UNIVERSITY OF MIAMI

BAT AND HUMMINGBIRD POLLINATION

OF TWO OF COLUMNAR CACTI:

EFFECTS ON FRUIT PRODUCTION AND POLLEN DISPERSAL

by

CATHERINETERESASAHLEY

A DISSERTATION

Submitted to the Faculty ofthe University ofMiami in partial fulfuillment of the requirements for the degree ofDoctor ofPhilosophy

Coral Gables, Florida

May, 1995

~ UNlVERSITY OF MIAMI

A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy

Bat and hummingbird pollination of two species of columnar cacti: Effects on fruit production and pollen dispersal

Catherine Teresa Sahley

Approved: -,.. 7f~~Hfh ~@ Dr. Tarek M. Khalil Professor of Biology Dean of the Graduate School Chairperson of the Dissertation Cornmittee (J\~l \ - \ Dr\J(eith Wáddington Dr. David Janos <:»: Professor ofBiology Associate Professor ofBiology

~c.~ Dr. Carol Horvitz qf.' j~mes riarnnck Associate Professor ofBiology Professor ofBotany •••• ~j :;-:tf~~,-. ¡,;.~-,

SAHLEY,CATHERINETERESA (Ph.D., Biology)

Bat and hummingbird pollination oftwo species of columnar cacti: Effects on fruit production and pollen dispersal.

(May 1995)

Abstract of a doctoral dissertation at the University ofMiami.

Dissertation supervised by Professor Theodore H. Fleming

No. of pages in text: 115

Two species of hummingbirds, Patagona ~ and Rhodopis vesper, and one species ofnectar-feeding bat, Platalina genovensium, pollinate the colurnnar

Weberbauerocereus weberbaueri in southwestern Peru. Initial fruit-set frorn pollinator exclusion experiments conducted in 1991 and 1993, just prior to and after the El Niño event of 1991-92, revealed that bats were the most important pollinators in 1991, but that hurnmingbirds were most important in 1993. In both years, however, autogamy and lepidopteran larval infestation of fruits reduced differences in mature fruit praduction among pollinator exclusion treatments so that differences among treatments in mature fruit-set were not statistically significant.

Platalina ~novensium populationsdecreased at the study site between 1.991and

1993. Because the data indicate that bats are specializing on W. weberbaueri flowers and fruits for food, I hypothesize that bats emigrated from Arequipa during the drought because cactus flower and fruit production declined during this time. Persistence of small bat populations in Arequipa may be threatened because bats are harvested for use in folkloric medicinal remedies.

In Sonora, Mexico, one species of bat, Leptonycteris curasoae, and one species of hummingbird, Cynanthis lasirostris, were the most important pollinators of the organ pipe cactus, Stenocereus thruberi. Bats and hummingbirds deposited large quantities of pollen on stigmas per flower visit. Pollinator exclusion experiments indicated that the praportion ...•. ~.•. ~....~--.

--~~-

of fruits produced by hummingbird pollination was constant while that by bats was greater in 1993 than in 1992. Pateroity exclusion analysis ofbat-and hummingbird-pollinated seeds indicated that most pollen probably carne from within 75 m although substantial amounts (5-11 %) carne from beyond 125 and 150 meters respectively, for the two ~ thurberi populations studied. To rny father, Theodore, who instilled in us a sense of adventure and respect for nature and rny rnother, Teresa, who raised us so that we would carry a little bit ofPeru in our hearts.

iii Acknowledgments

1 would like to thank the members of my dissertation committee, Drs. Carol

Horvitz, David Janos, James Harnrick, Keith Waddington and Theodore Fleming for their constructive comments regarding my research. Their editing of this manuscript has improved it significantly, and for this 1 am most grateful. 1 would like to give special thanks to my dissertation advisor, Theodore Fleming, for introducing me to the fascinating world ofbats, cacti, and deserts. Dr. Stephen Buchmann assisted my analysis ofpollen deposition at the Carl Hayden Agricultural Research Station in Tucson, Arizona. Dr.

Jim Harnrick allowed me to work in his genetics laboratory at the University of Georgia.

My research in Peru would not have been possible without the assistance and friendship of several people. Victor Pacheco of the University of San Marcos Natural

History Museum in Lima first suggested that 1 visit Arequipa. Once in Arequipa,

Professor Percy Jimenez ofthe Universidad Nacional de San Agustin gave me important logistic advice and introduced me to Luis Baraybar, a biology student who would become my field assistant. Luis provided excellent full time assistance during the course of my field research. His knowledge regarding and animal s at the study site as well as his dry sense ofhumor contributed greatly to this study. Joachim Ugarte and Jose

E stefanero , also biology students, assisted me in the field at the very outset ofthe study, when help was most needed. Eleana Linares Perea offered me advice on where to find columnar cacti in Arequipa, and patiently taught me how to identify many species at the study site. Horacio Zeballos provided me with important data on rainfall for Arequipa.

While 1 was away from my study site and in Arequipa, many people made me feel

iv as ifl was at home. Blenda Femenias and Pablo Vera de la Cruz offered me not only their friendshíp, but gave me an archeological and anthropological perspective of Arequipa and its surroundings. Yolanda Hurtado and her family treated me as one of their own. I thank

Luis Baraybar' s family: Mr. and Mrs. Baraybar, his two sisters, Marta and Mayu, and their assorted pets, for good food, drink, and laughter on weekends.

During my stay in Mexico, I was fortunate to be able stay at the Prescott School's field station at Bahía Kino. Tad Pfister moved out of his house and into a very small, uncornfortable trailer so that I could stay in cornfort at his home. I am most grateful to

Tad for introducing me to the Sea of Cortez, where I finally got to see great whales.

Vinicio Soza, although busy with his own field work, never hesitated to lend me a helping hand ifl needed one (and I often did).

Whíle I learned the techniques of electrophoresis at the "Hamrick lab", John

Nason, Victoria Aspit, Elizabeth Stacey, and Mary Hagin made me feel welcome and were extremely patient with the many questions I had regarding the alien world of isozymes and gels.

I willleave the University ofMiarni with many fond memories. My experience as a graduate student here was enhanced by the opportunity to interact with many people too numerous to name here. In particular, I would like to give special thanks to Mindy

Nelson, Karen Lips, Devon Graham, Maureen Donelly, and Roger Gunther, who at various times throughout my graduate career have sat patiently through practice talks and read many drafts of proposals and manuscripts. Dr. Jay Savage not only taught me about biogeography, but also about how to relax and have a good time when a day' s work was done. Dave Janos, in addition to editing my manuscripts, has offered me advice and discussion on many aspects regarding my career interests. I feel very fortunate to have

v had the opportunity to participate in his tropical biology field course, otherwise known as

"the hike". It added a bit of adventure to rny life, to say the least.

Finally, 1 rnust gratefully acknowledge the best support group one could ever hope for: rny farnily. My graduate education has been a farnily affair in every sense ofthe word. During rny graduate educatíon, rny parents Theodore and Teresa, rny sister Carol, rny brother Teddy, and rny aunt Mery, have offered me unfailing support and love.

Financial support carne frorn the New York Zoological Society/The Wildlife

Conservation Society, the World Wildlife Fund-Garden Club of America Scholarship in Tropical Botany, Bat Conservation Internatíonal, the University ofMiarni, and

Theodore and Teresa Sahley.

vi TABLE OF CONTENTS

CHAPTER 1: Introduction...... 1

CHAPTER 2: The natural history and population status ofthe nectar feeding

bat, Platalina genovensium (PhylIostomidae: Glossophaginae) in

southwestern Peru 12

CHAPTER 3: PolIination biology and breeding system of a columnar

cactus in southwestern Peru 39

CHAPTER 4: Vertebrate poIlination of the organ pipe cactus, Stenocereus

thurberi: Effects on polIen deposition, polIen dispersal, and fruit

production 78

vii CHAPTERONE

INTRODUCTION

The Cactaceae comprise a diverse and ecologically important New World plant family. Approximately 1500 species and 100 genera are distributed throughout all major arid and semi-arid ecosystems of North, Central, and South America (Barthlott &

Hunt 1993, Gibson & NobeI1986). Despite their prominence in New World habitats, astonishingly little is known about the basic ecology of these plants. Cacti probably playa critical role in maintaining biodiversity of desert habitats because they provide food and shelter for a variety of desert animals including insects, lizards, birds, and mammals.

Because of their tolerance to prolonged dry spells, some species of cacti probably are the only plants to consistently provide nectar, pollen and fruit for animal s during droughts; thus, they may be considered "keystone species" in New World desert habitats ( sensu

Terborgh 1986, Gilbert 1980).

Humans harvest cacti and/or cactus products for a variety ofpurposes. Cactus fruit (especially that ofthe Opuntia) is commercially valuable in both North and

South America, and many indigenous communities harvest fruits of other species for local consumption (pers. obs., Nabhan 1985). For example, the organ pipe cactus,

(Stenocereus thurberi), produces fruit that is harvested by the Papago Indians of SW

Arizona as well as the Seri tribe in NW Mexico (pers. obs., Nabhan 1985). Other species of cacti produce hallucinogenic chemicals used in religious ceremonies by Native

Americans throughout much ofthe New World. "Peyote" (Lopophora w..), used in

Mexico and the US Southwest (Anderson, 1979) and the "San Pedro" (Trichocereus m2.)

~ 2 used in Peru (pers. obs.), are but two of several examples. On an international scale, cacti are prized by both amateur and professional horticulturists. Unfortunately, habitat destruction and over-collection of plants for cornrnercial trade and has led to endangerment for some species. In response to this problem, all cacti have been listed as

Appendix 1 (endangered) or 2 (threatened) under the CITES treaty (Hunt 1992).

Given their prominence, ecological importance, and utility to humans, it is surprising that there is such a lack of information on natural history, ecology, and reproductive biology for most species of cacti. The purpose of my study was to investigate pollination biology and fruit production of two species of columnar cacti occurring in Peru and Mexico, respectively. Specifically, I focused on exarnining bat and hurnrningbird pollination ofWeberbauerocereus weberbaueri (: ) at a high elevation site in southwestern Peru, and Stenocereus thurberi (Cactoideae:

Pachycereeae; referred to as organ pipe or pitahaya), at a study site in Sonora, Mexico.

Bat and hurnrningbird pollination in colurnnar cacti

A striking feature of columnar cacti occurring within all major arid-and serni-arid regions is the prorninence offloral syndromes associated with bat and/or humrningbird pollination; in fact, the association between vertebrates and pollination appears to be more cornrnon in arid habitats than in more humid ecosystems (Vogel 1969, Porsche 1939).

Floral traits associated with bat pollination inelude stout, funnel-forrnlcampanulate corollas, white flower color, and nocturnal nectar production, while traits associated with humrningbird pollination inelude red, diurnal, and zygomorphic flowers. Both syndromes 3 may include the production oflarge amounts ofpolIen and nectar (Helverson 1993, Faegri and Van der Pijl 1979).

Why does there seem to be an association between bats, hurnmingbirds, and colurnnar cacti? Insect polIination is the prirnitive and most common mode of pollination for cacti, and occurs in alI three cactus subfamilies (Pereskiodeae, Opuntiodeae, and

Cactoideae). The high incidence ofvertebrate polIination syndromes in columnar cacti and the fact that they have probably evolved independently in several tribes (Gibson &

Nobel 1986, Barthlott and Hunt 1993), suggests that floral traits most attractive to vertebrate polIinators may confer specific advantages to these plants. It has been proposed that these advantages may include long-distance polIen dispersal, deposition of large polIen loads, and reliable polIination (Bertin 1982, Cruden 1977, Heithaus et al.

1974). These proposed advantages may also have micro-evolutionary consequences for plants polIinated by bats and hummingbirds. For example, reliable pollination by one type of polIinator may influence selection and result in the evolution of specialized mutualisms

(Howe 1984); depositing large polIen loads may influence polIen competition and have effects on both male and female components of fitness (Mulcahy 1983), and long-distance polIen dispersal wilI influence plant population genetic structure and neighborhood size

(Harnrick 1987). Although several recent studies have experimentalIy exarnined the importance ofbats and/or humrningbirds as pollinators of columnar cacti (Fleming et al, in review, Petit 1995, Nassar 1990) none have previously investigated patterns ofpolIen deposition and polIen-mediated gene flow due to vertebrate pollination. 4 The study organisms

The genus is restricted to Peru and ranges along the western cordillera of the Andes throughout the length of this country (Backeburg 1977). The tribe to which it belongs, the Trichocereeae, has as its center of diversity the arid and semi-arid regions of SW Andean South America. A recent estimate of generic diversity of this tribe indicates that approximately eight to fifteen genera occur in Peru, while the remaining eight genera are distributed in Bolivia, Chile, Argentina, and Brazil. Cactus genera belonging to this tribe display a range of floral pollination syndromes (bee, moth, bat, and hurnmingbird; see Table 1), although bat and/or hummingbird pollination syndromes predominate in this group. W. weberbaueri flowers open in the afternoon and remain open until approximately 11:00 a.m. the following morning. At my study site in southwestern Peru, they are visited by one species of nectar-feeding bat and two species of hurnmingbirds. Floral polymorphism is apparent at the study site, and flowers exhibit morphology consistent with both bat and hummingbird pollination syndromes.

The tribe Pachycereeae (to which Stenocereus thurberi belongs), comprised of approximately ten genera, has its principal center of diversity in the arid regions of

Mexico. Genera occurring within this tribe extend into the Caribbean islands as well as northern South America, but do not occur in Peru. Like the Trichocereeae, cacti occurring within the Pachycereeae also display a variety of pollination syndromes (Table

2), although vertebrate pollination predominates in this group as well. Stenocereus is a genus comprised of approximately 25 species which occur in Mexico, Central America, the West Indies, Venezuela and Colombia. Flowers are generally funnelform or 5 campanulate, and usually nocturnal, although two species (formerly ofthe genus

Rathbunia), have flowers which are narrowly tubular, zygomorphic, diurnal, red and apparentIy specialized for hummingbird pollination. Stenocereus thurberi occurs in

Mexico south from the state of Sinaloa, north into southern Arizona. Because its flowers open in the evening and begin closing between 7:30 and 9:00 a.m, it is visited and pollinated by bats and hummingbirds. Although floral polymorphism is not as marked for

.s.,. thurberi as it is for W. weberbaueri, flowers at the study site exhibit variation in corolla length, width, and color (which ranges from white to pink/red). Like its South American counterpart, .s.,. thurberi exhibits floral adaptation to both bat and hurnmingbird polIination.

Objectives of this study

In this study I investigate how bat and hummingbird pollination influence fruit production and polIen dispersal of columnar cacti. To my knowledge this is the first study to examine both pollination success and gene flow in a vertebrate-pollinated plant. In the second and third chapters I describe several aspects of the mutualism occurring between the columnar cactus W. weberbaueri and its pollinators at a study site in southwestern

Perno In the fourth chapter I examine bat and hurnmingbird pollination of the organ-pipe cactus, Stenocereus thurberi, at a study site in Sonora, Mexico, and investigate specific questions pertaining to pollen deposition and dispersal by bats and hurnmingbirds. 6 Chapter 2: The natural history and population status ofthe nectar feeding bat,

Platalina genovensium (PhylIostornidae, Glossophaginae) in southwestern Peru.

In this chapter I present data obtained for Platalina genovensium, a rare nectar feeding bat endernic to the arid regions of'Peru. It is the only species ofnectar feeding bat to occur at mid-to high elevation sites in southwestern Peru, The skull morphology ofthis bat, which is unusual because of its exceptionally long rostrum, has led investigators to hypothesize that it is highly specialized for feeding on flower nectar. Also, aspects of its lingual morphology strongly suggest that it has independently evolved the nectar-feeding habit of most other Glossophagine bats. However, prior to this study, no information on population status, natural history or ecological data had been collected for this species. In this chapter I present information suggesting that P. genovensium specializes primarily on columnar cacti for food, and feeds on fruits as well as nectar and pollen. I also describe how bat abundance and w. weberbaueri fruit and flower production decreased during a severe two-year drought at the study site, and suggest that emigration from the study area probably accounted for the decrease in bat numbers. I discuss how in Arequipa, bat populations may be threatened by human colIection for folkloric medicinal use.

Chapter 3: The polIination biology of the columnar cactus, Weberbauerocereus weberbaueri: The role ofbats, hummingbirds, and El Niño.

In this chapter I present data on flower morphology, genetic diversity and pollination effectiveness ofbats, hummingbirds, and insects for W. weberbaueri at a 2500 m elevation study site in southwestern Peru. W. weberbaueri displays floral polymorphism 7 at the study site, and exhibits characters associated with both bat and hummingbird pollination. TO determine the relative importance of bats, hummingbirds, and insects as pollinators ofW. weberbaueri , 1 conducted pollinator exclusion experiments on flowers just prior to and after a two year drought caused by an El Niño event. 1 discuss how shifting pollinator effectiveness among years, fruit mortality due to larval infestation, and tetraploidy interact to allow for the persistence offloral polymorphism in this cactus. 1 also emphasize the importance of El Niño events in influencing patterns ofrainfall and subsequent ecological interactions occurring at the study site.

Chapter 4. Vertebrate pollination ofthe organ pipe cactus, Stenocereus thurberi: consequences for pollen deposition, pollen dispersal, and fruit production ..

In this chapter 1 present data on pollination success ofbats and hummingbirds, quantify patterns of single-visit pollen deposition on stigmas, and examine the influence of bat and hummingbird pollination on pollen-mediated gene flow for ~ thurberi. 1 compare the data 1 obtained on pollination success with data obtained from a previous two-year study. The results ofthis study are discussed in terms ofbat and hummingbird flower visitation and foraging behavior.

~ 8 LITERA TURE CITED

Anderson, E.F. 1979. Peyote. The divine cactus. Tucson: University of Arizona Press.

Backeberg, C. 1977. Cactus Lexicon. Blandford Press, Poole, England.

Bartholott, W. & Hunt, D.R 1993. Cactaceae. In: The families and genera ofvascular plants, Vol 11: Flowering Plants-Dicotyledons. (Edited by K. Kubitzski, 1.G. Rohwer, and v. Bittrich). Springer- Verlag, Berlin.

Bertin, RI. 1982. Floral biology, humrningbird pollination and fruit production oftrumpet creeper (Campsis radicans, Bignoniaceae) Am. 1. Botany 69(1):122-134.

Cruden, RW. 1976. Pollinators in high-elevation ecosystems: Relative effectiveness of birds and bees. Science; 1439-1440.

Faegri K. & L. Van der Pijl. 1979. The principies ofpollination ecology. Pergamon Press, Oxford.

Fleming, T.H., Horner, M. & Tuttle, M.D. In review. The relative effectiveness of diurnal and nocturnal pollinators to three species of columnar cacti. Southwest American Naturalist.

Gibson, AC. & Nobel, P.S. 1986. The Cactus Primer. Harvard University Press, Cambridge, USA

Gilbert, L. 1980. Food web organization and the conservation of neotropical diversity. In: Conservation Biology: An evolutionary-ecological perspective (M. Soule & B. Wilcox, eds.). Sinauer Associates, Sunderland, Massachusetts.

Hamrick, 1.L. 1987. Gene flow and distribution of genetic variation in plant populations. In: Differentiation patterns in higher plants (K. Urbanskak, ed.) Acadernic Press, New York. Pp. 53-67.

Helverson, O.V. 1993. Adaptations of flowers to the pollination by glossophagine bats. Pp. 41-59 in: Plant-Animal Interactions in Tropical Environments (ed. W. Barthlott et al.). Museum Alexander Koening, Bonn.

Heithaus, E.R, Opler, P.A, Baker, H.G., 1974. Bat activity and pollination ofBauhinia pauletia: Plant -pollinator coevolution. Ecology 55:412-419. 9 Howe, H.F. 1984. Constraints on the evolution ofmutualisms. American Naturalist 123 :764-777.

Hunt, D. 1992. CITES Cactaceae Checklist. Royal Botanic Gardens Kew.

Mulcahy, D.L. 1983. Pollen competition in a natural population. In: Handbook of Experimental Pollination Biology. (E.e. Jones, & R.l. Little, eds.) van Nostrand Reinhold, New York, Pp. 330-337.

