on NM 8 8N 8N N 86888 S
w 2 m cm 9. cm 9. 9. u z _ 8 <0:
8-828 885% 8.852 8858 8-82-: 8-82-8 8-82-2 8-82-2 :85.
a N e m .. m N _ >20-50 ”=5

mh<n .364 Ewan!— OZ—ngwmmmco DZ< v—Hflg

 

.28. :80 “a x33 hog 8:28 hon A3280.— 8058 Mo 83:55 mofim 0388
98 .828 529.85 new: 838:: x83 mcficommufioo :05 85 5:00:00 33 “o 8:6 809280885 38. minus: ems—om

 

— X—A—Zm—AE<

117

 

 

 

 

 

 

 

83:28 in ooh
8N 8 2 2 8 m 2.88 .z 828“:
X «N 8 N2 N 882885
2 2 2 _ a8 02
88-8882 <2 88.8: 8.82-8 88.82-: Sz :26 gym

8 n
v m 8 8 8 8828 mm-nN ”88
N. 8 8 2 2 3 m m 2.898 .2 88.88”:
2 2 MN 8N 8N N. N 382885
3 N 2 2 N8 8N _ 38 22
88-858 88-852 88-85N 8.82.8 8.82.2 8.822 3.8553
8828 8888 628-28
2N v 8 N n 8 m $.88 .z 2.88”:
«N 8 N 2 2 NN N 8828888
8. 2 2 E 8 8N" z _ $8 5:
88.858 8.852 88.82; 8 8882-8 8.82.2 88-828 :85—
e m w n N _ >238 82.2

28¢;
.22 8555-.— czEszmmzoo nz< 8.8.85

8.283 a 596%?

118

$2 5

 

awe—ocean

 

 

 

<2 <2 5< 8 <2 a< : a< 2 a< m a: R a: 2 a: 2 was
E< ~

<2 <2 <2 <2 <2 a< «-2 a< z-a a< w-N saw on a: 2-: <2 82 cm
. < N

, <2 <2 <2 <2 a< «3 <2 a< 3 -32 R a: 8-2 a: eé 32 mm

~32 n 5< -

<2 :2 «3 <2 -a< mu a< 8-3 a< 3: a< 2-» -32 a .32 5a a: 3: 3: S-a 32 §
.a< v

<2 <2 <2 <2 <2 a< 2-: <2 a< 26 .32 a a: 5-8 a: 8.2 82 8
.32 N E< n

<2 34 2a -a< on a< 2-2 <2 a< 2-2 <2 a< :¢ .32 on :2 3mm <2 82 8

<2 » <2 <2 a< «2: <2 a< 2.” <2 5< 3 <2 32 2-2 <2 82 gm

«2 m
:2 NE -a< an <2 5< 3: <2 a< 2-5 <2 5< 3 a: 3-2 a: «S: a: :-n 8.: <3
a N. G e 3 m 5. v n u _ Ea
Mum?

 

 

.mcocatomno awe—ocean Ba 8353 5:8 be 339

= Eggs-Q

119

APPENDIX III
Composition of phenology transect.

AMILY ENUS AND SPECIES INDIVIDUALS
AME N TRANSECT
mosaceae sinuata Si
mosaccae izzia odoratissima Sele
bretaceae ' us 'olia "a
' ' racemosa e
ischofiaceac ' ' ' eelalu
ceiba uru
'accac ridelia retusa ironnc
'aceae uchanania lanzan '
'aceae Canthium dicoccum Ambe
'accae Canthium ' rum
'n 'aceae arborea lli
'naceae Cassia
'aceae Cordia le
abaccae ia lanceolaria ul
abaceae ia 'olia ite
ros montana ti
isia laucescens rianambu
serratus ikkilu
'accae ica elli
teculiaceae '
icus

urseraccae

Glochidion
crbenaccae Gmelina arborea
iliaceae Grewia hirsuta

vaceae
rstromia
allotus
'aceac eliosma innata
leaccae Olea landul era
'lionaceae enia 00 enesis
ersea macranta
abaceae
i 'aceae rmachera
accae Catunare
'caceae rma
' ra oleosa

v—NoooxooOx—Nmammmomamaoo—wwqahAMANucn—Nu—zm—

 

