GEB247.fm Page 359 Tuesday, June 5, 2001 3:53 PM Global Ecology & Biogeography (2001) 10, 359–367 RESEARCH LETTER ReproductiveBlackwell Science, Ltd ecology and pollen representation among neotropical trees MARK B. BUSH1 and ROBERT RIVERA21Department of Biological Sciences, Florida Institute of Technology, 150 W. University Blvd, Melbourne, Florida 32901, U.S.A., E-mail: mbush@fit.edu 2Department of Botany, Duke University, Durham, North Carolina 27708, U.S.A. ABSTRACT species or those zoophilous species exhibiting ‘messy’ pollination syndromes. Pollination mech- Three years of pollen trapping data from Barro anisms will predictably bias the fossil record Colorado Island, Panama, are compared with local against certain flower morphologies. These data vegetation inventories. Two hypotheses relating are of significance to those using the fossil pollen pollen representation to ‘messy’ pollination and record to reconstruct the timing and sequence of flower form are tested. Dioecious taxa were found angiosperm evolution. These data help prioritize to be over-represented in pollen spectra compared plants to be included in modern pollen reference with their occurrence in local forests. Similarly, collections and to focus the search for ‘unknown’ anemophilous and ‘messy’ pollination types were types on most-likely candidate families. found to be over-represented. While anemophilous taxa were the best dispersed pollen types, zoophilous Key words anemophily, Barro Colorado Island, taxa were also well-represented in dispersal classes dioecious, hermaphroditic, monoecious, Panama, of 20–40 m and > 40 m. Thus pollen arriving to pollen representation, pollination syndrome, trop- lake sediments is most likely to be from anemophilous ical rain forest. diversity of pollen in tropical pollen rain, it is INTRODUCTION probable that an unexpectedly large proportion The description of pollination syndromes (Faegri of pollen is coming from zoophilous taxa. & van der Pijl, 1979) pairs floral characteristics It is important to note that the pollen in a lake and likely pollinators. In temperate regions anemo- sediment or a pollen trap is wasted pollen, i.e. it philous taxa are well represented in pollen spectra failed to reach a stigma. Pollination strategies can while zoophilous species are not (Regal, 1982). range from those that produce and waste little Anemophilous species are commonly 70–100% of pollen to those that produce and waste vast amounts. the arboreal flora of temperate forest, but are It is the release of such wasted pollen that Horn usually just 2–3% of neotropical rain forest floras referred to as ‘messy pollination’ (Horn & Ramirez, (Bawa & Opler, 1975; Croat, 1978), while the 1990). Zoophily is often the product of co-evolution remainder of the flora is primarily zoophilous. of the plant and its pollinator and such specializa- Analysis of pollen trap data from New Guinea, tion, at least among temperate species, results in Ecuador and Panama (Flenley, 1979; Bush, 1991, less pollen being produced than in anemophilous 2000; Bush & Rivera, 1998) reveals that tropical counterparts (Faegri & van der Pijl, 1979). Remark- lowland pollen influx is similar in amount to that ably, despite being dominated by zoophilous of temperate forests. Given the paucity of anemo- species, tropical forests seem to produce similar philous species in the lowland tropics, and the high amounts of ‘wasted’ pollen as temperate forests. Determining a pattern in pollen representation * Corresponding author. sheds some light on the relationship between © 2001 Blackwell Science Ltd. http://www.blackwell-science.com/geb 359 GEB247.fm Page 360 Tuesday, June 5, 2001 3:53 PM 360 M. B. Bush & R. Rivera floral structure and pollinators (Bawa et al., 1985a, lesser potential ‘messiness’, we offer two testable 1985b; Renner & Feil, 1993) and helps to direct hypotheses: the construction of modern pollen reference 1 Pollen representation should follow the prob- collections toward the taxa most likely to be ability of outcrossing: dioecy > monoecy > hermaph- encountered in fossil pollen sequences. roditic self-incompatible > hermaphroditic selfers > cleistogamy (anthers and styles totally enclosed within flower). Pollination syndromes and flower structure 2 The more open the flower structure, the more A correlation has been suggested between dioecy likely the plant is to exhibit ‘messy’ pollination: and pollination systems in lowland tropical forest i.e. anemophilous taxa often have fully exposed plants (Bawa, 1974, 1994; Bawa & Opler, 1975; stamens dangling in the air, zoophilous open- Bawa et al., 1985a, 1985b; Ibarra-Manriquez & flowered taxa have exposed stamens, while taxa Oyama, 1992). Bawa & Opler (1975) noted that pollinated by birds, bats, butterflies and moths most dioecious