Petit, S. 1995. The mutualism between bats and columnar cacti in Curacao and conservation implications. Ph.D. Dissertation, University ofMiarni.

Porsche, O. 1939. Das Besubugleben der Kadkteen blute. Jahrb. Deutsche Kakt-Ges.

Nabhan, G.P. 1985. Gathering the deserto University of Arizona Press. Tucson, Arizona.

Nassar, J. 1991. Biologia reproductiva de cuatro cactaceas quitropterofilas Venezolanas (Cereae: Stenocereus griseus, Pilosocereus moritzianus, Subpilocereus repandus y ~ horrispinus) y estrategias de visita de los murcielagos associados a estas. Bacholors Thesis, Universidad Central de Venezuela. Caracas, Venezuela.

Terborgh, l. 1986. Keystone plant resources in the tropical forest. In; Conservation biology: The science of scarcity and diversity (M. Soule, ed.). Sinauer Associates, Massachusetts.

Vogel, K. 1969. Chiropterophilie in der Neotropischen flora. Neue Mitt. 11,111Flora abt. B, 158:185-323. 10 Table 1. Genera belonging to the tribe Trichocereeae according to Barthlott & Hunt (1993). Distribution, flower morphology, and pollination syndromes are from the above- mentioned paper as well as personal observations. The number of previously recognized genera that are subsumed within currently recognized genera are in parentheses.

Genus Main Flower morphology Pollination distribution syndrome Peru Tubular, zygomorphic, Humrningbird (11) diurnal, red to orange Peru Funnel-form to campanulate; Bee, humrningbird (8) nocturnal and diurnal bat Peru Tubular-campanulate, Bat, moth? (4) nocturnal, pale (includes Peru Tubular-funnelform, white- Bat, humrningbird Weberbauerocereus) red, nocturnal & diurnal (2) Peru Tubular-funnelform. diurnal, Humrningbird (3) orange-scarlett Mi/a Peru Small, funnelform- Bee campanulate, yellow Peru Tubular-funnelform, Hummingbird (2) zygomorphic, diurnal, red Peru Short funnel-form- Bee campanulate, yellow to red Leocereus ( 1 endemic sp.) E. Brazil Tubular, nocturnal Bat

Brachycereus (1 endemic sp.) Galapagos Funnelform, white Moth? (1 endemic sp.) E. Brazil Tubular-campanulate, Bat, moth? nocturnal Brazil Tubular, nocturnal Bat (2)

Semaipaticereus Bolivia Funnel-form, nocturnal Bat

Denmoza Argentina Tubular, diurnal, scarlett Humrningbird

Arthrocereus Brazil Funnelform, nocturnal Bat

Rebutia Bolivia, Small, diurnal, various Bee (4) Argentina colors Brazil, Bol. Funnel-form campanulate, Bee (2) Argentina diurnal, various colors Disocactus Brazil, Bol. Slender-tubular, nocturnal, Moth Para~ white 11

Table 2. Genera belonging to the tribe Pachycereeae according to Barthlott & Hunt 1993. Distribution, flower morphology, and pollination syndromes are taken from the above- mentioned paper, Fleming et al. (in rev.), Valiente-Banuet et al. (in press).

Genus Main distribution Flower Morphology Pollination syndrome Bergerocactus SW. US,NW. Small, diurnal, Bee (1 sp) Mexico pale yellow

Pachycereus Mexico Small to medium Primarily bat, some (12 sp) size; bird pollination nocturnal/ diurnal, funnel-form Carnegiea SWUS,NW White, Bat, bird (1 sp) Mexico nocturnal/ diurnal funnel-form Neobuxbamia Mexico Tubular -funnelform Bat, hummingbird (7 sp) nocturnal/ diurnal white to pink Cephalocereus Mexico Tubular- Bat, hummingbird? (3 sp) campanulate, nocturnal, white to pink Stenocereus Mexico, N. South Funnel-form to Bat, hummingbird (25) sp America, Caribean campanulate, usually nocturnal , white to pink Rathbunia NWMexico Tubular, red, diurnal Hummingbird (will be changed to Stenocereus)

Polaskia S Mexico Small campanulate, Insect? (2 sp) diurnal, white or Hummingbird? pink Escontria (1 sp) S Mexico Short, tubular, Insect? diurnal Hummingbird? Myrti llocactus Mexico, Guatemala Diurnal, small Insect? CHAPTER2

THE NATURAL mSTORY AND POPULATION STATUS OF THE NECTAR-

FEEDING BAT, PLATALINA GENOVENSIUM (PHYLLOSTOMIDAE:

GLOSSOPHAGINAE) IN SOUTHWESTERN PERU.

Co-author: L. Baraybar

ABSTRACT

We present data on morphology, diet, and population status ofthe nectar-feeding bat, Platalina genovensium, at a 2500 m site in Arequipa, Perno Our field observations and

813 carbon isotope ratios of toe muscle tissue indicate that P. genovensium depends primarily on the columnar cactus, Weberbauerocereus weberbaueri, for food at this site.

Numbers ofP. genovensium are low in Arequipa. In late 1991 we estimated bat abundance at 86 individuals in an area of approximately 1256 km 2. In 1992, during a severe drought caused by an El Niño event, cactus flower and fruit production decreased as did the number ofbats. In late 1993, after normal rains resumed, flower and fruit production increased, but bat numbers were still 60% lower than in 1991. We suggest that P. genovensium may have emigrated from our study are a due to drought conditions caused by El Niño. By influencing bat abundance, El Niño events may have significant influence on the co-evolution ofbat-plant mutualisms in southwestem Perno Persistence of small bat populations in Arequipa may be threatened because bats are harvested for use in folkloric medicinal remedies.

12 13 INTRODUCTION

Platalina genovensium (Glossophaginae, Phyllostomidae) is a rare nectar-feeding bat endemic to the arid regions of western Peru. This monotypic genus was first described by Thomas (1928) from a single individual collected in the department ofLima.

Collection localities inc1ude those from the departments ofPiura, Lima and Arequipa

(Aellen 1965, Jimenez & Pefaur 1982, Ortiz de la Puente 1951), as well as one location east ofthe Andes, in the arid valleys ofthe department ofHuanuco (Sanborn 1936).

These collection sites range in altitude from near sea-level (Lima) to 2500 m (Arequipa and Huanuco), and all are found within arid habitats. Tuttle (1970) hypothesized that

Platalina probably underwent divergence in coastal Peru prior to the Andean uplift, and later dispersed east of the Andes. He suggested that dispersal east of the Andes may have occurred through two trans-Andean passes in northwestern Peru that are approximately

2500 m above sea-level.

Evidence strongly suggests that Platalina, Lonchophylla, and Lionycteris share a common ancestor and independentIy evolved the nectar-feeding habit of other glossophagine bats (Griffiths 1982, Van den Bussche 1992). Platalina and its sister taxa,

Lonchophylla and Lionycteris, exhibit unique morphological attributes of the hyoid and lingual regions relative to other glossophagine bats. These inc1ude differences in tongue papillae, internal histology, and musculature (Griffiths 1982, Winkelman 1971). Based on a c1adistic analysis of these morphological characters, Griffiths (1982) recommended the elevation of these genera to subfamilial status. Recent restriction-site endonuc1ease data corroborate Griffith's morphological analysis (Van den Bussche 1992). 14 Species ofnectar-feeding bats that may share at least part ofthe range ofP.

genovensium inelude Lonchophylla hesperia, Anoura geoffroyi, and Glossophaga soricina.

L. hesperia is restricted to northwestern Peru, A. geoffroyi is not found south of the

department ofLima west ofthe Andes, and G. soricina does not occur above 1400 m

(Koopman 1978, Tuttle 1970). Thus, P. genovensium is the only species ofnectar-feeding

bat present at rnid-to-high elevation sites in southwestern Peru and is the only potential

bat -pollinator and seed disperser of plants in this region.

Despite its probable role as a pollinator and seed disperser of many species of

plants in arid habitats and its purported high degree of morphological specialization for

nectar-feeding (Winkelman 1971), nothing is known about the natural history, population

status, and ecological importance ofthis bat. Since 1990, we have investigated the role of

P. genovensium as a pollinator and seed disperser of columnar cacti belonging to the

genus Weberbauerocereus. The objectives ofthis paper are to surnrnarize the data

obtained between 1990 and 1994 regarding the natural history, ecology, and population

status of P. genovensium. We discuss the effects the El Niño event of 1991-92 has had on

bat populations, as well as the threat posed by human collection ofbats for use in

medicinal remedies.

STUDY AREA AND METHODS

The main study site is located in the department of Arequipa, Peru, 20 km south of the city of Arequipa (16.27° S, 71.30° W) on the western slope ofthe Andes, at

approximately 2500 m above sea level. At our main study site we observed bats visiting flowers of cacti and recorded cactus phenology and density. Additional bat census data 15 were recorded at various locations within a twenty km radius of the city of Arequipa.

Examination of a topographic map of this are a indicates that within the 1256 km2 area searched, at least 455 km2 is under residential or agricultural use. The sites where we found bat colonies all correspond to the subtropicallow-montane desert scrub life-zone classified by the Holdridge system (ONERN, 1976). Average annual rainfall for this life zone is 222 rnm. Between 1990 and 1993, however, a drought associated with an El Niño event resulted in a drastic decrease in rainfall in Arequipa.

Three species of colurnnar cacti occur at the study site: Weberbauerocereus weberbaueri, Browningia candelaris, and Coryocactus brevistylus. Of these, only

Browningia and Weberbauerocereus have long, tubular, nocturnal flowers that can be visited by bats. W. weberbaueri is the most abundant colurnnar cactus at the study site and around the city of Arequipa (Aragon 1982, pers. obs.). W. weberbaueri is visited and pollinated by the nectar-feeding bat P. genovensium, as well as two species of hurnmingbirds, Patagona ~ and Rhodopis vesper (pers. obs.). Davila et al. (1987), and

Zeballos & Davila (1993) provide additional information on the composition of the fauna in Arequipa, while Linares (1992) provides a detailed description of the composition and distribution of flora in this region.

Bats were located by searching all known mines and tunnels within a 20 km radius of Arequipa. No natural caves occurred within this radius. Three colony localities out of seven that we found were monitored from one to five times ayear until early 1994. Visual counts were made by entering the mines or waiting outside until bats exited. Captures were made using mist nets at one colony site in 1990 (Batolito mine #3) and at two sites 16 (Charcani and Batolito mine #3) in 1991-93. In 1990, LEB recorded relative age, forearm length, and reproductive condition of captured individual s; in 1991-93 we also recorded body mass. For three individuals, we measured and recorded wingspan and wing are a using methods described in Norberg and Rayner (1987).

Phenological data for Weberbauerocereus weberbaueri were obtained at the main study site by counting the numbers of buds, flowers, developing fruits and mature fruits, for 40 marked individuals between October 1991 and October 1993. These cacti were located approximately 40 km from the Charcani roost and approximately 15 km from the

Batolito roost. Plants were censused once a week when CTS was present at the site but only once a month whenever possible when only LEB was present. The density of adult

B. candelaris and W. weberbaueri cacti at the main study site was deterrnined by counting the number of cacti greater than one meter in height in ten 600 m2 plots that were separated by 50 meters.

Carbon isotope analysis of bat toe muscle tissue was performed to determine this species' dependence on CAM cacti for food. Muscle tissue from one toe per individual was obtained from three live bats captured in January 1991 and from three preserved specimens in the collection ofthe Universidad Nacional de San Agustin (no collection date was available for these). Analysis oftissues was performed according to methods outlined in Herrera et al. (1993) and Buchanon & Corcoran (1959). Due to extremely low numbers of bats found in 1992 through 1994, tissue samples were not collected in these years. No specimens were collected during the course of our study. 17 Statistical analysis was performed using SYSTAT version 5.3 (Wilkinson, 1991).

We performed t-tests on bat morphological measurements to determine if significant differences existed among sexes.

RESULTS

Morphology

P. genovensium is a relatively large nectar-feeding bat, third in size to

Leptonycteris curasoe and L. nivalis (Table 1; see Norberg & Rayner 1987 for comparative morphological data on glossophagine species discussed in this paper). Adult females tend to have a slightly greater body mass (1=2.954, f=0.004) and longer forearm length (1=3.152, f=0.002) than adult males. P. genovensium has a large wingspan and forearm length, second only in length to L. nivalis. The wing area (0.0202 m2) of

Platalina is the largest of any glossophagine recorded thus far, while its wing loading, (9.4

N/m2) is among the lowest.

Diet

We obtained 813C ratios of -10.294, -10.52, -10.90, -9.40, and -10.48 from tissue samples of five individuals of f. genovensium. A previous study found that tissues of

CAM cactus and agave plants frorn Mexico have 813 C values of -12.6 while bat-visited

C3 plants have an average 813C of -26.4 (Fleming et al. 1993). These data, along with our observations, indicate that P. genovensium is feeding almost exc1usively frorn cactus 18 flowers and/or fruits in Arequipa. Our estimates of cactus density indicate that W.

weberbaueri is much more abundant than B. candelaris at our site. Densities of adult

plants (> 1m in height) were 98 plants/ha for W. weberbaueri and 4 plants/ha for B.

candelaris. Observations offlower visitation behavior and ofbat droppings in day roosts

indicated that bats are feeding from both flowers and fruits ofW. weberbaueri.

Bat populations and cactus phenology

Four colonies were located in 1990 and an additional three were located in 1991

(Table 2). AlI sites occurred between 2200 and 2600 m above sea level, and all were

located within Weberbauerocereus habitat. Visual counts indicated that colony sizes

ranged from a minimum of one to a maximum of 50 individuals; prior to the 1992 drought,

the median colony size was five. Within a mine, bats would often be found in groups of 5-

7 at different roost microsites. In late 1991, a census of five of the seven mines revealed

the presence of approximately 74 individuals. Due to logistical constraints, in late 1991

we were not able to census the remaining two sites we had previously found. If we

assume these two sites contained a total of ten individuals (the maximum number of bats

found there in 1990), we estimate that the total number of individuals present in tunnels

and mines within a 20 km radius of the city of Arequipa in late 1991 was approximately 86

individuals.

In 1990, prior to the El Niño drought, groups containing males, females, and

subadults were present year round. Numbers of P. genovensium declined after the onset

of the most severe part of the El Niño drought in April 1992 (Fig.1 a & b). Fruit and

I " I "~ 19 flower production of W. weberbaueri also decreased sharply in late 1991 and early 1992 although it never stopped completely (Fig. 2). Flower density at the main study site dropped from 860/km2/day in October 1991 to 93/km2/day by Apri11992. Mature fruit production decreased from 193/km2/day to 14 fruits/km2/day during this same time periodo After rains returned in 1993, fruit and flower production in W. weberbaueri increased and bat numbers increased slightly but did not approach levels present before the drought. Only twenty-four individuals were found at the two largest colony sites

(Charcani and Batolito #3) in October 1993. This is a 60% decrease in number at these sites compared with late December 1991 and early January 1992 when 60 individuals were found. Bat numbers remained lower in 1994 than they were in 1991. For example,

Charcani cave, the site of the largest colony that once contained up to 50 bats in 1991, contained only three individuals in March 1994.

Sex composition of the roosts varied over time. Social groups contained only males or males and females (Fig. 3). For example, the Charcani cave, which contained approximately 50 individuals, contained only adult and subadult males from 1991 untillate

1993. Subadult males occurred in all male and mixed-sex colonies, but subadult females occurred only in mixed sex aggegations (Fig.4). Although capture data indicated the adult male and female sex ratio was almost exactly 1:1, we captured a larger number of subadult males than subadult females. It is possible we missed some maternity roosts and under- estimated the proportion of juvenile and/or subadult females present.

After March 1992, during the height of the drought, we did not capture any subadults or juveniles. We did not capture any reproductively active females (i.e., 20 pregnant or post-Iactating) from January 1991 through October 1993 (Fig.5). This two- year span corresponds with the period of drought and low fruit and flower production of

W. weberbaueri. Potentially reproductive adult males (those with medium to large testes), however, were found year-round in Arequipa, in all years from 1990 through 1993. In

1990, pregnant females were captured in March, August, and December, with the peak occurring in December. In 1993, pregnant and post-lactating females were captured in

October (Fig. 5). These months correspond to a period from late winter to early surnmer, just prior to and during the onset of the rainy season.

DISCUSSION

Our data indicate P. genovensium is a cactus specialist at our study site in southwestem Peru. Flower visitation observations and 613C values indicate that at

Arequipa, P. genovensium subsists almost entirely on CAM plants. AIthough P. genovensium probably visits flowers of the other co-occurring noctumally flowering columnar cactus, Browningia candelaris, this plant occurs at extremely low densities and flowers only two months ofthe year while W. weberbaueri flowers year-round. Thus, B. candelaris flowers and fruits probably comprise on1y a fraction ofthe diet ofP. genovensium. Our observations suggest that in our study area, P. genovensium obtains most of its nourishment from W. weberbaueri. Pollinator exclusion experiments conducted on flowers ofW. weberbaueri indicate that P. genovensium is also an important pollinator ofthis cactus (Sahley, chapter 3). 21 P. genovensium probably feeds from columnar cacti throughout much of its range, especially where it occurs along the western slopes of the Andes. This is because cactus abundance and diversity in Peru is greatest from 700 to 3000 m on the Pacific slope

(Sarmiento 1975, Weberbauer 1945). In addition, the western slope ofthe Andes is a center of diversity for the columnar cactus tribe Trichocereeae, of which approximately three to four genera exhibit floral adaptations consistent with bat pollination (see Barthlott

& Hunt 1993 for descriptions offloral morphology). Therefore, we hypothesize that P. genovensium populations may be most abundant at rnid-elevation (700-3000 m) sites on the western slopes of the Andes.

Our morphological data indicate that P. genovensium has a large wing area and wingspan relative to other glossophagine bats for which measurements are available.

Since air density decreases as altitude increases, the power required for hovering and forward flight also increases at altitudes above sea level (Norberg 1990). A large wingspan and low wing loading as found in P. genovensium will reduce the power required for flight at high altitudes (Norberg & Rayner 1987). Feinsinger et al. (1979) found that for humrningbirds occurring on the eastern Andean slopes ofPeru, mean wing disc loading (ratio ofwing area to weight) decreased with increasing elevation in humrningbirds. They suggested that the low wing loading found in humrningbirds at high elevations is a potential adaptation to reduce energetic expenditure during hovering and forward flight. It is possible that wing morphology ofP. genovensium may also represent adaptation to flight in montane desert habitats. 22 Despite its exceptionally elongated skull and its hypothesized specialization for nectar-feeding (Winkelman 1971), P. genovensium feeds on fruits as well as flowers ofW. weberberbaueri and may be an important seed dispersal agent for this cactus. We found large quantities of cactus seeds in bat droppings on roost floors. Other studies have shown that nectar-feeding bats eat the pulp of cactus fruits as well as nectar. For example, in Sonora, Mexico, Leptonycteris curasoae feeds from both flowers and fruits of three species of columnar cacti (Flerning et al in review, V. Sosa, pers. comm.). In arid

Venezuelan habitats, Glossophaga longirostris includes a higher proportion of cactus fruits than flower nectar and pollen in its diet (Soriano et al. 1991). Current research is addressing the relative importance offlowers and fruits in the diet ofP. genovensium (L.

Baraybar, in prep.).

We hypothesize that the long rostrum ofP. genovensium does not necessarily indicate specialization on nectar per se. Rather, this bat' s skull morphology probably represents adaptation to morphological attributes ofthe flowers upon which it feeds. For example, W. weberbaueri flowers exhibit floral attributes consistent with both bat and humrningbird pollination syndromes, and its flowers are pollinated by humrningbirds and bats (Sahley.chapter 3). Flowers can be over 10 cm long; corolla tube width is narrow, especially above the nectary, and ranges from 1.0 to 1.9 cm in diameter (Sahley, chapter

3.). Long tubular flowers such as these are typical ofthe "hurnmingbird pollination syndrome"(Faegri and van der Pijl 1979) which might select for a relatively long muzzle in P. genovensium. Thus, Platalina's skull morphology may be the result of a three-way co-evolutionary scenario that includes bats, the cacti they feed on, and humrningbirds. To 23 evaluate this hypothesis, we need additional information on the phylogenetics and ecology of the food plants that P. genovensium includes in its diet throughout the rest of its range.