120

Appendix III (cont’d)

 

 

 

 

 

 

 

 

 

 

 

 

 

I FAMILY GENUS AND SPECIES LOCAL N INDIVIDUALS ON
NAME TRANSECT
IOleaceae Schrebera sweitenoides Game 6
[Sterculiaceae Sterculia villosa Chouve 5
lBignoniaceae Stereospermum personatum Padure 2
Myrtaceac Syzygium cuminii Nerale 6
Edyrtaceae Syzygium malabaricum Neeranchi 5
ICombretaccae Terminalia bellerica Tare 5
[Combretaceae Terminalia chebula Arale 6
[Combretaceae Terminalia crenulata Matti 13
Eombretaceae Terminalia paniculata Uluge 9
meliaceae Trichilia connoroides Kanagojjali 3
Eaprifoliaceae Viburnum punctatum Thonde 3
[Rubiaceae Wendlandia thyrsoidea Koli 1

 

 

 

 

121

 

APPENDIX IV
Tree species in BRT which belong to the genera of pollens found in fecal samples.

AMILY AND SPECIES NAME
lanceolaria ul
olia
serratus
tuberculotus
iliaceae Grewia hirsuta
'aceae Grewia ' olia
Terminalia bellerica
Terminalia chebula
Terminalia crenulata
erstromia
cuminii erale
malabaricum eeranchi
ndaceae ra oleosa
'ndaceae
'aceae Cambium
'aceae Canthium dicoccum
ubiaceae Catunare
leaceae Olea era

 

122

 

APPENDIX V

Plant families and genera whose pollens were found in low frequencies in feces, and put
into the category designated as “other” for pollen analysis.

AMILY
thaceae
aceae
olorrhena

urseraceae
'naceae
'naceae
naceae

'aceae
'aceae Glochidion
aceae allotus
aceae
'aceae
mosaceae

ubiaceae ra
ubiaceae
'aceae eliosma

Trema
erbenaceae Clerodendrum

 

123

 

REFERENCES

Abrol, D. P. and R. P. Kapil (1996). Studies on abundance and importance of insect
pollinators for oilseed production. Journal of Insect Science 9(2): 172-174.

Aizen, M. A. and P. Feinsinger (1994). Forest fragmentation, pollination, and plant
reproduction in a chaco dry forest, Argentina. Ecology 75(2): 330-351.

Allen-Warden, G., P. Bernhard, R. Bitner, A. Burquez, S. Buchmann and et. al (1998).
The potential consequences of pollinator declines on the conservation of
biodiversity and stability of food crop yields. Conservation Biology 12(1): 8-17. I

Anupama, K., B. R. Ramesh and R. Bonnefille (1999). The modern pollen rain from the
Biligirirangan hills - Melagiri hills of southern Eastern Ghats, India. Review of
Palaeobotany and Palynology.

Appanah, S. (1981). Pollination in Malaysian primary forests. The Malaysian Forester
44(1): 37-42.

 

Balvanera, P., G. C. Daily, P. R. Ehrlich, T. H. Ricketts, S.-A. Bailey, S. Kark, C.
Kremen and H. Pereira (2001). Conserving Biodiversity and Ecosystem Services.
Science 291(5511): 2047.

Barth, F. G. (1991). Insects and Flowers: the Biology of a Partnership. Princeton, New
Jersey, Princeton University Press.

Bawa, K. and P. A. Opler (1975). Dioecism in tropical forest trees. Evolution 29: 167-
179.

Bawa, K. S. (1990). Plant-pollinator interactions in tropical lowland rain forest. Annual
Review of Ecology and Systematics 21: 399-422.

Bronstein, J. (1995). The plant-pollinator landscape, in Mosaic Landscapes and
Ecological Processes. L. Hanson, L. Fahri g and G. Merriam (ed.). London,
Chapman and Hall.