species have small, white or yellow- often have stamens hidden deep in a corolla tube. green flowers with an open structure, and suggested As a preliminary test of the first of these that this open structure shows no co-evolution hypotheses, Bush (1995) compiled information with a particular pollinator. From this observa- from 16 lowland neotropical pollen records. The tion, it was suggested that such flowers were relative percentages of taxa that were dioecious, pollinated by small diverse insects and small monoecious or hermaphroditic were computed, bees (collectively small generalist insects; Bawa as were percentages for different pollination systems. & Opler, 1975). In an analysis of pollination When these frequencies were compared with the records for such open-structured flowers, Renner expected occurrence of dioecious, monoecious & Feil (1993) found considerable pollinator and hermaphrodites (derived from Bawa et al., fidelity, with moths pollinating 25% of zoophilous 1985a), dioecy was found to be over-represented dioecious species, bees 22.5%, beetles 20%, flies among common pollen taxa. Those data also 12.5% and diverse insects just 10%. Renner & supported the suggestion that the more open the Feil (1993) noted that despite the appearance of flower structure, the greater the probability of allophily (adapted to pollination by any visitor) representation. in these flowers, they were, in most cases, euphilic The present study compares 3 years of pollen (adapted to a particular pollinator). Unlike Bawa rain data collected on Barro Colorado Island, & Opler (1975), who contended that pollen dis- Panama, with forest inventory data to determine persal by small generalist insects was strictly the relative representation of floral types. Specif- local, Renner & Feil (1993) found the pollinators ically, we wanted to test the model proposed by of these flowers to be highly motile. Bush (1995) that sexual mechanism and pollina- Stelleman (1984) has suggested that the open tion syndrome were important determinants of structure of these zoophilous dioecious flowers the representation of neotropical pollen. and their protruding stamens leads to a great deal of pollen being blown from the flower. Indeed, he METHODS conjectured that many of these plants may shed so much pollen into the air as to be secondarily Twenty-one pollen traps (Bush, 1991) spaced 100 m anemophilous; he therefore refers to such taxa as apart were set in the mature forest of the 50 ha ambophilous. permanent vegetation plot on Barro Colorado For the purposes of this paper, the actual mode Island (Hubbell & Foster, 1990). Each trap was of pollination is of secondary importance to the left in the field for a year and replaced on the day observation that plants with these open floral of collection. Data were gathered continuously structures are likely to release significant amounts from 1991 to 1994. Of a possible 62 traps (one was of pollen that fail to reach a stigma. It is these only run for 2 years), 56 were recovered and pro- wasted pollen grains that are most likely to be cessed (Bush & Rivera, 1998). found in pollen traps and lake sediments. In the laboratory, pollen was washed free of On the basis that each of the following the rayon trap fibres, spiked with Lycopodium hierarchies represents a trend from greater to spores, and processed following the methods of © 2001 Blackwell Science Ltd, Global Ecology & Biogeography, 10, 359–367 GEB247.fm Page 361 Tuesday, June 5, 2001 3:53 PM Neotropical pollen representation 361 Faegri & Iversen (1989). The final residues were and mean percentage of pollen influx. The ratio suspended in glycerine. Pollen identification was of pollen influx to basal area (signified ‘R-rel’ carried out using a Leitz Dialux light microscope after Davis et al., 1973) was used to estimate the at 400× and 1000× (for a more detailed discussion degree of over- or under-representation for each of methods see Bush & Rivera, 1998). of these taxa. The significance of this statistic is Richard Condit of the Smithsonian Tropical assessed using an of 21 paired samples of Research Institute provided data on all woody basal area and pollen influx data for each trap plants (> 1 cm d.b.h.) within a 40-m radius of location. each trap. The list of species found within 40 m Pollen dispersal distances are based on the of the traps was divided by sexual type (di- minimum source distance to the trap, i.e. all pollen oecious, monoecious or hermaphroditic) and their is assumed to have come from the nearest potential pollination systems (Bawa, 1974; Croat, 1978; source (Bush & Rivera, 1998). Bawa et al., 1985a, 1985b; Renner & Feil, 1993). When a species potentially had more than one RESULTS pollination system, each potential system was weighted equally. Renner