Census data.

Tuttle (1970) referred to P. genovensium as being "extremely rare" and our data are in accord. P. genovensium abundance is low at Arequipa. We estimate that prior to the drought, a total of 86 bats were present within a twenty kilometer radius of Arequipa.

Although it is possible we may have missed some mines or small tunnels within our study area, we feel that this population estimate is reasonable, because we used topographic maps and interviews with local people to locate potential colony sites. Because no natural caves occur within our study area, it is possible that the bats are adapted to roosting in rocky crevices or shallow depressions along ridges. We never observed bats emerging from these types of roosts, but, if P. genovensium bats are using them, then our censuses may have underestimated the number of bats present, since we censused only man-rnade structures. The pattem of low density we found contrasts with that of the arid-dwelling

Leptonycteris curasoae, a highly social, relatively abundant, and widespread nectar feeding bat (Arita 1993, Cockrum & Petryszyn 1991, Howell 1979) and with the common

Glossophaga longirostris and G. soricina,( Arita 1993, Soriano et al. 1991, pers. obs.). P. genovensium' s habit of roosting in relatively small groups may be most similar to that of the arid-dwelling Choeronycteris mexicana (Arita 1993). Unfortunately, little comparative information exists on the abundance of other species of arid-dwelling nectar-feeding bats.

Although bats ofboth sexes decreased in number after the onset of drought in

1992, we captured a few males but no reproductively active females during the most 24 extreme portion ofthe drought. Information obtained regarding the nectar-feeding bat

Leptonycteris curasoae indicates that males and females are geographically segregated during certain portions ofthe year and exhibit different migratory strategies (Cockrum

1991). This scenario may apply to P. genovensium as well.

Our data indicate that P. genovensium is resident around Arequipa year-round in years of normal rainfall, and may migrate during extremely dry El Niño years. We do not have enough information to determine the geographic scale ofthe 1992 migration or whether it was altitudinal or latitudinal. Fleming et al. (1993) proposed that in Mexico, certain species of columnar cacti and agaves provide a spatio-temporally predictable

"nectar corridor" that provides nutritional and energetic resources for migratory nectar- feeding bats. Currently, we cannot begin to speculate on the availability of a similar nectar corridor in Peru. Moreover, in Peru, such a nectar corridor may be spatio-temporally unpredictable because of irregular fluctuations in the frequency and severity of El Niño events (Cane, 1983). Unfortunately, data on the phenology of cacti in Peru during either normal or El Niño years is not available.

After flowering and fruiting of W. weberbaueri increased in 1993, P. genovensium populations remained low relative to 1991. Several factors could have accounted for the continued small populations. First, bats may not have returned from the area to which they had migrated, perhaps due to lack of a sequentially blooming nectar corridor as apparently occurs in Mexico (Fleming et al., 1993). Second, drought and/or migration may have resulted in higher than normal mortality in 1992 and 1993, and third, collecting ofbats for medicinal purposes might have played an important role in observed declines. 25 Although we had no evidence of increased mortality in colonies of P. genovensium because of drought, it is possible that a combination of all three factors may have been responsible.

Results of pollinator exclusion experiments indicate the low bat population in 1993 resulted in significantIy lower fruit set from bat-pollinated flowers ofW. weberbaueri in

1993 compared with 1991 (SahIey, chapter 3). We propose that local decreases in abundance ofP. genovensium triggered by El Niño events, may have had considerable influence on the co-evolution ofthis bat-cactus-hummingbird mutualism.

Conservationlthreats

A major threat to continued persistence ofP. genovensium around Arequipa is roost disturbance and collection ofbats for folkloric medicinal purposes. Bats are believed to cure a variety ofailments ranging from epilepsy to heart attacks (pers. obs.).

Bats can be obtained from healers, or "curanderos", and at medicinal stands within city markets, although at market stands in Arequipa we found only molossids and vespertilionids that had been collected in other areas. lnformation we obtained from informal interviews in the central market in downtown Arequipa suggests that several hundred bats might be sold annually at this market alone. We think that the absence ofP. genovensium from market stand s reflects its rarity, rather than non-random choice for the other bats. Despite their absence in these markets, a local, non-commercial demand for P. genovensium exists. One ofthe colonies we monitored (Batolito mine) was visited occasionally by medicinal collectors (pers. obs.). Unfortunately, because of their habit of roosting in old mines, P. genovensium bats are especially vulnerable, since the locations of 26 most mines are well known by local inhabitants. At this point in time, our best judgement is that any collecting ofbats in Arequipa, whether for scientific or medicinal use, is unsustainable.

CONCLUSIONS

Our data indicate that P. genovensium is a cactus specialist that occurs at low densities at our study site in Arequipa, Perno Census data obtained between 1990 and

1994 indicate that P. genovensium is a year-round resident in Arequipa during years of normal rainfall, but may be migratory during extremely dry El Niño years.

Currently, there is no data on the population status and natural history of~. genovensium throughout the rest of its range. Obtaining such information is especially important because ofP. genovensium's role as a pollinator of columnar cacti, the low numbers ofbats found in Arequipa, and potential over-harvesting ofbats for medicinal purposes.

ACKNOWLEDGEMENTS

M. Donelly, T. Fleming, C. Horvitz, D. Janos, K. Lips, S. Petit, K. Waddington and R. Winkelman reviewed this manuscript and provided valuable discussion, assistance and comments. L. Sternberg provided laboratory facilities for carbon isotope analysis. P.

Jimenez gave the authors important logistic advice and made our collaboration possible.

E. Linares assisted with plant identification. P. Baldizan graciously permited the authors to conduct research on his property. Funds for this research were provided by grants from 27 The Wildlife Conservation Society, a World Wildlife Fund-Garden Club of America

Scholarship in Tropical Botany, Bat Conservation Intemational, and a Maytag Graduate

Fellowship from the College of Arts and Sciences, University ofMiami (awarded to C.

Sahley). 28 LITERA TURE CITED

Aellen, V. 1965. Sur une petite collection de chiropteres du nord-ouest du Perou. Mammalia, 29:563-571.

Aragon, G. A. 1982. Cactaceas de los alrededores de la ciudad de Arequipa. Boletin de Lima, 4(21):86-94.

Arita, H. 1993. Conservation biology ofthe cave bats ofMexico. Journalof Marnmalogy, 74(3):693-702.

Buchanon, D. & Corcoran, B. 1959. Sealed tube combustions for the determination of 14C and total carbono Analytical Chemistry, 31:1635-1638.

Cane, M.A. 1983. Oceanographic events during El Niño. Science, 222(4629): 1189- 1194.

Cockrum, E.L. 1991. Seasonal distribution ofnorthwestern populations ofthe long- nosed bats, Leptonycteris sanborni: Family Phyllostomidae. Anal. Instit. Biol. Universidad Nacional Autononoma. Mexicana Serie Zoologia.

Cockrum, E.L., & Petryszyn, Y. 1991. The long-nosed bat, Leptonycteris: an endangered species in the southwest? Occassional Papers Museum Texas Tech University, 142: 1-32.

Davila, lD., Lopez, lE., Jimenez, P. 1987. Los mamiferos del departamento de Arequipa, Perno Boletin de Lima, 54: 11-13.

Faegri, K., and L. van der Pijl. 1979. The principles ofpollination ecology. 2nd edition. Pergamon Press, Oxford, England, pp 291.

Feinsinger, P., Colwell, R.K., Terborgh, l, Chaplin, S.B. 1979. Elevation and the morphology, flight energetics, and foraging ecology oftropical hummingbirds. American Naturalist, 113(4):481-497.

Fleming, T.H., Horner, M. & Tuttle, M.D. In review. The relative effectiveness of diurnal and nocturnal pollinators in three species ofMexican columnar cacti. Southwest American Naturalist. 29 Fleming, TH., RA. Nunez, & L. Sternberg 1993. Seasonal changes in the diets of migrant and non-migrant nectarivorous bats as revealed by carbon stable isotope analysis. Oecologia,94:72-75.

Griffiths, TA. 1982. Systematics ofthe New World nectar-feeding bats (Marnmalia, Phyllostornidae), based on the morphology of the hyoid and lingual regions. American Museum Novitates, 2742: 1-45.

Hererra, L. G., Flerning, TH., Findley, lS. 1993. Geographic variation in carbon composition ofthe pallid bat, Antrozous pallidus, and its dietary implicaitons. Journal ofMammalogy, 74(3):601-606.

Howell, D. 1979. Flock foraging in nectar-feeding bats: advantages to the bats and to the host plants. American Naturalist, 114:23-49.

Jimenez, P. & Pefaur, J. 1982. Aspectos sistematicos y ecologicos de Platalina genovensium (Chiroptera: Marnmalia). Actas del Octavo Congreso Latinoamericano de Zoologia, pp. 707-718.

Koopman, K. F. 1978. Zoogeography ofPeruvian Bats with special emphasis on the role ofthe Andes. American Museum Novitates, 2651: 1-33.

Linares E.L. 1992. Flora de la zona comprendida entre Yura y Chivay (2600-4800 m.s.n.m.) Caylloma-Arequipa. Bachelors Thesis, Universidad Nacional de San Agustin, pp. 200.

Norberg, D.M. 1990. Vertebrate flight. Mechanics, physiology, morphology, ecology, and evolution. Pp. 237-255. Springer-Verlag, Berlin, 291 pp.

Norberg, D.M., & Rayner, lM. 1987. Ecological morphology and flight in bats (Marnmalia; Chiroptera): wing adaptations, flight performance, foraging strategy and echolocation. Philosophical Transactions. ofthe Royal Society ofLondon B.,316:335-427.

Oficina nacional de evaluacion de recursos naturales. 1976. Mapa ecologico del Peru. Guia explicativa. ONERN. Lima, 76 pp.

Ortiz de la Puente, lO. 1951. Estudio monografico de los quiropteros de Lima y alrededores. Publicaciones del. Museo de Historia Natural Javier Prado, Universidad Nacional Mayor de San Marcos, 7: 1-48.

Sahley, C.T Pollination biology and breeding system of a columnar cactus in southwestern Peru. Chapter 3, this volume. 30

Sanbom, c.c. 1936. Records and measurements ofNeotropical bats. Field Museum Natural History Zoological Series, 20:93-106.

Sarmiento, G. 1975. The dry plant formations of South America and their floristic connections. Joumal ofBiogeography, 2:233-251.

Soriano, P.I, M.S. Sosa and O. Rossel. 1991. Feeding habits ofGlossophaga longirostris Miller (Chiroptera: Phyllostomidae) in an arid zone ofthe Venezuelan Andes. Revista de Biologia Tropical 39(2):263-268.

Thomas, O. 1928. A new genus and species of glossophagine, with a subdivision of the genus Choeronycteris. The Annals and Magazine ofNatural History London, 1:120-123.

Tuttle, M. 1970. Distribution and zoogeography ofPeruvian bats, with cornments on natural history. The University ofKansas Science Bulletin, 59:45-86.

Van den Bussche, R.A. 1992. Restriction-site variation and molecular systematics of New World leaf-nosed bats. Joumal ofMammalogy, 73(1):29-42.

Weberbauer, A. 1945. El mundo vegetal de los Andes Peruanos. Ministerio de Agricultura, Lima, pp. 776.

Winkelman, IR. 1971. Adaptations for nectar-feeding in glossophagine bats. Ph.D. Dissertation, Univ. Michigan, Ann Arbor, 121 pp.

Wilkinson, L. 1991. SYSTAT: The system for statistics. SYSTAT, Inc., Evanston, IIIinois.

Zeballos-Patron, H. & Davila-Florez, I 1993. Fauna del cerro San Ignacio (Tiabaya- Arequipa): Vertebrados, su relacion con el ecosistema agricola. Boletin del Museo de Historia Natural. Universidad Nacional de San Agustin de Arequipa, 1:1-8. 31

Table. l. --Average morphological measurements for P. genovensium individual s captured between 1991 and 1993. T-tests were used to compare adult males and females.

Age class

Subadults Adult females Adult males (pregant females excluded)

Mass (g) 16.92 20.24** 19.15 (SD=1.38, (SD=1.66, n=38) (SD=1.43, n=25) n=28)

Forearm length 49.16 50.78** 49.92 (rnm) (SD=.93, n=27) (SD=1.33, n=38) (SD=1.17, n=28)

Wing span (cm) .356 (n=3)

Wing are a (m2) .0202 (n=J)

Wing loading 9.4 (N/m2)

** Significant at r<.O1

31 32

Table 2.--Census data and roost sites for Platalina genovensium which were used to estimate total population size in late 1991 and early 1992. Data presented here were recorded prior to bat migration and are used to estimate population size prior to the drought. Although Batolito mine #2 was not censused after 1990, for purposes of our estimate we assumed it continued to house 5 bats. We also assumed that bats from the flooded Batolito tunnel #1 remained in the area until 1992.

Roost site Date of census Noof bats

Batolito mine #2 April1990 5 (not censused post-1990)

Batolito mine #3 January 1992 10

Poderosa mine December 1991 5

Rescate mine January 1992 2

Charcani tunnel December 1991 50

Coricancha mine January 1992 10

Batolito tunnel # 1 March 1990 5 (flooded in 1991) 33

FIGURE LEGENDS

Figure 1.a. Bat census and sex composition data for Batolito mine #3 from 1990-1993. Bars represent the number of individuals captured according to sex, while diamonds indicate number of individual s present as estimated by visual counts. Months when only visual counts were made lack sex composition data. The Batolito site was occasionally visited by bat collectors.

Figure l.b. Census and sex composition data for the Charcani colony site. Data from Charcani roost was obtained from late 1991 through early 1994. The Charcani roost was located on the premises of a hydro-electric facility; thus access to this site was restricted and bats were undisturbed.

Figure 2. Flower and mature fruit production of Weberbauerocereus weberbaueri in relation to rainfall, October 1991-0ctober 1993. Phenology data were taken at the main study site, 25 km. south of the city of Arequipa. Phenology was recorded for forty plants.

Figure 3.a. Age and sex composition ofbats at the Batolito mine #3.

Figure 3.b. Age and sex composition ofbats at the Charcani roost

Figure 4.a. Potentially reproductive males and females, Batolito mine #3.

Figure 4.b. Potentially reproductive males and females, Charcani roost.

Figure 5. Reproductive status of captured females. No reproductive activity was observed in 1992. In 1990 reproductive activity occured in late winter, spring, and summer. cEo"T1 ..•e No. of individuals No. of individuals (1) -"-"NNWW~~Ul -" -"N N WW ~ -' OUlOUlOUlOUlOUlO O Ul O Ul O Ul O Ul O C" 3/90--=::1 C» 12/91 ~ ~ 4/90

4/92 • ~ 5/90 I t ./ 8/90 3 10/92 Ir 1/91 DI iD en I 1/92 12/92 ...•.I 3/92 (l) e e 3 DI 93 DI 4/92 DI r+ r+ iD 1/ (1) en (1) 1 10/92

9/93 r" 11/92 --+1 O ~.. \ 12/92 ::::l 10/93 • O r/A I 1/93

8/93 1/94 I f 9/93 3/94 11 10/93

W ~ :!! G') Mean # Flowers and Fruits per e ::o Plant m NN (¡J (J1 (J1 (J1 N o o o o Oct-91 •. _ ..• Nov-91 -- 11I ",.. ",.. ~- . ",.. Dec-91 !I~ ,'/ Jan-92 I• , Feb-92 •• s: Mar-92 " \ 11I -+ ..,c: Apr-92 . ID ..,...•. c: May-92 ;::¡: (/) I e Jun-92 1» ••• I (1) Jul-92 "'TI - sO" Sep-92 ..,ID (/) Oct-92

Dec-92 t Jan-93 ::ti 11I ~...•. Apr-93 -tt 11I Aug-93 , " /"• Sep-93 . / ••• • I Oct-93 • • o N (¡J ~ (J1 m -..J o o o o o O o Rainfall (mm) :!! G') e No. of individuals No. of individuals ::c o •....NV• .l ~ VI ~ ~ 00 10 m •....• •....• N N w o VI o VI o VI •....• I I N C" --00 DJ -1-0 3/17/90 •....• 1!! : : ,...... • 3/18/90 ~ --~ --10 N 4/10/90

O 10 » --N Q. 10 5/12/90 3 -1-0 DI N

~ •....• » N 8/18/90 Q. e ...•. S» -•N....-• e ID ..•. s» (1) --10 ..•. 3 N (1) 1m 12/1/90 L •....• 3 --•....• DI iD VI --10 1/19/91 [].... V.l ...•.L ID 10 3 •....• 3/21/92 -•....-• --10 V.l 1/13/93 •....• -o- oN --10 V.l 10/19/93

W al 37

"'C •..el) 14 T a ::::1 +'c. 12 ea CJ 10 eaen 8 ::::1 "'C 6 .:; 4 ~ e 2 o O Z o o o N MM 0\ ~ ~ 0\ 0\ 0\ 0\ 0\ •.... --00 --o N --00 -- --•.... -- --M 0\-- •.... •.... •.... •....• N •.... •.... N--•..... •.... M-- ---.:t --V') 00 -- --M -- •--o....• Date

b 10 eaen ::::1 8 :2"'C .-> •..el) 6 "'C ::::1 .-e +' c. ~ ea 4 o CJ o 2 Z O 12/8/91 4/4/92 9/29/92 12/21/92 1/15/93 9/11/93 10/20/93 Date

O rep. male mi rep. female

FIGURE 4 oeld m 6aJd 13 daJN o

ateo ...- o I.-l ...- 00 VI .¡:.. I.-l N -...-- --N -...-- --...- --...- --...- \O ...- --...- 00 N o 00 -- -\O- -- -\O- -\O- ~ N 8 8 8 o o a s 2 o al •••o

(1)••• ~I 3 1» (D al en

~l CHAPTER3

POLLINATION BIOLOGY AND BREEDING SYSTEM OF A COLUMNAR

CACTUS IN SOUTHWESTERN PERU

ABSTRACT

1 present data on the pollination biology, breeding systern, and genetic diversity ofthe colurnnar cactus Weberbauerocereus weberbaueri at a 2500 m site in southwestern Peru. Weberbauerocereus weberbaueri is a self-cornpatible, auto- tetraploid cactus that is visited and pollinated by one species of rare endernic bat,

Platalina genovensiurn, and two species of hurnrningbirds, Patagona gigas and

Rhodopis vesper. W. weberbaueri exhibits pronounced inter-plant variation in floral color and size, and flowers exhibit traits corresponding to both bat and hurnmingbird pollination syndrornes. Initial fruit-set frorn pollinator exclusion experirnents conducted in 1991 and 1993, just prior to and after the El Niño event of 1991-92, revealed that bats were the rnost irnportant pollinators in 1991, but that hurnrningbirds were rnost irnportant in 1993. In both years, however, autogarny and lepidopteran larval infestation of fruits reduced differences in rnature fruit production arnong pollinator exclusion treatrnents so that differences in rnature fruit-set were not statistically significant. Reduced bat pollination in 1993 is attributed to the ernigration ofbats frorn the study site during a drought caused by El Niño. 1 hypothesize that interaction arnong several factors, including tetraploidy, autogarny, larval infestation of developing fruits, and variation in pollinator abundance due to El Niño events, rnay not result in strong

39 40 selection for a bat versus hurnmingbird floral morph, thus allowing the persistence of floral polymorphism in this cactus.