Buchmann, S. L. and G. Nabhan (1996). The Forgotten Pollinators. Washington DC,
Island Press/Shearwater Books.

Chittka, L., A. Shmida, M. Vorobyev and R. Menzel (1993). Bee color vision--the
optimal system for the discrimination of flower colors?, in Sensory Systems of
Arthropods. K. Wiese, S. G. Gribakin, A. V. Popov and G. Renniger (ed.). Berlin,
Birkhauser.

124

Crane, E. (1991). Apis species of tropical Asia as pollinators, and some rearing methods
for them. Acta Horticulturae 288: 29-48.

Crane, E. (1999). The World History of Beekeeping and Honey Hunting. New York,
Routledge.

Danielson, B. J. (1991). Communities in a landscape: the infleunce of habitat
heterogeneity on the interactions between species. American Naturalist 138(5):
1 105-1 120.

Daumer, K. (1958). Blumenfarben, wie sie die bienen sehen. Z. vergl. Physiol. 41: 49-
110. n

Dayanandan, S., J. Dole, K. Bawa and R. Kesseli (1999). Population structure delineated
with microsatellite markers in fragmented populations of a tropical tree, Carapa
guianensis (Meliaceae). Molecular Ecology. Oct. 8(10): 1585-1592.

Deodikar, G. B., A. I. Ghatge, R. P. Phadke, D. B. Mahindre, K. K. Kshirsagar, K. S.
Muvel and C. V. Thakar (1977). Nesting behaviour of Indian honeybees III.
Nesting behaviour of Apis dorsata Fab. Indian Bee Journal 39: 1-12.

 

Devy, M. S. (1998). Pollination of canopy and sub-canopy trees by social bees in a wet
evergreen forest of south Western Ghats, India. Ph. D. dissertation, Department of
Botany. Chennai, India, Madras Christian College.

Dick, C. (2001). Genetic rescue of remnant tropical trees by an alien pollinator.
Proceedings of the Royal Scoiety of London 268: 2391-2396.

Dramstad, W. E. (1996). Do bumblebees (Hymenoptera: Apidae) really forage close to
their nests? Journal of Insect Behavior 9(2): 163-182.

Dukas, R. and L. A. Real (1991). Learning foraging tasks by bees: a comparison between
social and solitary species. Animal Behavior 42: 269-276.

Dyer, F. C. (1985). Nocturnal orientation by the Asian honey bee, Apis dorsata. Animal
Behaviour 33(3): 769-774.

Dyer, F. C. and T. D. Seeley (1987). Interspecific comparisons of endothermy in honey-
bees (Apis): Deviations from the expected size-related patterns. Journal Of
Experimental Biology 127: 1-26.

Dyer, F. C. and T. D. Seeley (1991). Dance dialects and foraging range in three Asian
honey bee species. Behavioral Ecology And Sociobiology 28(4): 227-234.

Dyer, F. C. and T. D. Seeley (1991). Nesting behavior and the evolution of worker tempo
in four honey bee species. Ecology 72(1): 156-170.

125

Dyer, F. C. and T. D. Seeley (1994). Colony migration in the tropical honey bee Apis
dorsata F. (Hymenoptera: Apidae). Insectes Sociaux 41(2): 129-140.

Environmental Systems Research Institute, I. E. (1996). Arc View GIS and Spatial
Analyst. Redlands, CA, ESRI, Inc.

Faegri, K. and J. Iverson (1964). Textbook of Pollen Analysis. New York, Hafner
Publishing.

Faegri, K. and L. van der Pijl (1979). The Principles of Pollination Ecology. Oxford,
Pergamon Press.

Frankie, G. W., H. G. Baker and P. A. Opler (1973). Tropical plant phenology:
Applications for studies in community ecology, in Phenology and Seasonal
Modelling. H. Lieth (ed.). Berlin, Springer-Verlag.

