Diversity and Biogeography of Neotropical Vascular Epiphytes Author(S): Alwyn H

Diversity and Biogeography of Neotropical Vascular Epiphytes Author(s): Alwyn H. Gentry and C. H. Dodson Reviewed work(s): Source: Annals of the Missouri Botanical Garden, Vol. 74, No. 2 (1987), pp. 205-233 Published by: Missouri Botanical Garden Press Stable URL: http://www.jstor.org/stable/2399395 . Accessed: 04/10/2012 16:06

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Missouri Botanical Garden Press is collaborating with JSTOR to digitize, preserve and extend access to Annals of the Missouri Botanical Garden.

http://www.jstor.org DIVERSITY AND BIOGEOGRAPHY OF NEOTROPICAL VASCULAR EPIPHYTES'

ALWYN H. GENTRY AND C. H. DODSON2

In his classic work Schimper (1888), empha- epiphyticranks. Only 32 seed plantfamilies have sizingthe taxonomicdiversity of epiphytes,list- as many as fiveor more epiphyticspecies, 26 of ed 33 familiesand 232 generaof epiphytes.Until these with epiphytesin the Neotropics. It is on veryrecently, subsequent authors have generally the 42 families(Table 3) that contain epiphytes accepted Schimper'sfigures (Richards, 1957; Jo- in the Neotropics that this paper will focus. hansson, 1974). However,epiphytism (here used Even thoughthis analysis of epiphytediversity in a broad sense to include hemiepiphytes,see and distributionis largelyfocused on the Neo- Kress, 1986) was recentlyreported to exist in 65 tropics,a fewcomparisons with the Paleotropics differentvascular plant families(56 familiesex- are instructive.There are actually slightlymore luding ferns),38 of these with epiphytesin the familieswith epiphytesin the Paleotropics (43) Neotropics(Madison, 1977). Our own data (Ap- than in the Neotropics (42), with all of the pa- pendix 1) and additional records compiled by leotropical epiphyticfamilies having epiphytic Kress (1986) now record83 vascular plant fam- representativesin Australasia but only 15 in Af- ilies withepiphytic species.3 At least 876 genera rica and Madagascar. If only the 32 seed plant include at least one epiphyticspecies and there familieswith five or more epiphyticspecies are are perhaps 29,000 epiphyticspecies, ca. 10% of considered,there are also roughlyequal repre- all vascular plants (Table 1). Thus at firstsight sentationsof epiphyte-containingfamilies in the epiphytismseems a very widespread and suc- Neotropics (26) and Australasia (25) but only cessfullife-style, which very many unrelated taxa about half as many in Africa(14). have evolved. At the species level the storyis verydifferent. However, a closer examination suggeststhat There are many more epiphytesin the Neotrop- even though there are both many species and ics, at least half again as many as in Australasia highertaxa of epiphytes,few of the highertaxa and six times as many as in Africa. Although account for most of the species. Burger(1985), similarnumbers of genera and familiesevolved forexample, emphasized thatrelatively few lin- epiphytismin the differentregions, subsequent eages have been able to enterthe epiphytic niche, speciationas epiphyteswas dramaticallygreater presumablybecause of the complex suite of ad- in the Neotropics. A major objective of this pa- aptations needed. Thus even though it is true per, then,will be to tryto explain why there is thatthe evolutionof an epiphytichabit has been so much epiphytediversity in the Neotropics. a relativelycommon featureof vascular plant evolution,it is equally true that very few of the EPIPHYTE FAMILIES taxa that have evolved an epiphytichabit have One approach to an overview of neotropical radiated successfullyto produce otherepiphytic epiphytediversity is a taxonomic one. Table 3 species (Table 2). In most of the epiphyte-con- summarizesthe neotropicalepiphytic seed plant tainingfamilies, epiphytism is a ratherinsignif- floraby family.Of the 42 families represented icant anomaly. Indeed, eliminatinga mere 85 by at least one habituallyepiphytic species in the such "oddball" species from the roster of the Neotropics,the Orchidaceae are by farthe most world's epiphytesremoves 31 familiesfrom the importantwith ten timesas manyepiphytic neo-

I We thankthe National Science Foundation (INT-790-6840; BSR-834-2764) forsupport of our Ecuadorian fieldwork.Drs. H. Bedell, T. Croat, B. Hammel, J. Luteyn,S. Renner, P. Taylor, L. B. Smith, M. Dillon, and H. Robinson provided unpublished data on distributionof epiphytismin their taxonomic specialties. We especially thank B. Hammel formaking available unpublisheddata on the habit composition of the La Selva flora;B. Burger,T. Croat, B. Hammel, M. Madison, S. Renner, and E. Zardini for reviewingthe manuscript; and E. Zardini formuch technicalassistance in its preparation. 2 Missouri Botanical Garden, P.O. Box 299, St. Louis, Missouri 63166, U.S.A. 3The data set used in this paper was derived independentlyfrom that of Kress (1986) and is mostlybased on Madison's figuresas modifiedin Appendix 1.