INTRODUCTION

Nectar-feeding bats and hummingbirds are important pollinators of several hundred species ofplants in the Neotropics (Vogel1969, Stiles 1981). In arid habitats, pollination by these vertebrates may be especially cornmon, because a large proportion of plants belonging to the Cactaceae and Agavaceae exhibit floral adaptations for bat and/or hummingbird pollination (Vogel1969, Porsche 1939). Several recent field studies in Mexico, Curacao, and Venezuela have experimentally demonstrated the importance ofbats and hummingbirds as pollinators of colurnnar cacti (Fleming et al. in review, Nassar 1991, Petit in press). Traits associated with bat pollination ofcacti include large, funnelform, nocturnal, white, and pungent smelling flowers that produce large amounts of pollen and nectar. "Hummingbird" flowers, on the other hand, tend to be tubular, red, diurnal, and odorless, but may also produce copious nectar and large amounts of pollen (Faegri & Van der Pijl 1979, Helverson 1993). Perhaps because bat and hurnmingbird floral syndromes share several characteristics, evolutionary transitions between these syndromes within a plant lineage may occur relatively frequently (Helverson 1993). In fact, several plant families (including the Cactaceae), genera, and even species, exhibit floral traits corresponding to both bat and hummingbird pollination syndromes (e.g., Sazima et al. 1994, Kress & Stone 1993,

Helverson 1993, Barthlott & Hunt 1993, this study). 41 The purpose ofthis study was to investigate the importance ofbat and hummingbird pollination in Weberbauerocereus weberbaueri (tribe Trichocereeae), a columnar cactus occurring in southwestern Peru that exhibits floral traits associated with both bat and hurnrningbird pollination. The genus Weberbauerocereus currently contains five species which produce flowers that exhibit traits primarily conforrning to the chiropterophilous syndrome; some species also display floral traits associated with hummingbird pollination (Hunt 1992, Backeburg 1976). W. weberbaueri exhibits considerable within-population variation in several floral morphological characters including size, shape, and color. Among-plant variation in floral color ranges from bright pink-red to white. This variation in floral color, along with the presence oflong, stout, funnelform-tubular corollas and both diurnal and nocturnal nectar production, suggests adaptation to both bat and hummingbird pollination. Prior to this study, however, the relative importance ofbat and hurnrningbird pollination for fruit production in W. weberbaueri was unknown.

To elucidate possible ecological and genetic mechanisms responsible for the presence of floral morphological variation in this cactus, 1 addressed the following five questions: 1) What is the morphological and genetic diversity of W. weberbaueri at the study site? 2) What is the extent of outcrossing versus selfing occurring within the study population? 3) What is the relative importance ofbats, hummingbirds, and insects as pollinators ofW. weberbaueri and is the importance of pollinators constant between years? 4) What factors, in addition to effective pollination of flowers, influence successful development of maturing fruits? Finally, because this study 42 coincided with the occurence of the 1991-92 El Niño southern oscillation, 5) How did

climatic fluctuations associated with El Niño influence the pollination biology of this

cactus?

STUDY SITE AND STUDY ORGANISMS

The main study site is located in the department of Arequipa, Peru, 25 km south

ofthe city of Arequipa (16.27° S, 71.300W) on the western slopes ofthe Andes, at

approximately 2500 m elevation. The site corresponds to the subtropicallow-montane

desert scrub life-zone according to the Holdridge system (ONERN 1976). Average

annual rainfall for this life zone is 222 mm per year. Between 1990 and 1993, however,

a drought caused by an El Niño event resulted in a drastic decrease in rainfall at the site

(Fig. 1).

Weberbauerocereus weberbaueri is the most abundant columnar cactus in the

region around the city of Arequipa (Aragon 1982, pers. obs.). Although the population

around Arequipa has been classified as three species by Ritter (1981), the continuous variation among plants in flower morphology and color and the lack of any diagnostic

alleles among morphs, has led me to consider them as one highly polymorphic species,

W. weberbaueri, in accordance with Hunt (1992).

Flowers ofW. weberbaueri are hermaphroditic, begin opening between 1500

and 1700 hours, remain open throughout the night, and generally close by 1100 the following morning. Nectar production is continuous while the flower is open (see

below). Although flower and fruit production peaks in September and October, some 43 flower and fruit production occurs year-round (C. Sahley, chapter 1).

Weberbauerocereus weberbaueri is visited and pollinated by the rare and endemic nectar-feeding bat, Platalina genovensium (Phyllostomidae), as well as by two species ofhurnmingbirds, Patagona ~ and Rhodopis vesper (Trochilidae)(pers. obs.).

METHODS

Nectar production, flower and fruit morphology

Nectar production ofW. weberbaueri flowers was measured in 1991 by bagging

29 flowers on 29 randomly selected plants prior to opening and using a one cc syringe to measure the amount of nectar produced at two-hour intervals until the flowers closed the following morning. Sugar concentration was measured using a Reichert-

Jung refractometer. The value obtained, in mg sugar/mg solution, was converted to mg sugar/rnl nectar as described in Bolten et al. (1979). Average sugar content was calculated for each two hour interval.

My observations indicated that variation in floral morphs was among, rather than within plants, so I collected one flower from each of 97 cacti in 1991 and one flower from 22 additional cacti in 1993. I recorded outer corolla tube length, outer corolla tube width at both widest and narrowest points, petallength, stigma exertion, and style length. The number of stigma lobes, corolla tube color and petal color were also recorded. In 1993 I also recorded ovary length, width, and inner corolla tube length and width. Petal color was scored in one offour categories: white, white- reddish/brown, maroon, and bright pink-red. 44 1 collected one mature fruit from each of 115 plants. 1 recorded width and length of all fruits and external fruit color of 69 fruits. Seeds were counted from 14 open-pollinated fruits, and individual seed weights were recorded for five seeds per fruit for twenty-six fruits.

Genetic diversity and mating system

To sample genetic diversity in the study population, flower bud tissue was collected from 25 adults in 1993. Bud tissue was crushed with liquid nitrogen prior to addition ofPVP extraction buffer. For the mating system analysis, one mature fruit was collected from each of 31 plants. Seeds were germinated on moist filter paper in petri dishes inside growth chambers. Germination usually occurred within 10-14 days.

Six to 24 seedlings at the cotyledon stage (approximately three weeks old; mean # seedlings per fruit =14) per fruit were prepared for electrophoresis by crushing with a

PVP extraction buffer (see Mitton et al. 1979).

Starch gel electrophoresis was performed on prepared tissues using buffer systems and staining recipes ofSoltis et al. (1983). Eight enzyme systems with 15 loci were assayed for the bud tissues. Of these, one probable monomorphic (Pgm-l) and two polymorphic loci (Fe-2, Aat-l) were not consistently scoreable and were not used in analyses. The enzyme systems and loci used were: phosphoglucoisomerase (Pgi-1,

Pgi-2), alcohol dehydrogenase (Adh), phosphoglucomutase (Pgm-2), mannose reductase (Mnr-l), flourescent esterase (Fe-l), triose-phosphate isomerase (Tpi-l, Tpi-

2, Tpi-3), leucine-amino peptidase (Lap), 6-phosphogluconate dehydrogenase (6PGD), and aspartate amino transferase (Aat-2). Buffer system 8- was used for Fe, Mnr, and 45 Aat, system 11 for Pgm and 6PGD, and system 34 for Tpi, Adh, and Lap. Adult genotypes were seored for aII ofthe above loei. Only Adh-I, Lap, Tpi-3, Aat-I, Mnr, and 6PGD were used in mating system analyses. Maternal genotypes were inferred from the genotypes of the seedlings.

Eleetrophoretie data from flower bud tissue eolIeeted from adult plants were used to ealculate the following genetie parameters as in Harnriek and Godt (1990): pereentage polymorphie loei, mean number of alleles per loeus, and genetie diversity.

Genetie diversity (Hep) was ealculated for eaeh loeus as Hep = l-Ip¡2, where Pi is the frequeney of the ith allele. An overall mean Hep was obtained by averaging over all loei.

To determine the degree of outrossing versus selfing in the population, genotypie data from 416 seedlings of 31 open-pollinated fruits were analyzed using the multi-allelic version ofthe TETRAT program written by RitIand (1990) whieh is fully deseribed in Murawski et al. (1994). The program inferred maternal genotypes from their progeny arrays and ealculated population level multiloeus and single loeus '. estimates of outerossing rates. The program also ealculated standard errors of the estimates using the bootstrap method. Fifty among-farnily (i.e., fruit) bootstraps were used for population estimates.

Pollinator exclusion experiments

Pollinator exclusion experiments were eondueted from Septernber-Novernber in

1991 and 1993. Replieation of experiments in 1992 was not feasible due to the extremely low flowering frequeney ofW. weberbaueri that year. Flowers from 48 46 plants in 1991 and 51 plants in 1993 were subjected to six pollination treatments in a randomized block design, so that when possible, each plant underwent all six treatments. Because daily flowering frequency is low in W. weberbaueri, generally only one treatment per night per plant was applied; exclusion experiments thus took several weeks to complete. In a few instances, more than one treatment per night was applied to a plant. This occurred when it appeared that not enough flowers would be available in the future to complete the experiment. The pollinator exclusion treatments applied in 1991 and 1993 were: 1) Open control, which allowed access to all flower visitors;

2) Total exclusion, which prohibited access to all visitors; 3) Diurnal exclusion, which permitted bats and nocturnal insects access to flowers, but excluded diurnal visitors such as hurnmingbirds and insects; 4) Nocturnal exclusion, which permitted hurnmingbirds and diurnal insects access to flowers but excluded nocturnal visitors;

5) Bat and diurnal exclusion, which permitted access to nocturnal insects only; and

6) Hummingbird and nocturnal exclusion, which permitted access to diurnal insects only (Table 1). Materials used to exclude potential pollinators consisted of either 1 inch hexagonal wire mesh (used to exclude vertebrates, but not insects) or bridal veil netting (used to exclude all visitors). For treatment (1), no excluders were used, while for treatment (2) (total exclusion) bridal veil netting was placed over the flower prior to opening and removed after the flower closed the following day. The diurnal exclusion treatment consisted of putting bridal veil netting over the flower in the late afternoon prior to opening, removing the netting at sunset (between 1740 and 1800), and replacing it again just before sunrise (between 0445 and 0515). The nocturnal 47 exclusion treatment consisted of placing bridal veil netting over flowers after opening at sunset and removing the netting the following morning just prior to sunrise. Treatment

(5), the bat and diurnal exclusion, involved placing bridal veil over flowers prior to opening and replacing the bridal veil netting with wire mesh at sunset. Before sunrise, the wire mesh was again replaced with bridal veil netting, which was removed the following day after flowers had closed. Treatment (6), the humrningbird and nocturnal exclusion, involved placing wire mesh excluders over flowers prior to opening and replacing the wire mesh excluders with bridal veil netting at sunset. Before sunrise, the bridal veil netting was replaced with wire mesh, which was removed later that day after flowers had closed.

AlI cacti used in the pollinator exclusion experiments were tagged and numbered. Each experimental flower was also individually tagged and numbered using soft alurninum tags inserted into the cactus 2 cm below the flower. Each flower was censused once a week until the fruits matured. I recorded whether flowers and fruits were on or off the plant, whether fruit development had initiated (indicated by swelling of ovaries one week after treatrnent; this was used as an indication of effective pollination), and presence or absence oflepidopteran larval infestation. After fruits began to develop, 1/4" X 114" inch wire mesh was placed over them to prevent fruit predation by vertebrates. In 1991 all undamaged, mature experimental fruits were collected and their seeds were stored in labeled envelopes. Seeds frorn 14 open- pollinated fruits were counted and weighed, and all seeds frorn the other treatments were weighed. In 1993 fruit-set was recorded, but fruits were not collected. ------,------

48 Statistical analyses

Data from the pollinator exclusion experiments was analyzed using SYST AT

version 5.3 (Wilkinson 1991) and SAS (SAS Institute 1989). Differences among

treatments in the frequency of fiuit initiation and fruit maturation were analyzed using

an experimentwise likelihood ratio r} statistic (G- statistic; Sokal & Rolf 1981) for

each year and for both years combined. Individual treatment G-statistics were also

calculated for each year and both years combined following the procedure in Sokal and

Rohlf (1981). The G-statistic was also used to analyze differences in fruit initiation

frequencies and the proportion of mature fruit set between years as well as between-

year differences in frequency of larval infestation and fruit mortality due to larval

infestation.

Seeds from mature fruits collected in 1991 were weighed to the nearest .0001

g. A Pearson product-moment correlation coefficient was calculated to determine the

relationship between total seed weight and seed number per fruit for the 14 open-

pollinated fruits. A one-way analysis of variance was performed to analyze differences

in total seed weight per fruit among treatments. Because treatments (5) and (6) did not

differ significantly from the total exclusion treatment for both fruit initiation and mature

fruit set in either year or in both years combined, seeds from these treatments were

po oled with the total exclusion treatment in order to increase sample size. 49

RESULTS

Nectar production, flower and fruit morphology

W. weberbaueri flowers were tubular to slightly funneiform in shape; flowers were long, ranging from 5.2 cm to 10.3 cm, and corolla tube width was narrow, especially just above the nectary (Table 2). The most variable floral morphological characters were petallength, ovary length, style length, and length of stigma exsertion

(Table 2). Despite variation in style and stigma exsertion length, most flowers (82%) exhibited some degree of exsertion. Petal color among plants ranged from white to bright pink-red, with white, and white with red-brown petal colors predominating (Fig.

3a). Only 20% ofthe plants had either maroon or pink-red flowers. Within-plant flower color was constant and did not exhibit seasonal or annual variation. All flowers, regardless of shape or color, opened between 1500 and 1700 hours (prior to sunset) and remained open untillate morning (between 0900 and 1100). Nectar production was continuous while flowers were open. Average sucrose concentration ofnectar was approximateiy 22% but dropped in the late morning; nectar production was highest at night and also dropped in the morning (Fig. 2).

Mature fruits ranged in color from bright red to orange, yellow, and green (Fig.

3b). The most common colors were in the red to orange range (75%), followed by green (15%); yellow fruits were encountered relatively infrequently (4%). Seed number, and seed weight in W. weberbaueri were more variable than fruit length and 50 width (Table 3). The number ofseeds per fruit ranged from 181 to 1533; individual seeds on average weighed slightly less than .001 g.

Genetic diversity and mating system.

Banding patterns obtained from assays of all polymorphic enzyme systems in all individuals indicated that W. weberbaueri is an autotetraploid. Both balanced and unbalanced heterozygote banding patterns, corresponding to equal and unequal combinations of alleles, were present in adult and offspring tissues. Fixed heterozygotes, which would be expected to occur in allotetraploids, were never observed for any enzyme system.

Ofthe 12 resolvable loci assayed, eight (66.6%) were polymorphic. The number of alleles in adult heterozygotes ranged from two to four per locus per individual plant (Table 4); the mean number of alleles per locus was 2.25. Mean genetic diversity (Hep) ofadults in the study population was .257.

Analysis of the mating system of open-pollinated seedlings from 1991 indicated that population outcrossing levels were high based on both multi-locus (tm=.937, s.e.=.042) and single locus (ts=.869, s.e.=.047) estimates.

Pollinator exclusion experiments

Fruit initiation frequencies. Data on fruit initiation frequencies indicated that bats and hurnmingbirds effectively pollinate W. weberbaueri flowers but that flowers also produce fruits in the absence of flower visitors (Table 5). In 1991, fruit initiation frequencies among pollinator exclusion treatments differed significantly (G=11.177, d.f=5, P<.05). The diurnal exclusion treatment (bat pollination) accounted for a 51 significantly higher proportion offruits initiated (70%; G=5.534, d.f.=I, P<.05).

Remaining treatments were not significantly different from expected values. Forty

percent of flowers from which all flower visitors were excluded initiated fruit

development, indicating the presence of self-compatibility and autogamy in W.

weberbaueri.

In contrast to 1991, differences in fruit initiation frequencies among pollinator

exclusion treatments conducted in 1993 were not significant (G=8.671, d.f.=5, P>.05).

Examination of the contributions of individual treatments to the overall G-statistic

indicates that in 1993, hummingbird-visited flowers (nocturnal exclusion) resulted in a

significantly higher proportion (76%) offruits initiated (G=4.927, d.f.=I, P<.05)

relative to expected frequencies. As in 1991, a relatively high proportion of flowers

(48%) developed fruits in the absence offlower visitors.

When data from 1991 and 1993 are combined, experimentwise differences

among treatments were significant, with the nocturnal exclusion treatment

(hurnmingbird pollination) contributing a significantly greater proportion of developing

fruits (G=3.861, d.f.=I, P<.05) and the total exclusion treatment contributing a

significantly smaller proportion offruits (G=5.1099, d.f.=I, P<.05).

Mature fruit set. In 1991, the diurnal exclusion treatment (bat pollination) resulted in the largest proportion of mature fruits (48%), while in 1993, the nocturnal exclusion

treatment (humrningbird pollination) yielded the largest proportion (35%) ofmature fruits (Table 5). However, differences in mature fruit-set among treatments in both 52 1991 and 1993 were not significant at the experimentwise or individuallevel, nor were they significant when data for 1991 and 1993 were combined.

Total seed weight per fruit was highly correlated with the total number of seeds

(r=0.915, P<.OOI). However, differences in total seed weight for mature experimental fruits collected in 1991 were not significant among treatments (one-way ANOV A,

F=.727, P=0.539; Table 6).

Between year comparisons. In 1993, fruit initiation frequencies were significantly higher for the nocturnal exclusion treatment than in 1991 (G=3.920; d.f.=I, P<.05;

Table 7). However, differences in mature fruit set between years for the nocturnal exclusion treatment were not significant. The most marked difference between years was the decrease in mature fruit set by bat-visited flowers (diurnal exclusion) in 1993 relative to 1991 (G=9.243, P<.OOI). Despite the variability in bat and humrningbird pollination effectiveness, both fruit initiation frequencies and mature fruit set for open- pollinated flowers did not differ significantly between years. The proportion ofboth autogamous initiated and mature fruits was also constant from year to year.

Larval infestation of developing fruits

Infestation of developing fruits by lepidopteran larvae occurred in both 1991 and 1993 and had a pronounced effect on fruit survivorship (Fig. 4). Fruit mortality in developing, open-pollinated fruits ranged from 36% in 1991 to 61% in 1993; of these,

10% and 39% respectively, were positively identified as being due to larval infestation.

This difference in fruit mortality due to larval infestation between years was statistically significant (G=7.74, d.f=I, P=.005). Differences in fruit mortality among pollinator 53 exclusion treatments were not significantly different in 1991 or in 1993 (G=5. 051, d.f.=5, p=.410; G=3.689, d.f.=5, p=.595, respectively)

DISCUSSION

Flower shape, color, and nectar production ofW. weberbaueri at the study site are consistent with bat and hummingbird pollination syndromes. Results of this study indicate that both bats and hummingbirds visit and pollinate flowers ofW. weberbaueri.

Nevertheless, a large proportion (26%) offlowers from which all visitors were excluded set fruit, indicating the presence of self-compatibility and autogamy in this cactus. Mating system analysis, however, suggests that if flowers are visited, they are highly outcrossed. Diurnal and/or nocturnal insect visitation to flowers did not significantly contribute to either fruit initiation or mature fruit-set in either year.

However, differences in mature fruit-set among pollination treatments within ayear were statistically undetectable, as were differences in seed production. The importance ofbats as pollinators varied between years with the proportion ofbat-pollinated flowers being greater in 1991 than in 1993. In 1993, hummingbirds were the most important pollinators. Lepidopteran larval infestation and subsequent fruit mortality was greater in

1993 than in 1991. Despite variability in pollinator effectiveness as measured by fruit- initiation frequencies, mature fruit-set of open-pollinated flowers did not differ significantly between 1991 and 1993. 54

Genetic and morphological diversity

Polyploidy is frequent (35%-47% of species) in angiosperms (Stebbins 1971,

Grant 1971), and the Cactaceae are no exception. Pinkava et al. (1985) found that

27.9% of 551 cactus taxa analyzed were polyploid. Polyploidy has been hypothesized as having several advantages and/or consequences. For example, polyploidy enhances the capacity ofindividuals and populations to "store" genetic variation (Briggs &

Walters 1990). Several theoretical and empirical studies have shown that autotetraploids have higher heterozygosity relative to diploids (Moody et al. 1993,

Soltis & Soltis 1989, Wolfet al. 1990, Ness et al. 1989). Genetic diversity ofW. weberbauerocereus in the study population is high compared to diploid plants with similar mating systems, but is similar to estimates in other autotetraploid plants (Table

8).