Free, J. B. (1970). Effect of flower shapes and nectar guides on the behavior of foraging
honeybees. Behaviour 37: 269-285.

Frisch, K. v. (1914). Der farbensin und formensinn der biene. Zool. Jahrb. Abt. allg.
Physiol. Tiere 35: 1-182.

Frisch, K. v. (1919). Uber der Bienen und seine blutenbiologische Bedeutung. Zool. Jb.
Physiol. 37: 1-238.

Frisch, K. v. (1967). The Dance language and Orientation of Bees. Cambridge, Harvard
University Press.

Gadgil, M. (1992). This F issured Land: and Ecological History of India. Delhi, Oxford
University Press.

Gentry, A. H. (1974). Flowering phenology and diversity in tropical Bignoniaceae.
Biotropica 6(1): 64-68.

Giurfa, M., J. Nunez, L. Chittka and R. Menzel (1995). Colour preferences of flower-
naive honeybees. Journal of Comparative Physiology A177: 247-259.

Hall, P., S. Walker and K. Bawa (1996). Effect of forest fragmentation on genetic
diversity and mating system in a tropical tree, Pithecellobium elegans.
Conservation Biology 10(3): 757-768.

Heinrich, B. (1979). Bumblebee Economics. Cambridge, MA, Harvard University Press.

Howe, H. (1990). Seed dispersal by birds and mammals: implications for seedling

demography, in Reproductive Ecology of Tropical Forest Plants. K. Bawa and M.
Hadley (ed.). Paris, UNESCO and Parthenon Publishing Group. 7.

126

 

Huang, T. S. (1972). Pollen flora of Taiwan, National Taiwan University, Botany
Department Press.

Ims, R. (1995). Movement patterns related to spatial structures. London, Chapman and
Hall.

J anzen, D. H. (1971). Euglossine bees as long-distance pollinators of tropical plants.
Science 171: 203-205.

Jhansi, P., T. P. Kalpana and C. G. K. Ramanujam (1991). Pollen analysis of rock bee
summer honeys from the Prakasam district of the Andhra Pradesh, India. Journal
of Apicultural Research 30(1): 33-40.

Johnson, S. D. and K. E. Steiner (2000). Generalization versus specialization in plant
pollination systems. Trends in Ecology and Evolution 15(4): 140-143.

Kareiva, P. (1982). Experimental and mathematical analyses of herbivore movement:
quantifying the influence of plant spacing and quality on foraging discrimination.
Ecological Monographs 52(261-282).

Kastberger, G., G. Raspotnig, S. Biswas and O. Winder (1998). Evidence of Nasonov
scenting in colony defence of the giant honeybee Apis dorsata. Ethology 104(1):
27-37.

Kastberger, G., O. Winder, T. Hoetzl and G. Raspotni g (1996). Behavioural features of a
periodic form of massed flight activity in the giant honeybee Apis dorsata.
Apidologie 27(5): 381-395.

Keams, C. and D. W. Inouye (1993). Techniques for Pollination Biologists. Niwot, CO,
University of Colorado Press.

Keams, C., D. W. Inouye and N. Waser (1998). Endangered mutualisms: the
conservation of plant-pollinator interactions. Annual Review of Ecology and
Systematics 29: 83-112.

Kiew, R. (1997). Analysis of pollen from combs of Apis dorsata in peninsular Malaysia,
in Proceedings of the International Conference on Tropical Bees and the
Environment. M. Mardan, A. Sipat, K. M. Yusoff, R. Kiew and M. M. Abdullah
(ed.). Serdang, Malaysia, Beenet Asia and Southbound.

Koeniger, G., M. Mardan and F. Ruttner (1990). Male reproductive organs of Apis
dorsata. Apidologie 21(2): 161-164.

Koeniger, N. and G. Koeniger (1980). Observations and experiments on migration and
dance communication of Apis dorsata in Sri Lanka. Journal of Apicultural
Research 19(1): 21-34.

127

 

Lawton, J. H. (1994). What do species do in ecosystems? Oikos 71: 367-374.