A second advantage to polyploidy is the release of genetic variation and a consequent increase in ecological amplitude and geographical distribution relative to diploids (Briggs & Walters 1990, Brochmann & Elven 1992). Previous authors (Ritter

1981, Backeberg 1976) have noted the morphological variation present within and among populations belonging to the genus Weberbauerocereus but have assumed that this variation was due to interspecific differences. This study suggests that only one species is present in the vicinity of Arequipa and that much of the intra-populational variation may be because ofhigh heterozygosity caused in part by autotetraploidy. 55 The presence of morphological variation in flowers raises some intriguing questions that have yet to be addressed. For example, what is the origin of such variation? Do differences in floral characteristics among plants result in different preferences for a particular flower morph by bats and hummingbirds and subsequent diverging selection pressure for floral traits? Ifbats prefer a suite of characteristics including white flower color and copious nocturnal nectar production, this might indicate that at the main study site, W. weberbaueri plants are primarily adapted to bat- pollination. Another source ofvariation within this species could be past inter-specific hybridization; hybridization is a common phenomenon within the cactus family (Gibson

& Nobel 1986); however, this possiblity has yet to be investigated. Other species within the genus Weberbauerocereus also exhibit variation in flower color (Backenburg

1976), which suggests that similar ecological and genetic forces may be operating throughout the range of this genus in Perno

Mating system

Analysis ofthe mating system ofW. weberbaueri based on fruits collected in

1991 indicated that most seedlings were outcrossed (tm=.937); this was probably because both bats and hurnmingbirds were present that year and high visitation rates resulted in high levels of outcrossing. Unfortunately 1 did not estimate outcrossing rates in 1993, when the proportion ofbat pollination was low. Because the increase in homozygosity due to selfing is far less for a tetraploid than for a diploid (Moody et al.

1993), facultative autogamy and self-compatibility in W. weberbaueri may be 56 an effective strategy in dry El Niño years, when the most important pollinators, bats,

are absent.

The high degree of stigma exsertion exhibited by most tlowers likely decreased

the possibility of self-pollination during visits by bats or hummingbirds in 1991. Motten

and Antonovics (1992) examined outcrossing rates in Datura stamonium and found that

the most important determinant of outcrossing rate was the degree of stigma exsertion

exhibited by tlowers. Flowers with stigmas positioned above the anthers had higher

outcrossing rates than did those with overlapping stigmas and anthers. Results

obtained for the bat-pollinated Mexican columnar cactus ~ pringlei show that open-

pollinated hermaphrodite tlowers were found to have a low outcrossing rate (tm=.301)

even though high tlower visitation rates were observed (Murawski et al. 1993, Fleming

et al. 1994). Murawski et al. (1994) hypothesized that the high degree of selfing in the

presence oftlower visits was due to absence of stigma exsertion in tlowers ofP.

pringlei.

In addition to being able to produce fruits during especially dry years, another

potential advantage to autogamy was pointed out by Murawski et al. (1993). They

hypothesized that the tetraploid, bat-pollinated cactus P. pringlei. whose seeds may be

dispersed long distances by bats, may experience selection pressure for isolated

individual s to produce fruits via autogamy. Because W. weberbaueri is also bat-and bird-dispersed, such a scenario may hold true for this cactus as well. 57

Variation in pollinator effectiveness and larval infestation among years: the role ofEI

Niño

As in the rest of western Peru, the arid western slopes of the Peruvian Andes are subject to dramatic climatic changes associated with the El Niño southern oscillation (Rasmusson & Wallace 1983). Changes in sea level and temperature during this time cause considerable reduction in primary productivity of upwelling areas off the coast ofPeru that result in the elevation ofmortality rates and disruption ofbreeding schedules in seabirds and marine mammals (Castro 1986, Schreiber & Schreiber 1984,

Barber & Chavez 1983). However, except for a study that exarnined the influence of

El Niño on breeding schedules of Andean condors, (Wallace & Temple 1988), the effects that such inter-year variability in climate have on terrestrial ecological interactions in Peru have previously received little attention.

A major finding ofthis study was that mature fruit production attributable to bat pollination decreased significantly in 1993 relative to 1991 because of emigration of bats from the study area. The 1991-93 El Niño event led to three years of low rainfall at the study site and a rainless period which spanned 17 months. Fruit and flower production ofW. weberbaueri dropped considerably during this time as did bat abundance (Sahley, chapter 1). The reduction in flower and fruit availability in 1992 may have caused the nectar-feeding bat P. genovensium to ernigrate from the area, which resulted in fewer bat-pollinated fruits in 1993 (Sahley, chapter 1). Surprisingly, after fruit and flower production increased in 1993, bat populations remained low 58 relative to 1991. This suggests that either bats suffered high mortality rates during this time or have an irregular rnigratory schedule. As of mid-1994, bat populations had still not returned to levels observed in 1990 and 1991 (Sahley chapter 1). If, as the data suggest, bats abundance decreases near Arequipa during and after El Niño years, then

W. weberbaueri flowers may be subject to shifting selection pressures wherein most years bats are the most important pollinators, but during and after El Niño years hurnmingbirds may be most important.

El Niño events occur irregularly and it has recently been proposed that this irregularity may be due to low-order chaotic behavior (Tziperman et al. 1994), although on average, El Niño events happen every four years (Cane 1983). Dramatic shifts in rainfall availability at the study site because ofEI Niño events have implications not only for the pollination biology of columnar cacti, but likely have a large influence on variation in larval infestation frequency, fruit survivorship, seed dispersal and seedling establishment. Larval infestation and fruit mortality varied between years in this study, probably as a result of differences in lepidopteran abundance during and after the drought. After rains resumed in 1993 insects, especially moths and butterflies, were observed more frequently than they had been in late 1991 and 1992 (pers. obs.).

While I did not study seed dispersal and seedling establishment ofW. weberbaueri, my observations indicate that very little or no seedling establishment occurred during the drought. Parker (1993) exarnined long-term regeneration trends ofcolumnar cacti in the northern Sonoran desert and found that establishment of cacti was sensitive to variability in climatic patterns. Thus, El Niño events, by influencing rainfall, plant 59 phenology, and pollinator abundance, have played an important role in the evolution of this pollination mutualism.

Maintenance ofvariation in W. weberbaueri

An important result of this study was that although bat and hummingbird pollination treatments resulted in significantly higher fruit initiation frequencies than the other four treatments, differences in mature fruit set among treatments within ayear were not significant. In addition to the small differences in mature fiuit-set among treatments, it is likely that autogamy limited differences among treatments. Also, as fruits matured, many were subject to larval infestation and subsequent fruit death. This further reduced differences among treatments. These results suggest that even ifbats

and hummingbirds do prefer different flower types, in any given year selection for a bat versus hummingbird flower morph may be weak.

1 propose that interaction among several factors, including variation in pollinator importance among years, low within-year differences in pollinator importance, and high levels of genetic variation in part due to autotetraploidy, have led to weak selection pressure for any one flower morph and have allowed the persistence

ofmorphological variation in the population ofW. weberbaueri that 1 studied. El Niño

events play an important role in influencing the pollination biology of this cactus by

causing drought conditions and reducing the abundance of important pollinators such

as nectar-feeding bats. Data on the relationship between plant morphology, visitor

preference, and pollinator effectiveness at this and other sites within the range ofW. 60 weberbaueri would help to evaluate this hypothesis. In addition, phylogenetic studies of the genus Weberbauerocereus as well as other members of the tribe Trichocereeae would be extremely helpful for understanding the evolution ofthe mutualism among bats, hummingbirds, and columnar cacti.

ACKNOWLEDGEMENTS

M. Donnelly, T. Fleming, 1. Hamrick, D. Janos, C. Horvitz, K. Lips, M. Nelson,

K. Waddington, and members ofthe plant-animal interactions graduate discussion group ofthe University ofMiami provided valuable discussion, assistance, and cornments. L. Baraybar provided assistance in the field throughout the duration ofthe study. 1. Estefanero and 1. Ugarte provided assistance at the outset ofthe study. P.

Jimenez gave important logistic advice. E. Linares assisted with plant identification. P.

Baldizan graciously permitted the author to conduct research on his property. 1.

Harnrick ofthe University ofGeorgia kindly allowed me the opportunity to work in his laboratory. Funds for this research were provided by grants from New York

Zoological Society- The Wildlife Conservation Society, a World Wildlife Fund-Garden

Club of America Scholarship in Tropical Botany, Bat Conservation International, and a

Maytag Graduate Fellowship from the College of Arts and Sciences, University of

Miami. 61 LITERA TURE CITED

Aragon, G.A. 1982. Cactaceas de los alrededores de la ciudad de Arequipa. Boletin de Lima, 4(21):86-94.

Backeberg, C. 1976. Cactus Lexicon. Blandford Press, Poole, England.

Barber, T.T. & Chavez, F.P. 1983. Biological Consequences ofEI Niño. Science 222: 1203-1210 ..

Barthlott, W. & D.R. Hunt. 1993. Cactaceae. In: Families and genera ofvascular plants. ed. by K. Kubitzki. Springer Verlag, Berlin. Pp. 161-196.

Briggs, D. & Walters, S.M. 1990. Plant variation and evolution. Cambridge University Press, Cambridge.

Brochmann, C. & Elven, R. 1992. Ecological and genetic consequences of polyploidy in arctic Draba (Brassicaceae). Evolutionary Trends in Plants. 6(2): 111-124.

Cane, M.A. 1983. Oceanographic events during El Niño. Science 222(4629):1189- 1194.

Castro, G. 1986. Repercusiones biologicas del fenomeno "El Niño" en la costa Peruana. Boletin de Lima 44:71-79.

Faegri K. & L. Van der Pijl. 1979. The principies of pollination ecology. Pergamon Press, Oxford.

Fleming, T.H., Horner, M. & M.D. Tuttle. In review. The relative effectiveness of diurnal and nocturnal pollinators to three species of colurnnar cacti. Oecologia. Southwest American Naturalist.

Gibson, A.c. & P.S. Nobel. 1986. The Cactus Primer. Harvard University Press, Cambridge, USA.

Grant, V. 1971. Plant speciation. Columbia University Press. New York & London. 62 Hamrick, J.L & M.J. Godt. 1990. Allozyme diversity in plant species. Pages 43-63 in: H.D. Brown, M.T. Clegg, AL. Kahler, & B.S. Weir, editors. Plant population genetics, breeding, and genetic resources. Sinauer Associates Inc. Sunderland, USA.

Helverson, O.V. 1993. Adaptations of flowers to the pollination by glossophagine bats. Pages 41-59 in: W. Bartholott et al. Plant-Animal Interactions in Tropical Environments. Museum Alexander Koenig, Bonn.

Hunt, D. 1992. CITES Cactaceae Checklist. Royal Botanical Gardens, Kew, UK.

Kress, W. J. & D .E. Sto ne. 1993. Morphology and Floral Biology of Phenakospermum (Strelitziaceae), an arborescent herb ofthe Neotropics. Biotropica 25(3):290-300.

Mitton, J.B., Linhart, Y.B., Sturgeon, B.K., and J.L. Hamrick. 1979. Allozyme polymorphism detected in mature needle tissue of ponderosa pine, Pinus ponderosa. Laws. J. Heredity, 70:86-89.

Moody, M. E., L.D. Mueller, & D.E. Soltis. 1993. Genetic variation and random drift in autotetraploid populations. Genetics 134:649-657.

Motten, AF. & 1. Antonovics. 1992. Determinants of outcrossing rate in a predominantly self-fertilizing weed, Datura stramonium (Solanaceae). Am. J. Botany 79(4):419-427.

Murawski, D.A, Fleming, T.H., Ritland, K., & 1.L. Hamrick. 1993. Mating system of Pachycereus pringlei: An autotetraploid cactus. Heredity 72:86-94.

Nassar. 1. 1991. Biologia reproductiva de cuatro cactaceas quiropterofilas Venezolanas (Cereae: Stenocereus griseus, Pilosocereus moritzianus, Subpilocereus repandus y ~ horrispinus) y estrategias de visita de los murcielagos associados a estas. Bachelors Thesis, Universidad Central de Venezuela. Caracas, Venezuela.

Ness, B.D., D.E. Soltis, & P.S. Soltis. 1989. Autoployploidy in Heuchera micrantha (Saxifacaceae). American Journal ofBotany 76(4):614-626.

Oficina Nacional de Evaluacion de Recursos Naturales. 1976. Mapa ecologico del Perno Guia explicativa. ONERN. Lima, Perno 63 Oficina Nacional de Evaluacion de Resursos Naturales. 1974. Inventario, evaluacion y uso racional de los recursos naturales de la costa: Cuencas de los rios Quilca y Tambo, Vol.3. ONERN. Lima, Peru.

Parker, K.C. 1993. Climatic effects on regeneration trends for two columnar cacti in the northern Sonoran deserto Annals of the Association of American Geography 83(3):452-474.

Petit, S. 1995. The effectiveness oftwo bat species in the pollination oftwo species of columnar cacti on Curacao. Ph.D. dissertation, University ofMiarni.

Pinkava, DJ., M.A. Baker, B.D. Parfitt, M.W. Nohlenbrock, & RD. Worthington. 1985. Chromosome numbers in some cacti of western North America- V. Systematic Botany 10:471-483.

Porsche, O. 1939. Das Besaubugsleben der Kadkteenblute. Jahrb. Deutsche Kakt-Ges.

Rasmusson E.M. & lM Wallace. 1983. Meteorological aspects ofthe El Niño Southern Oscillation. Science 222:1195-1202.

Ritland K.L. 1990. A series ofFortran computer programs for estimating plant mating systems. Journal ofHeredity, 81:235-237.

Ritter, F. 1981. Kakteen in Sudamerika. Band 4. F. Ritter Sebbstverlag, Spangenberg, Germany.

Sahley, C. Chapter 1. The natural history and population status ofthe nectar- feeding bat, Platalina genovensium (phyllostornidae: Glossophaginae) in southwestern Peru.

SAS. 1989. SAS version 6.08. SAS Institute. Carey, North Carolina ..

Sarmiento, G. 1975. The dry plant formations ofSouth America and their floristic connections. Journal ofBiogeography, 2:233-251.

Sazima, M., l., Sazima, & S. Buzato. 1994. Nectar by day and night: Siphocampylus sulfureus (Lobeliaceae) pollinated by hummingbirds and bats. Plant. Systematics and Evolution 191:237-246.

Schreiber, RW. & E.A. Schreiber. 1984. Central Pacific seabirds and the El Niño southern Oscillation: 1982 to 1983 perspectives. Science 225:713-716.

Sokal, RF. & F.J. Rohlf 1981. Biometry. W.H. Freeman & Company, New York, USA. 64 Soltis, D.E. & P.S. Soltis. 1989. Genetic consequences ofautopolyploidy in Tolmiea (Saxifacaceae). Evolution 43(3):586-594.

Soltis, D.E., Haufler, C.H., Darrow, D.C. and G.l Gastony. 1983. Starch gel electrophoresis offerns: a compilation of grinding buffers, gel and electrode buffers and staining schedules. Am. Fer. Journal, 73: 9-15.

Stebbins, G. 1971. Chromosomal evolution in higher plants. Amold Press, London, UK.

Stiles, F.G. 1981. Geographical aspects of bird-flower coevolution, with particular reference to Central America. Annals of the Missouri Botanical Garden 68:323-351.

Tziperman, L. Stone, M. Cane, H. Jarosh. 1994. El Niño chaos: Overlapping of resonances between the seasonal cycle and the Pacific Ocean-atmosphere oscillator. Science 264:72-74.

Vogel, K. 1969. Chiropterophilie in der Neotropischen flora. Neue Mitt. II, III Flora abt. B, 158: 185-323.

Wallace, M.P. & Temple, S.A. 1988. Impacts ofthe 198201983 El Niño on population dynamics of Andean condors in Perno Biotropica 20(2): 144-150.

Wilkinson, 1991. SYSTAT. The system for statistics. SYSTAT, Inc., Evanston, Illinois, USA.

Wolf, P.G., D.E. Soltis, & P.S. Soltis. 1990. Chlorplast-dNA and allozymic variation in diploid and autotetraploid Heuchera grossularifolia (Saxifragaceae). American Journal ofBotany 77(2):232-244. 65

Table 1. Pollinator exclusion treatments performed in 1991 and 1993 for Weberbauerocereus weberbaueri.

Treatment Pollination Treatment Description of flower Flower visitors number treatment allowed

1 Open No treatment Insects, bats, humrningbirds. Bridal veil during 2 Total exclusion entire time flower None open.

3 Diurnal exclusion Bridal veil netting Bats, nocturnal during day insects

4 Nocturnal exclusion Bridal veil netting Hurnmingbirds, during night diurnal insects

5 Diurnal and bat exclusion Bridal-veil netting N octurnal insects during day and wire mesh during night

6 Nocturnal and Bridal-veil netting Diurnal insects hurnmingbird exclusion during night and wire mesh during day Table 2. Morphological measurements for flowers ofWeberbauerocereus weberbaueri. Coefficients ofvariation are in parentheses.

Outer Outer Outer corolla Petal Style Stigma Number of Ovary Ovary corolla corolla tube width length length exertion stigma length width tube length tube width (narrowest) (cm) (cm) (rnm) lobes (rnm) (rnm) (cm) (widest) (cm) (cm)

Mean 7.57 2.20 1.45 2.08 7.49 7.2 11.49 7.02 7.4 (.11 ) (.12) (.11 ) (.15) (.15) (.79) (.12) (.32) (.13) n=96 n=71 n=69 n=95 n=92 n=92 n=51 n=21 n=22

Minimum 5.2 1.5 1.0 1.2 3.5 O 8 3.7 5.8

Maximum 10.3 2.9 1.9 2.9 10.5 2.5 15 14.9 9.1

0'1 0'1 67

Table 3. Fruit rnorphology ofWeberbauerocereus weberbaueri. Coefficients of variation are in parentheses.

Fruit length Fruit width Seed nurnber Individual (rnrn) (rnrn) per fiuit seed wt. is2

Mean 27.0 30.6 1067 .0009 (.15) (.14) (.39) (.31) n=115 n=113 n=14 n=133

Minirnurn 15.7 18.4 181 .0004

Maxirnurn 37.5 41.5 1533 .0020 68 Table 4. Allele frequencies of eight polymorphic loci for adult plants at the main study site. Tpi-1, Tpi-2, Pgi-1, Pgi-2 were monomorphic. Locus Allele Allele Genetic diversity designation frequency (Hep)

FE-1 1 .408 .704

3 .276

4 .197

5 .118

MNR 2 .927 .135

3 .073

PGM-2 2 .307 .425

3 .693

6P-2 2 .546 .516

3 .432

4 .023

AAT-2 3 .875 .219

4 .125

TPI-3 3 .890 .196

4 .110

ADH 1 .060 .267

3 .850

4 .080

5 .010

LAP 2 .227 .625

3 .546

4 .113

5 .114 Table 5. Proportion fruits initiated and fruits matured for Weberbauerocereus weberbaueri according to pollinator exc1usion treatment. G- statistics are in parentheses. 1991 1993 1991-93 combined

Treatment Flowers Fruits Fruits Flowers Fruits Fruits Flowers Fruit Mature (Flower visitors) initiated matured initiated matured initiation fruit set

Open 51 .55 .35 44 .64 .25 95 .59 .30 (.2086) (.0532) (.168) (.0145) (.2613) (.0552)

Total exc1usion 40 .40 .28 40 .48 .25 80 .44 .26 (Autogamy) (2.2048) (.7256) (2.813) (.0132) (5.1099)* (.3968)

Bats & 40 .70 .48 43 .60 .16 83 .65 .32 nocturnal insects (5.534)* (3.211) (.0004) (2.0058) (2.6120) (.2029)

Hummingbirds 41 .56 .37 46 .76 .35 87 .67 .35 & diurnal insects (.3172) (.144) (4.927)* (2.0902) (3.8610)* (1.194)

Nocturnal 32 .38 .31 39 .54 .26 71 .46 .28 insects (2.6060) (.0913) (.7394) (.0004) (2.785) (.0542)

Diurnal insects 30 .47 .20 42 .60 .26 72 .54 .23 (.3053) (2.77) (.0214) (.0034) (.1393) (1.222)

Experimentwise 234 11.177* 7.003 254 8.671 4.128 488 14.769* 3.127 values

en *p<.05 ~ 70

Table 6. Total seed rnass per fruit by pollinator exclusion treatrnent in 1991. Data frorn the no visit (autogarnous) treatrnent were pooled with bat and hurnrningbird exclusion treatrnents.

Pollinator exclusion treatrnent No.of Avg. total seed rnass per S.D. fiuits fruit llil Open 22 .838 .460

No visits (Autogarny) 23 .842 .550

Bat pollination 28 l.00 .406

Hurnmingbird pollination 15 .789 .292

j 71

Table 7. Fruit initiation and mature fruit-set compared between years for Weberbauerocereus weberbaueri. Numbers indicate G-statistics.

Comparison between years: 1991 vs. 1993

Treatment Flowers Fruits initiated Fruits matured

Open 95 .747 1.190

No visits 80 .457 .065 (Autogamy)

Diurnal exclusion 83 .832 9.243**

Nocturnal exclusion 87 3.920* .010

Nocturnal insect 71 1.236 .273 pollination

Diurnal insect 72 1.166 .077 pollination

*P<. 05; **P<. 01: (Likelihood ratio X2 for deviation of expected ratio of 1:1 for each pollination treatment)

...1 72

Table 8. Levels ofheterozygosity in W. weberbaeuri and other autotetraploid and anirnal- outcrossing diploid plants.

Plant species Heterozygosity Source

Autotetraploid species

Weberbauerocereus weberbaueri .257 This study (Cactaceae)

Pachycereus pringlei (Cactaceae) .209 Harnrick & Flerning (unpubl. data) Heucheria grossularifolia .159 (Saxifragaceae) Wolf et al. (1990)

Tolmiea menziessi .237 Soltis & Soltis (Saxifragaceae) (1989)

Diploid species

Mixed-animal outcrossing .120 Harnrick & Godt (mean value, N=85) (SD=.015) (1990) 73 FIGURE LEGENDS

Figure l. Rainfall data for 1962-1971 and 1981-1993 for the Characato weather station,

(2400 m above sea-level). This station is approximately 15 km from the main study site and occurs within the same life-zone. El Niño events occurred in 1963-65, 1969, 1972-

73,1976,1982-83, and 1991-93 (no rainfall data was available for 1969 and 1972-73 events). Rainfall data from 1961-1976 from ONERN, 1974.

Figure 2. Average nectar volume and concentration for W. weberbaueri flowers.

Data were recorded at two-hour intervals.

Figure 3a. Frequency distribution offlower colors for 113 individuals ofW. weberbaueri at the study site.

Figure 3b. Frequency distribution offruit colors for 69 individuals ofW. weberbaueri at the study site.

Figure 4. Frequency oflarval infestation and fruit mortality due to larval infestation of developing fruits in 1991 (n=28) and 1993 (n=28). 350

300

250 -E E 200 cu 1: 150 "¡¡ a: 100

50

01 I I III I I I I I I I I I II I I I ~ 62 63 64 65 66 67 68 69 70 71 81 82 83 84 85 86 87 88 89 90 91 92 93 Year

FIGURE 1

"~ 75

24 0.2 ~ - j - ....-••....- ..•, Q) 22 'r....---.. 0.18 tn / , 20 •O.. •• 0.16 (J 18 \ ::l \ 0.14 _ tn- tn 16 \ o e \ 0.12 _E ~--Q) 14 \ Q) oc-ea 12 0.1 .- > \ E ea .-::l \ ::l -••• C'" 0.08 'O 1: Q) 0.06 > Q) lB (J e ! 0.04 o o 2 0.02 O O 1700 1900 2100 2300 100 300 500 700 900 1100 -. - Conc. Time • Volume

FIGURE 2 76

45 a 40 ~en al 35 3: 30 o q::.•.. 25 c:: 20 al (J 15 ~ oal.. 10 5 O II white white- maroan pink-red red/brown Color category

b 30 I 25

en :!: 20 ;:, ~ .•...•.. c:: 15 al (J ~ oal.. 10

5

O Red Red- Orange Orange Orange green Yellow Green orange yellow Color category

FIGURE 3 77

60 1993

50 I 1993 en 1991 .~:;:, 40 ~ - '+- +'e 30 G) ~CJ G) Q. 20 I I 1991 10 - O Fruit Fruit mortality infestation

FIGURE 4 CHAPTER4

VERTEBRATE POLLINATION OF THE ORGAN PIPE CACTUS, STENOCEREUS

THURBERI: EFFECTS ON POLLEN DEPOSITION, POLLEN DISPERSAL

AND FRUIT PRODUCTION

ABSTRACT

I examined three components ofbat and hurnmingbird pollination ofthe organ-pipe cactus, Stenocereus thurberi, in Sonora, Mexico. I quantified the amount ofpollen deposited by bats and hummingbirds per flower visit on stigmas of conspecifics, examined variability in pollinator success within and among tlowering seasons, and determined levels of pollen-mediated gene flow. My results indicate that both bats and hummingbirds deposit large quantities of pollen on a per-visit basis with pollen-ovule ratios on stigmas after one flower visit being, on average, 9. 13 and 14.9 for bats and hummingbirds, respectively. Pollinator exclusion experiments indicate the proportion of fruits produced by hummingbird pollination was constant (x=14.5%) in both spring of 1992 and 1993, while that by bats was significantly greater in 1993 (17%) than in 1992 (3%). In 1993 I examined within-year variation in pollination success and found that fruit production attributable to both bats and hurnmingbirds dropped significantly in late June (0% and 3% for bats and hummingbirds respectively) relative to spring (17% for each), after the organ- pipe flowering peak was over. I compared results obtained in spring of 1992 and 1993 with those obtained from a previous two year study, and found that in all four years the proportion offruits produced by hurnmingbird pollination remained constant (X=17%), while the proportion of bat-pollinated fruits varied significantly, ranging from 3-17%.

Paternity exclusion analysis of vertebrate-pollinated seeds indicated that most paternal gametes probably carne from within 75 m, although substantial amounts of gene flow (5% and 11%) carne from beyond 125 and 150 m respectively, for the two ~ thurberi populations I studied. 1 hypothesize that the majority oflong-distance pollen 78 79 dispersal events are caused by long foraging flights undertaken by the nectarivorous bat

Leptonycteris curasaoe.

1 conclude that at the study site, both bats and hummingbirds are reliable and effective pollinators of Stenocereus thurberi, even though rnaxirnurn fruit set was not obtained in either year of rny study. Slight shifts in the flowering phenology of organ pipe cactus and migratory schedules ofbats rnay account for the arnong-year differences in bat pollination that were observed. Within-year differences in pollination effectiveness by both bats and hummingbirds are probably attributable to changes in abundance of these migratory flower visitors. 1 hypothesize that pollen dispersal capabilities ofthe nectar- feeding bat L. curasaoae rnay result in low inter-population genetic differentiation in ~ thurberi. 80

INTRODUCTION

In temperate are as, animal pollination of plants occurs primarily by insects. In the

New World tropics, however, nectar-feeding bats (Phyllostomidae; Glossophaginae) and hurnmingbirds (Trochilidae) are important pollen vectors for several hundred species of plants occurring within a wide variety of habitats, inc1uding tropical wet and dry forests, montane ecosystems, and deserts (Helverson 1993, Stiles 1981, Vogel 1968, Porsche

1939). Both bats and hummingbirds are relatively large, homeotherrnic pollinators that require large amounts of energy per individual relative to small-bodied pollinators such as insects (Stiles 1978, Baker 1975). Consequently, on a per flower basis they are expensive for plants to feed. Despite this potential cost, proposed benefits ofvertebrate versus insect pollination may inc1ude 1) longer distance pollen dispersal, which may be especially advantageous for sparsely distributed plants or plants occurring within isolated fragments

2) reliable pollination in inc1ement weather, such as may occur at high elevations and

3) the deposition and uptake of large amounts of pollen per flower visit (Bertin 1982,

Cruden 1976, Heithaus et al. 1974). It is likely that differences in body size, morphology, and foraging behavior of vertebrates relative to insects manifest themselves to a plant as differences in the amount and quality of pollen uptake, deposition, and dispersal (Lertzman

& Gass 1983). These differences may influence pollination effectiveness and fitness components of plants such as seed production and viability (Ramsey 1988, Bertin 1982) as well as pollen dispersal and population genetic structure of plants (Schmitt, 1980).

For hummingbirds and bats to be effective pollinators, they should maximize high quality seed output of a plant and/or a component of male fitness. These criteria are affected by both the quantity and genetic quality of pollen transferred (Waser & Price

1989, Stanton et al. 1986, Willson 1979). Because they may transfer greater quantities of pollen as well as disperse pollen over a larger area than do insect pollinators, it is possible that bat and humrningbird pollination has different micro-evolutionary consequences for 81 plants than does insect pollination. Large amounts of pollen deposited on a stigma, for instance, can result in pollen competition and have important implications for male fitness components and female choice (Mulcahy 1983, Stephenson & Bertin 1983), as well as the type of sexual system exhibited by a plant. Consumption of pollen by bats and the spread of large amounts of pollen over a bat or bird' s body may result in large amounts of pollen being wasted, which may also influence plant sexual systems (Bawa & Beach 1981,

Willson 1979). For example, some authors have noted a correlation between the occurrence of andramonoecious breeding systems in plants and pollination by birds or bats

(Hopkins 1984, Rarnirez et al. 1984). In addition, long-distance pollen dispersal is an important force genetically homogenizing populations within a plant species (Hamrick

1987). Furthermore, because bats and hummingbirds are long-lived and can track floral resources fram year to year, they may be more constant than insects in their selection pressures on floral traits over time (Herrera 1988, Schemske & Horvitz 1984, Stiles

1978). Such constancy in selection pressure over time can pramote the evolution of specialized mutualisms (Horvitz & Schemske 1990).

The purpose of this study was to investigate bat and hummingbird pollination of the organ-pipe cactus, Stenocereus thurberi (Cactaceae: Pachycereeae), a plant which is visited and pollinated by both humrningbirds and bats. Specifically, I addressed three questions regarding vertebrate pollination. 1) How successful are bats and hummingbirds at pollinating flowers of organ pipe cactus, and does their effectiveness vary over time?

2) How much pollen do these visitors deposit on a single-visit basis? and 3) How does bat and humrningbird pollination influence patterns ofpollen-mediated gene flow? To answer these questions, I determined the relative effectiveness of bat and humrningbird pollination within and between flowering seasons for ~ thurberi , quantified patterns of single-visit pollen deposition on stigmas, and determined the extent of pollen-mediated gene flow using plant isozymes and paternity exclusion analysis on vertebrate-pollinated 82 progeny arrays. 1 discuss the results with reference to previous research conducted at the study site regarding pollinator effectiveness and bat foraging behavior (Fleming et al. in review, Fleming et al. in prep.).

STUDY SITE AND ORGANISMS

The study site is located within the central gulf coast region of the Sonoran desert, at Bahia Kino, Sonora, Mexico (290 N, 1100 W). Four species of columnar cacti occur at the site: Stenocereus thurberi (organ pipe), Carnegia gigantea (saguaro), Pachycereus pringlei (cardon), and Lophocereus schotti (senita). ~ thurberi produces flowers from

April through September, with a flowering peak in late May or early to mid-June.

Flowers, which are bisexual and self-incompatible, open shortly after sunset, and during the night are visited by moths and the nectar-feeding bat, Leptonycteris curasoae

(Phyllostomidae; Glossophaginae; Fleming et al., in review). Flowers of ~ thurberi remain open after sunrise, but generally close by about 9 a.m. and do not re-open. During this time, they are visited by honeybees, hurnmingbirds, and rarely, woodpeckers (pers. obs., Fleming et al., in review). The most common hummingbird visitor to organ pipe flowers in late May through July is the broad-billed hummingbird, Cynanthis lasirostris

(Trochilidae, Trochilinae), although Costa's hummingbird, Calypteae costae (Trochilidae,

Trochilinae), also visits organ pipe flowers.

METHODS

Organ Pipe flower morphology,phenology, and pollen production

1 collected one flower from each of 37 cacti and recorded outer corolla tube length, outer corolla tube width at both widest and narrowest points, petallength, stigma exsertion, and style length for each flower. 1 also recorded the number of stigma lobes, corolla tube color and petal color. Petal color was recorded in two categories: red-white 83

versus white, and corolIa-tube color was recorded in three categories: red, red-green, and

green. The number of ovules in flowers was determined by counting all ovules

occurring in 21 flowers.

Phenology was recorded weekly for 40 cacti > 1 m in height; 19 of these plants

occurred within mapped "maternal core" areas (see below), while the remainder were

randomly selected. In 1992, phenology data were colIected from the last week in May

through the third week of July, while in 1993, data were colIected from the final week in

May through the last week of June. For each plant, the number of open flowers,

developing fruits, and mature fruits was recorded. In 1992, estimates ofbat numbers at a

cave located 7 km west of the study site were obtained by either waiting outside the

dayroost and counting bats as they exited in the evening, or entering the roost during the

day and counting the number ofbats present. Unfortunately because oflogistic

difficulties, 1 was not able to obtain estimates ofbat numbers in 1993.

1 determined the amount of pollen produced by organ pipe flowers by colIecting and counting pollen from one flower from each oftwenty-two plants. Flowers were

colIected just after they had opened and dehisced in the early evening (collection prior to this time inhibited fulI opening ofthe flower). The nectary ofthe flower was removed by

cutting off the lower 1/2 of the corolIa tube. The top half of the flower was placed upside down on a piece of aluminum foil and left over-night; in the morning polIen remaining on anthers was removed by holding the flower upside down and vigorously tapping it. PolIen was stored in a refrigerator until transported to the laboratory. In the laboratory, pollen grains were acetolyzed (see Buchmann & Shipman 1990 for details) and counted using a

HIAC-ROYCO model PS-320 particle size analyzer. 84

Pollinator exclusion experiments

Pollinator exclusion experiments were conducted once in 1992 and twice in 1993.

In 1992, experiments were conducted in late May through early June. In 1993, experiments were conducted in late May and were replicated in late June, after the organ pipe flowering peak. Flowers from 19 plants in 1992 and 27 plants in 1993 were

subjected to three pollination treatments in a randomized block design so that each plant underwent all three treatments. Sample sizes in 1992 were substantially larger than in

1993, because an unsuccessful attempt was made in 1992 to collect sufficient diurnal-only

and nocturnal-only pollinated fruits for later use in comparative genetic analyses.

Three treatments were applied to flowers: 1) open pollination, which allowed

access to all flower visitors; 2) diurnal exclusion, which permitted bats and nocturnal insects access to flowers, but excluded all diurnal visitors; and 3) nocturnal exclusion, which permitted hummingbirds and diurnal insects access to flowers but excluded nocturnal visitors. Excluders designed to separate the effects of vertebrate and insect pollination were not used because previous research showed that insects, primarily honeybees, are negligible pollinators (Fleming et al., in review). Bridal veil netting placed

over flowers was used to exclude flower visitors. For treatment (1), open control, no

excluders were used, while for treatment (2) diurnal exclusion, netting was placed over a flower in the morning just prior to sunrise. For the nocturnal exclusion (3), netting was placed over a flower before it opened in the evening and removed prior to sunrise.

AlI cacti used in the pollinator exclusion experiments were tagged and numbered.

Each experimental flower was also individually tagged and numbered using soft aluminum tags inserted in the cactus 2 cm below the flower. Each flower was censused at least once a week until fruits developed. 1 recorded whether flowers and fruits were on or offthe plant and whether fruit development had initiated (indicated by swelling of ovaries 5-7 days after treatment); this was used as an indicator of successful pollination. 85

Pollen deposition by bats and hummingbirds

One adult male L. curasoae and one adult male C. lasirostris were captured for use in pollen deposition experiments. Both bat and hummingbird deposition experiments occurred in a 12'x 12' screen tent placed in a well-ventilated garage, where both individuals were allowed to acclimate to captivity for two days prior to the experiment.

During this time, they had access to freshly picked flowers, ripe organ pipe fruits, and fruit juice in an artificial feeder. During the day, the bat was captured with a hand net, placed inside a smaller cage, and stored in a cool closet in order to prevent it from becoming overheated or dehydrated.

Each evening prior to opening, flowers in the field were covered with bridal veil netting to eliminate visitation. As soon as flowers began opening, small cones made of wax paper were manually inserted into two-thirds of the netted flowers so that the paper lay smoothly against the inner corolla tube, but prevented the stigma from coming into contact with the anthers. One-half ofthese treated flowers were used as recipient flowers in the experiment, while the other half served as unvisited controls. Pollen donor flowers were treated only with bridal veil netting. After insertion ofthe wax-paper, flowers were covered again with netting, and all flowers were left on the plant until fully open and were then picked and transported back to the tent. Upon arrival, anthers from the flowers containing wax paper cones were trimmed off with scissors in an effort to reduce self- pollen contamination.

Pollen deposition experiments with the bat were conducted in the late evening, while the hummingbird pollen-deposition experiments were conducted in the early morning. At the outset ofthe experiment, an untreated (i.e., pollen donor) flower was placed in a wire cylinder hanging from the tent ceiling in a such a way that the position of the flower closely mimicked its position on the cactus. The bat (or hummingbird) was 86 allowed one visit to the flower, whereupon the flower was removed and replaced with a recipient flower. The bat or hummingbird was then allowed one visit to the recipient flower. Because hummingbirds often hover at flowers for a longer period of time than bats, a hummingbird was allowed to hover at a flower and probe the corolla-tube a maximum ofthree times. After being visited, the recipient flower was removed. The stigmas of recipient and unvisited control flowers were clipped off, stored in 70% ethanol, and later transported to the laboratory. After twenty-two replicates over four days, experiments were completed and both bat and hummingbird were released unharmed. In the laboratory, pollen grains on the recipient and control stigmas were removed by sonicating stigmas, and were counted with a particle size analyzer as described above.

Pollen-mediated gene flow

At the main study site, plants in two "maternal core" areas each occurring within a larger "paternal core" area were labeled and mapped (Fig. 1). The study areas were separated by approximately 500 m. Maternal core plants were chosen on the basis ofhigh flowering activity and fruit production and were also used in the pollinator exclusion experiments described previously. In addition to these maternal core plants, paternal core areas 1 and 2 were created by labeling and mapping all flowering and fruiting individual s within 150 m ofthe outermost cacti in maternal core area 1 and within 125 meters from the outermost cacti in maternal core are a 2. The density of ~ thurberi in each study area was determined by counting all cacti >lm tall in ten 30 x 30 m plots separated by 50 m within each site.

Small pieces of bud tissue were collected from all mapped plants in each site and stored in a cooler on ice for 2-4 days before shipment to the laboratory. Bud tissue was then crushed with liquid nitrogen prior to addition of a PVP extraction buffer (Milton et al. 1979). In addition to bud tissue, 1-3 open-pollinated fruits from the maternal core 86 allowed one visit to the flower, whereupon the flower was removed and replaced with a recipient flower. The bat or hummingbird was then allowed one visit to the recipient flower. Because hummingbirds often hover at flowers for a longer period of time than bats, a hummingbird was allowed to hover at a flower and probe the corolla-tube a maximum of three times. After being visited, the recipient flower was removed. The stigmas of recipient and unvisited control flowers were clipped off, sto red in 70% ethanol, and later transported to the laboratory. After twenty-two replicates over four days, experiments were completed and both bat and hummingbird were released unharrned. In the laboratory, pollen grains on the recipient and control stigmas were removed by sonicating stigmas, and were counted with a particle size analyzer as described above.

Pollen-mediated gene flow

At the main study site, plants in two "maternal core" areas each occurring within a larger "paternal core" are a were labeled and mapped (Fig. 1). The study areas were separated by approximately 500 m. Maternal core plants were chosen on the basis of high flowering activity and fruit production and were also used in the pollinator exclusion experiments described previously. In addition to these maternal core plants, paternal core areas 1 and 2 were created by labeling and mapping all flowering and fruiting individual s within 150 m ofthe outermost cacti in maternal core area 1 and within 125 meters from the outermost cacti in maternal core are a 2. The density of ~ thurberi in each study area was determined by counting all cacti >lm tall in ten 30 x 30 m plots separated by 50 m within each site,

Small pieces of bud tissue were collected from all mapped plants in each site and stored in a cooler on ice for 2-4 days before shipment to the laboratory. Bud tissue was then crushed with liquid nitrogen prior to addition of a PVP extraction buffer (Milton et al. 1979). In addition to bud tissue, 1-3 open-pollinated fruits from the maternal core

~ l

87 plants at each site were collected and their seeds sto red in labeled paper envelopes. In the laboratory seeds were germinated on moist filter paper in petri dishes inside growth chambers. Germination usually oeeurred within four days. For most fruits, twenty-four seedlings per fruit at the cotyledon stage (approximately one week old) were prepared for eleetrophoresis by erushing with a PVP extraetion buffer (see Mitton et al. 1979). A total of 456 seedlings from twenty-three progeny arrays from twelve maternal cacti in site one, and 586 seedlings from 30 progeny arrays eolleeted from nine maternal cacti in site two were prepared for isozyme analysis.