Lima, S. and P. A. Zollner (1996). Towards a behavoral ecology of ecological
landscapes. Trends in Ecology and Evolution 11(3): 131-135.

Lindauer, M. (1957). Communication among the honeybees and stingless bees of India.
Bee World 38(1): 3-14, 34-39.

Mardan, M. and P. G. Kevan (1989). Honeybees and 'yellow rain'. Nature 341: 191.

May, R. (1975). Patterns of species abundance and diversity, in Ecology and Evolution of
Communities. M. L. Cody and J. Diamond (ed.). Cambridge MA, Belknap Press.

Michener, C. D. (1979). Biogeography of the bees. Annals of the Missouri Botanical
Garden 66(3): 277-347.

Momose, K., T. Yumoto, T. Nagamitsu, M. Kato, H. Nagarnasu, S. Sakai, R. Harrison, T.
Itioka, A. Hamid and T. Inoue (1998). Pollination biology in a lowland
dipterocarp forest in Sarawak, Malaysia. 1. Charactersitics of the plant-pollinator
community in a lowland dipterocarp forest. American Journal of Botany 85(10):
1477-1501.

Morse, R. A. (1969). Apis dorsata in the Philippines. Monograph of the Philippine
Association of Entomologists 1: 1-96.

Murali, K. S., U. Shankar, R. U. Shaanker, K. N. Bawa and G. K. S. (1996). Extraction of
non-timber forest products in the forests of Biligiri Rangan Hills, India. 2. Impact
of NTFP extraction on regeneration, population structure, and species
composition. Economic Botany 50(3): 252-269.

Murali, K. S. and R. Sukumar (1994). Reproductive phenology of a tropical dry forest in
Mudumalai, southern India. Journal of Ecology 82: 759-767.

Murawski, D. A. and K. S. Bawa (1994). Genetic structure and mating system of
Stemonoporus oblongifolius (Dipterocarpaceae) in Sri Lanka. American Journal
of Botany 81(2): 155-160.

Myers, N., R. A. Mittermeier, C. G. Mittermeier, G. A. B. da Fonseca and J. Kent (2000).
Biodiversity hotspots for conservation priorities. Nature London. Feb. 403(6772):
853-858.

Nason, J. D. and J. L. Hamrick (1997). Reproductive and genetic consequences of forest

fragmentation: Two case studies of Neotropical canopy trees. Journal of Heredity
88(4): 264-276.

128

 

Nason, J. D., E. A. Herre and J. L. Hamrick (1998). The breeding structure of a tropical
keystone plant resource. Nature 391(6668): 685-687.

N ayar, T. S. (1990). Pollen flora of Maharashtra state, India. New Delhi, Today and
Tomorrow's Printers and Publishers.

Nguyen, Q. T. P. H. C., H. T. Pham and V. Mulder (1997). Rafter beekeeping with Apis
dorsata: Some factors affecting the occupation of rafters by bees. Journal of
Apicultural Research 36(1): 49-54.

Osborne, J. L., S. J. Clark, R. J. Morris, 1. H. Williams, J. R. Riley, A. D. Smith, D. R.
Reynolds and A. S. Edwards (1999). A landscape-scale study of bumble bee
foraging range and constancy, using harmonic radar. Journal of Applied Ecology.
Aug. 36(4): 519-533.

Pleasants, J. M. and M. Zimmerman (1979). Patchiness inthe dispersion of nectar
resources: evidence for hot and cold spots. Oecologia 41: 283-288.

Power, M., D. Tilman, J. Estes, B. Menge, W. J. Bond, L. S. Mills, G. C. Daily, J. C.
Castilla, J. Lubchenco and R. T. Paine (1996). Challenges in the quest for
keystones. Bioscience 46(8): 609-620.

Punchihewa, R. W. K., N. Koeniger, P. G. Kevan and R. M. Gadawski (1985).
Observations on the dance communication and natural foraging ranges of Apis
cerana, Apis dorsata and Apisflorea in Sri Lanka. Journal of Apicultural
Research 24(3): 168-175.