I performed stareh gel eleetrophoresis on prepared tissues using buffer systems and staining reeipes of Soltis et al.(1983) using three enzyme systems. Ten polymorphic loei were assayed from bud and seedling tissues. The enzyme systems and loci used were: fluorescent esterase (Fe-l, Fe-2), triose phosphate isomerase (Tpi-l), diaphorase (Dia), phosphoglueoisomerase (Pgi-2), phosphoglucomutase (Pgm-2) with buffer system 34; aspartate aminotransferase (Aat-l, Aat-2) on buffer system 8-, and 6-phosphoglueonate dehydrogenase (6PGD-l, 6PGD-2) on buffer system 11.

Paternity exclusion analysis was used on genotypes of assayed progeny to exclude subsets of potential paternal plants oeeurring in 25 m increments away from eaeh maternal plant. This procedure, deseribed fully in Ellstrand (1984) and Ellstrand & Marshall

(1985), involves eomparing the maternal genotype with genotypes of progeny in order to determine the paternal eontribution. The paternal gamete that must have been contributed is then compared with the set ofpossible fathers. The Paseal program used (written by 1.

Nason) determined the number of progeny from a fruit that had paternal gametes that eould not have been produeed by eaeti occurring within a particular distanee interval.

Analyses were repeated by adding all potential fathers within the next distanee interval until a distanee of 150 m for Site one and 125 m for Site two was reached. In this way the number of gametes that could not have been produced by any of the paternal core area 88 plants could be identified. Using this method, a minimum pollen dispersal distance curve was constructed for both study populations. It is important to note that these curves likely under-estimate the actuallevels ofpollen-mediated gene flow occurring because the exclusion procedure 1 used did not allow me to discriminate between immigrant and local polIen when genotypes of immigrant pollen grains were identical to those occurring within the mapped populations. 1 was not able to use Monte-CarIo methods as recommended by

Devlin & Ellstrand (1990) to estimate levels of such "cryptic" gene flow because it is likely that some seeds within progeny arrays of individual fruits were not fathered independently. This is likely given the large number of pollen grains flower visitors deposit per visit.

Statistical analysis

Data from pollinator exclusion experiments were analyzed using SYSTAT version

5.3 (Wilkinson 1991). Differences among treatments in the frequency offruit production were analyzed using the likelihood ratio X2 statistic (G-statistic, Sokal & Rohlf 1981).

Log-linear analysis of contingency tables was used to determine if a significant interaction existed between year and treatment for 1992 versus 1993 comparisons. Contingency table analysis was used to determine if significant differences across four years (1989, 1990,

1992, 1993; data for 1989 and 1990 are from Fleming et al. in review) existed for the three pollination treatments.

To determine if significant differences existed among treatments for the pollen deposition experiment, variables were log-transformed to eliminate heteroscedasticity, and one-way analysis ofvariance was performed .. Post-hoc comparisons among means were made using Tukey's HSD test. For analysis ofpollen-mediated gene flow, linear regression was performed to determine the effect of distance on the frequency of pollen gene flow events. I ------

89

RESULTS

Organ pipe flower morphology and pollen production.

~ thurberi flowers are stout and funnelform in shape, and white to white-red in

color, conforming to a bat/hummingbird pollination syndrome (Table 1). The average

number ofpollen grains produced per flower is 233,388 (n=22, SD=79,862). Flower petal

color ranged from white to white-red, with the majority of flowers having white-red petal

color (81% white-red vs. 16% for white only). Colors of corolla tubes ranged from green

to red, with the majority offlowers falling in the red/green category (74%). A substantial

number of plants contained flowers with red-only corolla tubes (19%), while a rninority of

flowers fell in the green-only corolla-tube category (7%). The average amount of ovules

present per flower was 734 (n=9, SD=199).

Phenology and bat abundance

Average flower production per plant was higher in June 1992 than in June 1993,

but in both years, flower production decreased in July (Fig. 2a). Despite higher flower

production in 1992, however, mature fruit-set was greater in 1993 than it was in 1992

(Fig. 2b). This finding is in accord with the results of pollinator exclusion experiments I

conducted in 1992 and 1993 (see below).

Bat census data for L. curasoae indicates that near the study site, bat abundance is

highly variable between months (Fig. 3), a pattern which has been found in previous years

(Fleming et al, in review). In 1992, bat numbers were very low in late spring, greatest

near the beginning of July, and very low again near the end of July. Weeks in which L.

curasao e bats were abundant in 1992 are different than in 1990, a finding suggestive of a

variable migratory schedule.

~ 90

Pollinator exclusion experirnents

Exclusion experirnents conducted in spring 1992 resulted in significant differences in fruit production arnong treatrnents (n=468, G=14.25, p=.OOl) with hurnrningbirds producing a significantly greater proportion (12%) offruits than bats (3%; n=403, df=l,

G=10.184; p=.OOl; Table 2a). Differences arnong treatrnents for the experirnent conducted in spring 1993 were also significant (n=116, G=6.3, p<.05). However, the proportion of fruits attributable to either bat (17%) or hurnmingbird (17%) pollination did not differ significantly as it did in 1992 (n=83, df=l, G=.002, p=.961; Table 2a). When the experirnent was replicated in late June 1993, after the organ-pipe flowering peak, fruit set in all three treatrnents was significantly lower than in May 1993 (Table 2b). Variation in pollinator effectiveness between years was apparent, as log-linear analysis of a three- way table for both years indicated a significant interaction between year and treatrnent

(n=474, G=329.64, p<.OOl). Fruit production by open and bat, but not hurnrningbird, pollinated flowers was significantly greater in 1993 than in 1992 (Table 2a). Cornparison of pollinator effectiveness across 1989, 1990, 1992 and 1993 indicate that bat and open pollination varied significantly arnong years (n=377, G=10.567, df=3, p=.014 for bats; n=169, G=8.926, df=3, p=.030 for open pollination) but that hurnrningbird pollination did not (n=245, G=2.221, df=3, p=.528).

Pollen deposition by bats and hurnmingbirds

Bats and hurnrningbirds both deposit large arnounts of pollen on stigrnas per flower visit, (Fig. 4). The hurnmingbird and bat deposited a significantly greater arnount of pollen on stigrnas per visit than did the unvisited control (one-way ANO VA, n=69, df=2,

F=49.809, p<.OOl). Tukey's HSD test revealed that the arnount of pollen grains that the bat and hurnmingbird deposited were not significantly different. Because organ pipe 91

flowers contain an average of734 ovules (SD=199), pollen-ovule ratios on stigmas after a

single flower visit were, on average, approximately 9.1 for the bat and 14.9 for the

hummingbird.

Pollen-mediated gene flow

Both study populations had similar allele frequencies, although some alleles were

present at site 1 but not at site 2 (e.g., alleles ofPgm, Fe-1, and 6P-1; Table 3). Organ

pipe cactus density for Site 1 was 19 plants/ha, and density at Site 2 was 28 plants/ha.

Despite these differences in plant density, pollen dispersal curves obtained from paternity

exclusion analysis for both study populations exhibited similar leptokurtic distributions

(Fig 5). Examination of data from both sites indicated that most paternal gametes could

have originated from plants occurring within 75 m ofmaternal plants. In fact, a substantial

proportion (50% at Site 1; 65% at Site 2) of seeds produced could have been pollinated

by paternal plants occurring within 25 m, i.e., primarily by nearest neighbor plants.

However, a significant proportion of seeds were fathered by plants from outside the study

populations (5% >150m at site 1, and 18% >125m, at site 2). Linear regression analysis

of the relationship between pollen dispersal distance and frequency of pollen dispersal

indicated a significant inverse relationship at both sites (Table 4).

Analysis of the frequency of pollen dispersal events on a fruit by fruit basis indicates that most fruits (87% at site 1; 100% at site 2) contained at least one or more

seeds that could have been fertilized by paternal plants occurring within 25 meters (Fig. 6).

However, a substantial percentage offruits (50% at site 1, 97% at site 2) also contained at least one or more seeds fertilized by fathers occurring further than 125 meters. Thus, a single fruit may contain seeds that were fertilized by plants occurring within as well as outside the study populations. ------i

--- J

92

DISCUSSION

Pollinator effectiveness

At Bahia Kino, both bats and humrningbirds visit organ pipe cacti and contribute to

fruit production. Examination of data collected over a four-year period indicates that in

three offour years hummingbirds contributed a greater proportion offruit-set than bats,

(see Flerning et al. in review for data collected in 1989 and 1990). In addition, the

proportion of fruits resulting from hummingbird-pollination remained constant across all

four years. Bat pollination resulted in a greater proportion of fruits produced in 1993 than

in 1992. Thus, despite some variation among years in pollinator effectiveness, both bats

and humrningbirds were present and contributed to fruit production in every season, and

the relative importance ofboth was similar. Small differences in flowering peaks among

years in organ pipe cacti, combined with slight shifts in the migratory schedules ofbats

could account for among-year variation in bat pollination success.

The pattern observed for bat and humrningbird pollination of.s.,. thurberi contrasts

with patterns observed for many speeies of insect pollinated plants. Horvitz &

Sehemske(1990) found considerable variation over a four year period in the taxonomie

composition of insect pollinators of a neotropical herb. Herrera (1988) investigated an

inseet pollinator assemblage over a six-year period and found marked variation in the

diversity, eomposition, and abundance of pollinators between years. His review of

available literature on variability of inseet-pollinated assemblages led him to conclude that

variation in such assemblages is "the rule, rather that the exeeption". Howe (1984)

suggested that sueh variation in mutualisms will limit the degree to whieh specialization

oecurs. Beeause both nectar-feeding bats and hummingbirds oecur throughout the entire

range of organ pipe eaeti (Arita & Humphrey 1992, True 1993), it is likely that these

vertebrates are pollinators of this cactus thoughout its range, although this has yet to be

~ 93

confirmed. Floral morphology and nocturnal and early morning nectar production of

organ pipe is suggestive of adaptation to attract both types of visitors. At the study site at

least, there appears to be a constant selection regime favoring the maintenance of a bat-

humrningbird pollination syndrome.

It is probable that the relative importance ofbats vs. hummingbirds may vary over

the range of ~ thurberi. For example, in the summer, L. curasao e bats are more numerous

near Organ Pipe Cactus National Monument than they are at Bahía Kino (Cockrum &

Petryszyn 1991); therefore, bats likely contribute a greater proportion of pollination

services there than they do at the study site. One might hypothesize that at this northern

site, not only would bats contribute a greater proportion of fruit-set than hummingbirds,

but that patterns of pollen-mediated gene flow might also reflect a greater frequency of bat

pollination.

Examination of withín-year data indicates that pollination by both bats and

hummingbirds dropped in late June 1993, after the peak flowering season of organ pipe

had passed. ~ thurberi phenology differs from the two other vertebrate-visited species of

columnar cacti occurring at the site (Carnegia gigantea and Pachycereus pringlei) in that

its flowering season extends over most of the spring and summer, while flowering in the

other cacti is restricted to about two months in the spring (Fleming et al., in review).

During most of its flowering season, therefore, ~ thurberi does not compete with C.

gigantea and P. pringlei for flower visits. Census data for L. curasaoe indicate that bat

abundance is highly variable over the organ pipe flowering season (Fleming et al. in rev.; this study), with bat abundance being highest in the spring and dropping sharply over the

surnmer. Bahia Kino occurs along a proposed migratory corridor that has been postulated for L. curasoae and it is probable that bats from this are a continue north into southern

Arizona (Fleming et al. 1993, Wilkinson & Flerning in prep.). C. lasirostris abundance also varies over the spring and summer at Bahia Kino; observations of humrningbird

~ , 1_---

94

flower visits in April and early May 1989 and 1990 (Fleming et al., in review) indicated

that C. lasirostris hurnmingbirds were not the most common hummingbird visitors to

cactus flowers as they were in late May and June 1992 and 1993, during my study. If bat

and hummingbird abundance varies at Bahia Kino during the spring and summer, then it is

not unreasonable to expect pollination success in S. thurberi to vary over a flowering

season, as my data indicate.

Pollen deposition

Although several studies have examined patterns of hummingbird pollen deposition

in several species of temperate and tropical plants (Feinsinger & Tiebout 1991, Feinsinger

et al. 1986, Bertin 1982, Price & Waser 1982), similar information for bats is lacking. The

number of pollen grains deposited on a stigma influences fruit production and seed-set

(Waser & Price 1991, Bertin 1990), and avian flower visitors have been found to deposit

more pollen on stigmas than insects (Ramsey 1988, Bertin 1982). Data from this study

indicate that for ~ thurberi flowers, both bats and hummingbirds deposit in a single visit

9-15 times more pollen grains than the number of ovules present in flowers. Therefore,

both flower visitors are probably highly effective pollinators on a single-visit basis,

although I did not conduct field experiments to test this. It is possible that more than one

visit per flower may be necessary to ensure maximum fruit or seed set per fruit; benefits of

higher visitation rates to flowers may include a greater probability of receiving outcrossed

pollen and/or pollen high genetic quality, which has positive effects on seed production

and viability in some plant species (Waser & Price 1991, Bertin 1990, Marshall &

Ellstrand 1986, Ellstrand 1984, Schemske & Pautler 1984, Janzen 1980).

Although differences between average amounts of bat and hummingbird pollen

deposited per visit were not statistically signficant in my experiment, differences in their

flower visitation behavior were apparent. First, bat flower visits are brief, lasting only a

c~ 95 few seconds or less, while those of hurnmingbirds may last up to a few minutes (pers. obs). Second, bats probe the flower only once, placing their head and neck into the corolla and coming into contact with anthers and stigma; hummingbirds often probe flowers several times. Third, while they also extend their head and neck into the corolla, because of their smaller size, hummingbirds do not come into contact with as many anthers as do bats. The small size of hurnmingbirds relative to ~ thurberi flowers probably explains why their patterns of pollen deposition may be more variable relative to bats. It is possible that allowing the hurnmingbird to probe flowers up to three times may have influenced the outcome of this experiment. Also, although my observations suggest visitation behavior to naturally occurring and treatment flowers did not differ, 1 cannot cornpletely elirninate the possibility that treatment of recipient flowers may have somehow influenced flower visiting behavior during the experiment.

In addition to influencing maternal fitness through seed and fruit production, a large amount of pollen deposited per visit may have important effects for levels of pollen- rnediated gene flow. Examination of allele frequencies within the progeny arrays suggest that seeds within a fruit were not all independentIy fathered. Thus, because one flower visit by a bat or hummingbird may result in more than one seed within a fruit being sired by one plant, one long-distance pollination event can result in the production of several

"long-distance" progeny. This effect rnay be magnified ifthere are significant amounts of pollen carryover, as has been shown for hurnmingbirds (Price & Waser 1982). 1_-- 1

96

Pollen-mediated gene flow

The results of this study indicate that flowers of organ pipe cacti receive a mixture

ofnear-neighbor and long-distance pollen. Paternity exclusion analysis indicates that most

pollen comes from within 75 m ofthe maternal plant, although substantial amounts of

pollen did come from outside the mapped study areas. Pollen dispersal from outside the

mapped area may be higher than reported here because, as mentioned previously, the

paternity exclusion method used in this study gives a minimum pollen dispersal-distance

curve and may have underestimated actuallevels ofpollen-mediated gene flow. Previous

studies (Godt & Harnrick 1993, Devlin & Ellstrand 1990) have demonstrated that the

apparent rate of gene flow is one-half to one-third that of the total rateo

It is not surprising that most pollen dispersal may be occurring primarily within a

radius of75 m; many other published studies have demonstrated that pollen dispersal

often follows a leptokurtic distribution (Levin 1984). Moreover, studies that have focused

on humrningbird pollen dispersal patterns have found substantial nearest neighbor pollen

dispersal (Parra et al 1993, Linhart et al. 1987, Murawski & Gilbert 1986, Webb & Bawa

1983, Linhart 1973); this is usually attributed to the widespread occurrence of territorial

behavior in hummingbirds. Pollen dispersal in these studies, however, has probably been

somewhat underestimated because pollinator flight distances and/or dyes were used to

estímate gene flow. These methods may not lead to an accurate picture of dispersal

patterns and give no indication of whether or not a viable fruit has been produced (Broyles

& Wyatt 1991, Fenster 1991, Hamrick 1989). Recent work has demonstrated that pollen-

mediated gene flow often occurs over larger areas than previously thought (Godt &

Hamrick 1993, Kohn & Casper 1992, Broyles & Wyatt 1991, Hamrick & Murawski 1990,

Ellstrand & Marshall 1985). In addition, studies have demonstrated that a considerable

degree of pollen carryover may occur with hummingbirds, possibly leading to long

distance dispersal events (Price & Waser 1982).

J 97

The hummingbird responsible for the majority of successful pollination events of S. thurberi, Cynanthis lasirostris, exhibits territorial behavior while foraging, often chasing intruders away from flowers (pers. obs.; Parra et al 1993). Observations of C. lasirostris flower visits revealed that although some visits were to nearest neighbor cacti, many times hummingbirds flew out ofview after visiting a cactus (pers. obs.). Although in this study 1 did not estimate territory sizes of C. lasirostris, territory size may be large enough so that non-nearest neighbor pollination by hummingbirds is a common event.

It is probable that a large proportion of long-distance pollen dispersal events are due to flower visits by the nectar-feeding bat, Leptonycteris curasaoe. Previous radio- tracking studies have shown that in addition to being migratory, this bat is highly mobile on a nightly basis. For example, bats at Bahía Kino often commute 25-30 km one-way between their day roosts and foraging areas (Sahley et al. 1993). Whíle they feed on cactus flowers and fruit, they may typically forage over a 1 km2 patch before flying up to

10 km to reach other foraging patches; total distance flown in one night of foraging may be up to 100 km (Fleming 1992, Fleming et al. in prep.). We do not yet have detailed information on their small-scale plant visiting patterns (e.g. how many nearest neighbor plant visits they make), but given the large areas over which they forage, a large amount of pollen-mediated gene flow is likely to occur, perhaps over distances of several kilometers.

Given such potential for inter-population gene flow, 1 hypothesize that genetic differentiation of organ pipe cactus populations occurring within the L. curasao e migratory corridor is likely to be low.

CONCLUSION

This study has revealed that bats and hurnmingbirds deposit large pollen loads on stigmas of ~ thurberi, and can disperse pollen over long-distances, even though considerable pollen dispersal may originate within 75 meters of maternal plants.

~ 1__ - 1

98

Furthermore, results from this and a previous study (Fleming et al. in review) indicate that

the proportion offruit-set due to hurnmingbirds was constant over a four year period,

while fruit-set due to bat pollination varied. Despite such variation in pollinator

effectiveness over time, both bats and hummingbirds were present during each flowering

season and, qualitatively, the relative importance of pollinators was similar across three of

four years, with hummingbird pollination contributing a greater proportion of fruit set in

1989, 1990, and 1992. Presence of L. curasaoe at the study site across years is likely

because ofre-use of the same nectar corridor from year to year. Differences in bat

pollination effectiveness observed in 1993 relative to other years may be due to slight

shifts in both the migratory schedules of pollinators and phenology of plants. Because

both bats and hummingbirds are migratory, a prolonged flowering season can result in

different levels of pollination success within a year, as occurred in 1993 for ~ thurberi.

More information is needed on additional bat-and/or hummingbird-pollinated

plants to determine whether or not the patterns for fruit production and pollen dispersal

observed here are consistent across a range of plant species and habitats in the neotropics.