Pyke, G. H. (1978). Optimal foraging: movement patterns of bumblebees between
inflorescences. Theoretical Population Biology 13: 72-98.

Ramesh, B. R. (1989). Evergreen forests of the Biligirirangan Hills, South India: ecology,
structure, and floristic composition. Ph.D. dissertation, Dept. of Botany. Madras,
University of Madras.

Ramesh, B. R., S. Menon and K. Bawa (1998). Vegetation map of the Biligiri
Rangaswamy Temple Wildlife Sanctuary, Kamataka, India, French Institute of
Pondicherry & Ashoka Trust for Research in Ecology and Environment.

Roubik, D. W. (1990). Niche pre-emption in tropical bee communities: A comparison of
Neotropical and Malesian faunas, in Natural History of Social Wasps and Bees in
Equatorial Sumatra. S. Sakagami, R. Ohgushi and D. W. Roubik (ed.). Sapporo,
Hokkaido University Press.

129

 

Roubik, D. W. (1993). Loose niches in tropical communities: why are there so many trees
and so few bees?, in Eflects of Resource Distribution on Animal-Plant
Interactions. M. D. Hunter, T. Ohgushi and P. W. Price (ed.). Orlando, Academic
Press.

Roubik, D. W. (2000). The foraging and potential outcrossing pollination ranges of
African honey bees (Apiforrnes: Apidae; Apini) in Congo forest. Journal of the
Kansas Entomological Society 72(4): 394-401.

Roubik, D. W. and M. Aluja (1983). Flight ranges of Melipona and Trigona in tropical
forest. Journal of the Kansas Entomological Society 56(2): 217-222.

Rudrappa, S. (1996). Indigenous Knowledge among the Soligas at Biligri Rangana
Swami Temple Wildlife Sanctuary (BRT), Southern India. M.S. thesis, dept of
Environmental Sciences. Madison, University of Wisconsin.

Ruttner, F. (1988). Biogeography and taxonomy of honeybees. Berlin, Springer-Verlag.

 

Sakai, 3., K. Momose, Y. Takakazu, T. Nagamitsu, H. Nagamasu, A. Hamid and T.
Nakashizuka (1999). Plant reproductive phenology over four years including an

episode of general flowering in a lowland Dipterocarp forest, Sarawak, Malaysia.
American Journal of Botany 86(10): 1414-1436.

SAS Institute (1999). JMP Statistical Discovery Software. Cary, NC, SAS Institute.

Saville, N. M., W. E. Dramstad, G. L. A. Fry and S. A. Corbet (1997). Bumblebee
movement in a fragmented agricultural landscape. Agriculture Ecosystems and
Environment 61(2—3): 145-154.

Schaller, G. (1967). The Deer and the Tiger: a Study of Wildlife in India. Chicago,
University of Chicago Press.

Schmid-Hempel, P. (1984). The importance of handling time for the flight directionality
in bees. Behavioral Ecology and Sociobiology 15: 303-309.

Schneider, S. S. (1989). Spatial foraging patterns of the African honey bee, Apis mellifera
scutellata. Journal of Insect Behavior 2(4): 505-522.

Schneider, S. S. and L. C. McNally (1992a). Factors influencing seasonal absconding in
colonies of the African honey bee, Apis mellifera scutellata. Insectes Sociaux
39(4): 403-423.

Schneider, S. S. and L. C. McNally (1992b). Seasonal patterns of foraging activity in

colonies of the African honey bee, Apis mellifera scutellata, in Africa. Insectes
Sociaux 39(2): 181-193.

130

Seeley, T. D., S. Camazine and J. Sneyd (1991). Collective decision-making in honey
bees: How colonies choose among nectar sources. Behavioral Ecology and
Sociobiology 28(4): 277-290.

Seeley, T. D., R. H. Seeley and P. Akaratanakul (1982). Colony defense strategies of the
honeybees in Thailand. Ecological Monographs 52(1): 43-64.