Comparative studies focusing on how differences in size, morphology, and plant-visiting

behavior of bats versus hummingbirds influence fruit production and pollen dispersal

would also be useful. Knowledge regarding the effect these flower visitors have on

pollination success and gene flow for cacti as well as other tropical tree species is

important not only from a theoretical perspective, but is also important for management

considerations in fragmented or logged habitats.

ACKNOWLEDGMENTS

T. Fleming, 1. Harnrick, D. Janos, S. Petit, and K. Waddington reviewed this

manuscript and provided valuable discussion and assistance throughout all phases of this

project; 1. Hamrick kindly let the author use his laboratory facilities at the University of 1:---

99

Georgia. 1. Nason provided much advice and wrote the Pascal program used in the

paternity exclusion analysis. S. Buchman ofthe USDA Carl Hayden Research Center

generously assisted with polIen and ovule counting. V. Sosa provided assistance in the

field. T. & T. Sahley provided logistic and financial support. Portions of this work were

conducted with support from World-Wildlife Fund/Garden Club of America Scholarship in

Tropical Botany, and a graduate teaching assistantship from the University of Miami.

.j 100

LITERA TURE CITED

Baker, HG. 1975. Sugar concentrations in nectars from hummingbird flowers. Biotropica 7:37-41.

Bawa, K.S., & Beach lH 1981. Evolution ofsexual systems in flowering plants. Annals ofthe Missourri Botanical Garden 68:254-274.

Bertin, RI. 1990. Effects ofpollination intensity in Campsis radicans. American Journal ofBotany 77(2): 178-187.

Bertin, RI. 1982. Floral biology, hummingbird pollination and fruit production oftrumpet creeper (Campsis radicans, Bignoniaceae) Am. 1. Botany 69(1):122-134.

Broyles, S. B. & Wyatt, R. 1991. Effective pollen dispersal in a natural population of Asclepias exaItata: The influence of pollinator behavior, genetic similarity, and mating success. Am. Naturalist 138(5)1239-1249.

Buchmann, S.L. & Shipman, e.W. 1990. Pollen harvesting rates for Apis mellifera L. on Gossypium (Malvaceae) flowers. J. Kansas Entomological Society 63: 92-100.

Cockrum E. L., & Petryszyn Y. 1991. The long-no sed bat, Leptonyceteris: an endangered species in the southwest? Occas. Papo Mus. Texas Tech Univ., 142: 1-32.

Cruden, RW. 1976. Pollinators in high-elevation ecosystems: Relative effectiveness of birds and bees. Science: 1439-1440.

ElIstrand, N.e. 1984. MuItiple paternity within the fruits ofthe wild radish, Raphanus sativus. Am. Nat. 123:819-828.

ElIstrand, N.e. & Marshall, D.L. 1985. Inter-population gene flow by pollen in wild radish, Raphanus sativus. Am. Nat. 126(5):606-616.

Devlin, B. & ElIstrand, N.C. 1990. The development and application ofa refined method for estimating gene flow from agiosperm paternity analysis. Evolution 44:248- 259.

Feinsinger, P., & Tiebout, HM. 1991. Competition among plants sharing hummingbird pollinators; laboratory experiments on a mechanism. Ecology 92: 1946-1952. 101

Feinsinger, P., LH. Beach, Y.B. Linhart & W.H. Busby. 1986. Floral neighborhood and pollination success among Costa Rican cloud forest plants. Ecology 68: 1294- 1306.

Fenster, CB. 1991. Gene flow in Chamaecrista fasciculata (Leguminosae). 1. Gene dispersal. Evolution 45:298-409.

Fleming, T.H. 1992. How do fruit- and nectar- feeding birds and mammals track their food resources? In: Effects of Resource Distribution on Animal-Plant Interactions. Pp. 355-389.

Fleming, T.H., Numez, R.A., Sternberg, L. 1992. Seasonal changes in the diets of migrant and non-migrant nectarivorous bats as revealed by carbon stable isoptoe analysis. Oecologia 94:72-75.

Fleming, T.H., Horner, M., Tuttle, M.D. in review. The relative effectiveness of diurnal and nocturnal pollinators to flowers of three species of columnar cacti. Southwest American Naturalist.

Fleming, T.H., Horner, M., Sahley,C. in preparation. The foraging behavior ofthe nectarivorous bat, L. curasoae.

Godt, M.W. & Hamrick, lL. 1993. Patterns and levels ofpollen-mediated gene flow in Lathyrus latifolius. Evolution 47(1):98-110.

Hamrick, lL. 1989. Isozymes and the analysis of genetic structure in plant populations. In: Isozymes in plant biology. (Soltis, D.E. and Soltis, P.S., eds.). Dioscorides Press, Portland, OR. pp.87-105.

Hamrick, J.L. 1987. Gene flow and distribution of genetic variation in plant populations. In: Differentiation patterns in higher plants (K. Urbanskak, ed.) Academic Press, New York. Pp. 53-67.

Hamrick, lL. & Murawski, D.A. 1990. The breeding structure oftropical tree populations. Plant Species Biology 5:157-165.

Heithaus, E.R., Opler, P.A., Baker, H.G., 1974. Bat activity and pollination ofBauhinia pauletia: plant -pollinator coevolution. Ecology 55:412-419.

Helverson, O. V. 1993. Adaptations of flowers to the pollination by glossophagine bats. Pp. 41-59 in: Plant-Animal Interactions in Tropical Enviornments (ed. W. Barthlott et al.). Museum Alexander Koening, Bonn.

, j 102 Herrera, e.M. 1988. Variation in mutualisms: the spatio-temporal mosaic ofa pollinator assemblage. BioI. Journal ofthe Linnean Society 35:95-125.

Hopkins, H.e. 1984. Floral biology and pollination ecology of the neotropical species of Parkia. Journal ofEcology 72: 1-23.

Horvitz, C.e. & Schemske, D.W. 1990. Spatiotemporal variation in insect mutualists ofa neotropical herb. Ecology 71:1085-1087,

Howe, H.F. 1984. Constraints on the evolution ofmutualisms. Am. Nat. 123:764-777.

Janzen, D.H., DeVries, D.W., Gladstone, M.L. Higgins & Lewinsohn, T.M. 1980. Self and cross-pollination ofEncyclia cordigera (Orchidaceae) in Santa Rosa National Park, Costa Rica: Biotropica 12:72-74.

Kohn, & Caspar 1992. Pollen-mediated gene flow in Curcubita foetidissima (Cucurbitaceae) Am. l Bot. 79:57-62.

Levin, D.A. 1984. Immigration in plants; An excercise in the subjunctive. In: Perspectives in plant population ecology. (R. Dirzo & J. Sarukhan, eds.). Sinauer, Sunderland, Mass. Pp. 242-260.

Lertzman, K.P. & Gass, C.L. 1983. Alternative models ofpollen transfer. In: Handbook ofExperimental Pollination Biology. Eds. Jones, e.E. & R.l Little). Van Nostrand Reinhold, New York. pp 474-489.

Linhart, y.B. 1973. Ecological and behavioral determinants of pollen dispersal in hummingbird-pollinated Heliconia. Am. Nat. 107:511-523.

Linhart, Y.B., Busby, W.H., Beach, lH., & Feinsinger, P. 1987. Forager behavior, pollen dispersal, and inbreeding in two species of hurnmingbird-pollinated plants. Evolution 41 (3)679-682.

Marshall, D.L. & Ellstrand, N.C. 1986. Sexual selection in Raphanus sativus: Experimental data on non-random fertilization, maternal choice, and consequences of multiple paternity. Am. Nat. 127:446-461.

Mitton, lB., Linhart, Y.B., Sturgeon, B.K., and lL. Hamrick. 1979. Allozyme polymorphism detected I mature needle tissue of ponderosa pine, Pinus poderosa. Laws. J. Heredity, 70:86-89.

Mulcahy, D.L. 1983. Pollen competition in a natural population. In: Handbook of experimental pollination biology. (E.e. Jones, & RJ. Little, eds.) van Nostrand- Reinhold, New York. Pp. 330-337. 103

Murawski, D.A. & Gilbert, L.E. 1986. Pollen flow in Psiguria warscewiczii: a comparison ofHeliconius butterflies and hummingbirds. Oecologia 68:161-167.

Parra, v., Vargas, c., Eguiarte, L.E. 1993. Reproductive biology, polIen and seed dispersal, and neighborhood size in the hummingbird-pollinated Echeveria gibbiflora (Crassulaceae). Am. 1. Bot. 80(2):153-159.

Porsche, O. 1939. Das Besaubugsleben der Kadkteenblute. In Jahrb. Deutsche Kakt.- Ges. 142 pp.

Price, M.V. & Waser, N.M. 1982. Experimental studies ofpolIen carryover: hummingbirds and Ipomopsis aggregata. Oecologia 54:353-358.

Ramirez, N., Sobrevila c., Enrech, N., & Ruiz-Zapata, T. 1984. Floral biology and breeding system ofBauhinia ungulata: a bat-pollinated tree in the Venezuelan llanos. Am. 1. Botany 71:273-281.

Ramsey, M.W. 1988. Differences in pollinator effectiveness ofbirds and insects visiting Banksi menziensii (Proteaceae). Oecologia 76:119-124.

Sahley, C.']", Homer, M.A., & Fleming, T.H. 1993. Flight speeds and estimated mechanical power outputs for the nectar-feeding bat, Leptonycteris curasoae (Phyllostomidae; Glossophaginae). 1. Marnm. 74(3):594-600.

Schemske, D.W. & Horvitz, C.C. 1984. Variation among floral visitors in pollination ability: A precondition for mutualism specialization. Science 225: 519-521.

Schemske, D.W. & Pautler, L.P. 1984. The effects ofpollen composition on fitness components in a neotropical herb. Oecologia 62:31-36.

Schmitt, 1. 1980. Pollinator foraging behavior and gene dispersal in Senecio (Compositae). Evolution 34:914-943.

Sokal, R.F., & F.1. Rohlf. 1981. Biometry. W.H. Freeman & Company, New York, USA.

Soltis, D.E., Haufler, C.H., Darrow, D.C. and G.J. Gastony. 1983. Starch gel electrophoresis offems: a compilation of griding buffers, gel and electro de buffers and staining schedules. Am. Fer. Joumal, 73:9-15. 104 Stiles, G. 1981. Geographical aspects ofbird-flower coevolution, with particular reference to Central America. Ann. Miss. Bot. Gard. 68:323-351.

Stiles, F.G. 1978. Ecological and evolutionary implications ofbird pollination. Am. Zoo1. 18:715-727.

Stanton, M.L., Snow, AA, & Handel, S.N. 1986. Floral evolution: Attractiveness to pollinators increases male fitness. Science 232:1625-1627.

Stephenson, AG., & Bertin, R.I. 1983. Male competition, female choice, and sexual selection in plants. In: Pollination Biology (L. Real, ed.) Academic Press, Orlando. pp. 110-140.

True, Dan 1993. Hummingbirds ofNorth America: Attracting, feeding, and photographing. University ofNew Mexico Press. Albuquerque.

Waser, N.M., & Price, M.V 1991. Outcrossing distance effects in Delphinium nelsonii: Pollen loads, pollen tubes, and seed set. Ecology 72(1): 171-179.

Waser, N.M. & Price, M.V 1989. Optimal outcrossing in Ipomopsis aggregata: Seed set and offspring fitness. Evolution 43(5):1097-1109.

Webb, c.l. & Bawa, K.S. 1983. Pollen dispersal by hummingbirds and butterflies: A comparative study oftwo lowland tropical plants. Evolution 37:1258-1270.

Wilkinson, L. 1991. Systat. The system for statistics. Systat, Inc., Evanston, Illinois, USA

Wilkinson, G.S. & T.H. Fleming. Migration oflesser long-no sed bats, Leptonycteris curasoae, inferred from mitochondrial DNA Unpublished manuscript.

Willson, M.F. 1979. Sexual selection in plants. Am. Nat. 113:777-790.

Vogel, 1969. Chiropterophilie in der Neotropischen Flora. N eue Mitt. II, III Flora abt. B,158:185-323.

.J Table 1. Stenocereus thurberi flower morphology. Measurements were taken from 37 flowers. Coeffcients ofvariation are in parentheses.

Outer Outer Outer Petal Style Stigma Number of corona corona corona length length exertion stigma tube tube tube (cm) (cm) (rnrn) lobes length (cm) width width (widest) (narrowest) (cm) (cm)

Mean 6.2 3.3 2.3 2.1 5.2 .15 11 (10.2) (22.2) (19.9) (17.9) (15.3) (2.4) (11.0)

Minimum 4.9 1.7 1.7 1.6 2.3 O 9 (O is mode) Maximum 7.9 3.5 3.5 3.3 6.5 l.5 14

•..... o U1

11Io...- 106

Table 2. Results of pollinator exclusion experiments.

A. Proportion fruit set for spring (late May/early June) 1992 vs 1993. G-statistics and probablities are for between year comparisons.

1992 1993 Treatment N Fruit- N Fruit-set G D.F. P value set statistic Open 65 .16 33 .39 7.844 1 .005

Nocturnal 267 .03 42 .17 9.34 1 .002 pollination

Diurnal 136 .12 41 .17 .744 1 .388 E,ollination

B. Comparison of proportion fruit-set for pollinator exclusion experiments in May and June 1993. G-statistics are for between month comparisons.

Fruit-set Fruit-set Treatment N May 1993 N June 1993 G-statistic P value Open 33 .39 32 .06 11.01 .001

Nocturnal 42 .17 37 O 9.4 .002 pollination

Diurnal 41 .17 35 .03 4.588 .03 pollination

I ~ 107 TabIe 3. Isozyme alleIe frequencies for ~ thurberi study popuIations.

AlIele Freguency Locus AlIeIe code number Site 1 Site 2 Pgi 2 .799 .806 3 .201 .194

Pgm 1 .041 O 2 .861 .894 3 .098 .106

Tpi 2 .681 .673 3 .319 .327

Dia 1 .979 .944 2 .021 .056

Fe-1 1 .008 O 2 O O 3 .808 .74 4 .016 .010 5 .033 .010 6 .133 .229

Fe-2 1 .094 .189 3 .261 .349 6 .645 .462

6P-1 1 .007 O 2 .927 .921 3 .065 .078

6P-2 2 .041 .019 3 .944 .961 4 .014 .019

Aat-1 2 .022 .01 3 .977 .99

Aat-2 2 .112 .085 3 .888 .915

d 108 Table 4. ANO VA table for regression analysis ofthe relationship between pollen dispersal distance and the frequency of pollen dispersal.

Source df R2 t P Site 1:

Distance 1,149 .268 9.90 <.001

Distance 1,5 .636 3.95 .01 (population mean)

Site 2

Distance 1,179 .237 11.19 <.001

Distance 1,4 .321 2.07 >.05 (population mean) df, degrees of freedom R2, coefficient of determination t, t-test statistic p, probability of t-test statistic

----~- j 109 FIGURE LEGENDS

Figure 1a & b. Map of ~ thurberi study populations used in analysis of pollen-mediated

gene flow. Maternal core plants are indicated by circIes. Site two is located

approximateiy 500 m southwest of Site one.

Figure 2a & b. ~ thurberi flower and fruit production respectively for 1992 and 1993.

Data were recorded for forty labeled plants; "W" stands for week in which

data were recorded.

Figure 3a &b. L. curasao e census data for Sierra Kino day roost, 1990 and 1992.

(1990 data is from Flerning, et al. in review).

Figure 4. Number ofpollen grains deposited in single visits by a bat and a hurnrningbird.

Figure 5 a & b. Minimum pollen distance dispersal curve for Site 1(A) and Site 2(B).

The number of progeny produced from paternal gametes occurring within each

distance interval is indicated on the graphs.

Figure 6 a & b. Proportion offruits containing at least one seed that could have been

produced within each distance interval indicated for Site l(A) and Site 2 (B). 110

400 Site 1 350 ~~ - 300 ~ ~ ~ ~~ (1) ~ ~ ~ ~ ~ (1) 250 ~ - ~ E ~ - 200 ~ ~~ • ~ ~ ~ ; 150 • 8 ~ ~ • ••• t • • ~ ~ 100 /:& -l ••• ~ ~ ~ ~ ~~ 50 + • \ b.~ lA O I .•. O 50 100 150 200 250 300 350 400 Distance (meters)

275 250 Site 2 ~~ 225 ~ ~ ~ ~~ -e 200 ~ (1) ~ ~~ ~ ~ ~ -(1) 175 E 150 •• •• -(1) ~ •'" CJ 125 • c: ~ ~.• ~ ~~ ns 100 . ~ 1/) ~ ~ - 75 ~ e ~ ~ 50 • ~ ~ ~ 25 ~ ~ o ,~ .... .•. ~ o 25 50 75 100 125 150 175 200 225 250 275 Distance (meters)

FIGURE 1 111

3.5

t: 3 -CU Q. 2.5 .--1992 -e --...-.·1993 (1) 2 ~ O 1.5~------A' ¡¡: e cu 1 (1) ~ 0.5

O May June June June June July July W4 W1 W2 W3 W4 W2 W3 Date

1.6T B lA, 1.4 I , , e I 1992 I • -cu 1.2 I t I Q. '. --...-·1. 993 ti) 1 ::- I ;. ::::J 0.8 u"-. 0.6 cue (1) 0.4+ ,. ~ 0.2 O May June June June June July July W4 W1 W2 W3 W4 W2 W3 Date

FIGURE 2 112

A 3000 l In 1990 .-nsc 2500 el) ns 2000 O In ns 1500 :l"- 1000 (J -Í 500 ~ O April-5 April-30 June-8 June-28 July-20 Date

In 3000 l B 1992 .-nsc 2500 el) ns 2000 o In ns 1500 :l"- 1000 (J -Í 500 ~ o April28 June 10 July 1 July 20 Date

FIGURE 3 r

¡uaw¡eaJl

pJ!q5u!WWnH lea IOJlU08 o s: (1) D) ooo~ :::l ~ "'tI 00017 2.

(1) 0009 :::l te.., OOOS D) :::l oooo~ (fI ¡uaw¡eaJ¡ Jad ZZ=u ? (fI ooo~~ -

Ell 114

(A) 70 T Site 1 60

>- 50 e ~ 40 o ~ 30 o~ 20 10

O <25 >25<50 >50<75 >75<100 >100<125 >125<150 >150

Distance (m)

70 T Site 2 (B) 60

50 >- e cv 40 el o D"..- 30 o~ 20

10

O <25 >25<50 >50<75 >75<100 >100<125 >125 Distance (m)

FIGURE 5 115

(A) Site 1 IV U c: 90 S r--- .!!! 80 ~ .!: 70 - -g e 60 IV O CII el 50 - r--- ••••. IV 40 1\ 1; s: U •• 30 r--- 'i 20 ~ 10 2 ["1 LL O ~ e <25 >25<50 >50<75 >75<100 >100<125 >125 Distance(m)

(B) Site two c: 100 r- 90 r--- ~~ g¡ O 80 .•.•.CII, elf:! 70 1\ C'CI 60 s: U r--- ~ IV 50 ~ ~ 40 CII C'CI 30 r--- r-- r--- •-•••• CII ;:, .- 20 U:~ 10 e~ O <25 >25<50 >50<75 >75<100 >100<125 >125 Distance(m)

FIGURE 6 VITA

Catherine Teresa Sahley was born in Cleveland, Ohio on 26 December 1964.

Her parents are Theodore A. Sahley and Teresa Sahley and she has a sister Caroline

M. Sahley and a brother Theodore A. Sahley. She attended primary school at

Highland Drive Elementary in Brecksville, Ohio and secondary school at Revere

High in Richfield, Ohio. During this time her interests in nature and travel were encouraged by her farnily. Catherine attended the University of Akron from 1982 to

1986 and graduated with a Bachelor of Science in Biology and a minor in Spanish.

From 1986-1988 she assisted on several research projects and worked in Washington

DC, Virginia, Florida, Venezuela, and Brazil. She pursued graduate studies at the

University ofMiarni, Department ofBiology from 1988-1995 and obtained the degree

Doctor ofPhilosophy. During this time she conducted her dissertation research in

Mexico and Peru. She plans to continue conducting ecological and conservation- oriented research in Latin America.