Shankar, U., K. S. Murali, R. U. Shaanker, K. N. Ganeshaiah and K. S. Bawa (1998).
Extraction of non-timber products in the forests of Biligiri Rangan Hills, India: 4.

Impact on floristic diversity and population structure in a thorn scrub forest.
Economic Botany 52(3): 302-315.

Sidhu, A. S. and S. Singh (1962). Role of honeybees in cotton production. Indian Cotton
Growers Review 16(1): 18-23.

Sihag, R. C. (1986). Insect pollination increases seed production in cruciferous and
umbelliferous crops. Journal of Apicultural Research 25(2): 121-126.

Simberloff, D. (1998). Flagships, umbrellas, and keystones: Is single species
management passe inthe landscape era? Biological Conservation 83(3): 247-257.

Sinha, S. N. and S. S. Atwal (1996). Pollination requirement in sunflower hybrid seed
production: H. Effect of date of sowing and fertilizer on honeybee foraging
behaviour on parental lines. Seed Research New Delhi 24(2): 110-115.

Sokal, R. and F. J. Rohlf (1981). Biometry. New York, W.H. Freeman and Co.

Starr, C. K., P. J. Schmidt and J. O. Schmidt (1987). Nest-site preferences of the giant
honey bee, Apis dorsata (Hymenoptera: Apidae), in Borneo. Pan Pacific
Entomologist 63(1): 37-42.

Steffan-Dewenter, I. and T. Tscharntke (1999). Effects of habitat isolation on pollinator
communities and seed set. Oecologia 121: 432-440.

Stephens, D. W. and J. R. Krebs (1986). Foraging theory. Princeton, NJ, Princeton
University Press.

Tabachnick, B. and L. S. Fidell (1983). Using Multivariate Statistics. New York, Harper
and Row.

Thompson, J. (1997). Evaluating the dynamics of coevolution among geographically
structured populations. Ecology 78(6): 1619-1623.

Tissot, C., H. Chikhi and T. S. Nayar (1994). Pollen of Wet Evergreen Forests of the
Western Ghats, India, Institute Francais de Pondichery, trav. Sec. Sci. Tech.

131

 

Turchin, P. (1991). Translating foraging movements in heterogeneous environments into
the spatial distribution of foragers. Ecology 72(4): 1253-1266.

Vasanthy, G. (1976). Pollen of the South Indian hills, Institute Francais de Pondichery,
trav. Sec. Sci. Tech.

Venkatesh, G. and C. C. Reddy (1989). Rates of swarming and absconding in the giant
honeybee, Apis dorsata F. Proceedings of The Indian Academy of Sciences
Animal Sciences 98(6): 425-430.

Verma, S. K. (1987). Preliminary studies on the effect of honey bees on the yield of
greater cardamom. Indian Bee Journal 49: 25-26. !

Visscher, P. K. and T. D. Seeley (1982). Foraging strategy of honeybee colonies in a
temperate deciduous forest. Ecology 63(6): 1790-1801. -

Waddington, K. D., P. K. Visscher, T. J. Herbert and M. R. Richter (1994). Comparisons
of forager distributions from matched honey bee colonies in suburban
environments. Behavioral Ecology And Sociobiology 35: 423-429.

 

Waser, N. (1983). The adaptive nature of floral traits: ideas and evidence, in Pollination
Biology. L. A. Real (ed.), Academic Press.

Waser, N., L. Chittka, M. Price, N. Williams and J. Ollerton (1996). Generalization in
pollination systems, and why it matters. Ecology 77(4): 1043-1060.

Wehner, R. (1971). The generalization of direction stimuli in the honey bee, Apis
mellifera. Journal of Insect Physiology 17: 1579-1591.

Winston, M. (1987). Biology of the Honey Bee. Cambridge MA, Harvard University
Press.

Zimmerman, M. (1981). Optimal foraging, plant density, and the marginal value theorem.
Oecologia 49: 148-153.

132

   

IIIIIIIIIIIIIIIIIII

illllllllllllllllllllié
3 1293 023