ANN. MissouRI BOT. GARD. 74: 205-233. 1987. 206 ANNALS OF THE MISSOURI BOTANICAL GARDEN [VOL. 74

TABLE 1. Taxonomic distributionof vascular epi- and suggested20,000-25,000 as the best esti- phytes(modified from Madison, 1977 and Kress, 1986). mate of orchid species numbers. Since 70% of the total number of orchid species should ap- Fami- Gen- proximate the number of species of epiphytic lies era with with Species orchid,there should be betweenca. 12,000 (from Epi- Epi- of Epi- AiryShaw's estimate)and ca. 20,000 (fromMad- phytes phytes phytes ison's estimate).Unfortunately, the 8,000 species Pteridophytes 13 92 2,593 "slop" betweenthese two estimatesis as greatas Gymnosperms 2 2 4 thetotal number of epiphytes in all otherfamilies combined!Madison (1977) used thehigher figure Monocots forhis calculations,whereas Dressler (1981) gen- Withoutorchids 80 2,657 erallyopted for more conservativeestimates of Orchids (fide Madi- son) 500 20,000 orchid species numbers.We have mostlytaken Orchids (compiled the highervalues since we know of many cases fromDressler, whereDressler's figuresfor species numbersare 1981) 460 15,000 significantunderestimates but none wherehe has Orchids (fide Kress, overestimated.For example, Dressler suggested 1986) 440 13,951 that there are 830 species in subtribe Epiden- Total (fideKress, drineae,whereas in Ecuador alone thereare 500 1986) 17 520 16,608 species in the single genus Epidendrum and we 17 540 Total (our estimate) 22,657' think 1,200 species would be a bettersubtribal Dicots 51 262 4,251 estimate. Total 83 896 29,505 Moreover, new orchid species are being dis- 1 Includes Madison's orchid figure(see text). covered at an astonishingrate, especially in the northernAndean region, again suggestingthat Dressler'sestimates of species numberswill have tropicalspecies as runners-upAraceae and Bro- to be adjusted dramaticallyupward. For exam- meliaceae. All threeof these most speciose neo- ple, about 2,315 orchid species are now known tropical epiphyte families are monocots. One fromEcuador, ca. 700 of these describedonly in other monocot family,Cyclanthaceae, also has the last 15 years.Nevertheless, more than 1,500 a significantnumber of epiphytes.Commelina- additional Ecuadorian orchid "morphospecies" ceae, Rapateaceae, and Philesiaceae, although have notbeen identifiedwith any published name. with few species, have an exclusivelyepiphytic Even if the 300 unaccounted fornames should genus (respectively,Cochilostemon, Epidryos, all prove applicable to the unidentifiedspeci- Luzuriaga, and (in our experience)Philesia). As mens at hand, it is inevitable that most of the summarizedby Madison (1977), the rest of the unidentifiedtaxa will prove undescribedand the epiphyticmonocot floraof the Neotropics con- list of orchids forEcuador alone will increase to sists of single species of Burmannia and Yucca well over 3,000 species. and a few Central American species of Smila- The neotropicalepiphytic dicot florais more cina. diversein familiesbut much less diversein species Since orchidsare so overwhelminglythe most than the monocots. Twenty-ninedicot families diverse group of epiphytes(about 70% of their have at least one habituallyepiphytic species in species are epiphytic),estimates of orchiddiver- the Neotropics. The largestof these are Pipera- sityare criticalto an evaluation of epiphytedi- ceae (ca. 500 spp.), Gesneriaceae (483 spp.), Me- versity. Unfortunately,orchids are amazingly lastomataceae(227 spp.), Ericaceae (ca. 300 spp.), poorly known taxonomically (compare the ca. Cactaceae (133 spp.), Guttiferae(ca. 90 spp.), 12 orchid taxonomistswith the ca. 200 system- and Marcgraviaceae (87 spp.). In addition there atists specializingin the similar-sizedCompos- are perhaps 110 neotropicalspecies of Moraceae itae). Estimatesof the numberof orchid species stranglersin the genera Ficus and Coussapoa. range from 12,000 ("some authors" fide Dress- The only otherdicot familieswith more than 20 ler, 1981) or 17,000 (AiryShaw, 1973) to 30,000 epiphyticspecies in the Neotropics are Aralia- (Madison, 1977) or 35,000 ("some authors" fide ceae, Bignoniaceae,Compositae, Rubiaceae, and Dressler, 1981). In the best available review Solanaceae (Table 3). Dressler (1981) counted almost 20,000 species In addition, Bombacaceae, though with few 1987] GENTRY & DODSON-NEOTROPICAL VASCULAR EPIPHYTES 207

TABLE 2. Largestepiphyte families (in part fromMadison, 1977).

No. Genera with No. Epiphytic Percent Family Epiphytes Species Total No. Species Epiphytes Orchidaceae 460 20,000 30,000 67 (-13,9511) (-19,1281) (73) Bromeliaceael 26 1,144 2,500 46 Araceae 15 1,1002 2,500+2 42 Polypodiaceael 53 1,023 1,100 93 Piperaceae 2 710 3,000 24 Melastomataceae 33 ca. 6473 4,7703 14 Gesneriaceae 28 5984 3,0004 20 Moraceae (incl. stranglers) 3 521 1,400 37 Ericaceae 28 4785 4,000 23 Hymenophyllaceael 2 400 600 67 Aspleniaceael 1 400 675 59 Dryopteridaceael 10 292 1,920 15 Rubiaceae 21 217 6,000 4 Lycopodiaceael 1 200 400 50 Davalliaceael 8 139 150 10 Asclepiadaceae 6 135 2,000 7 Cactaceae 25 133 2,000 7 Cyclanthaceae 7 1256 205 61 Vittariaceael 9 112 112 100 Guttiferae 6 92 1,000 9 Marcgraviaceae 7 897 117 76 Araliaceae 5 73 700 10 I Kress, 1986. 2 Croat, pers. comm. 3Renner, 1986 and pers. comm. 4Wiehler, 1983. 5Luteyn, pers. comm. 6 Hammel, pers. comm. 7Bedell, pers. comm. species,has an epiphyticgenus (Spirotheca). The typesof epiphytepropagule. The greatmajority other 16 epiphyticneotropical dicot familiesare of epiphytegenera and species have tiny dust- representedby only occasional epiphyticspecies like wind-dispersed sporochores, often with of predominantlyterrestrial genera. highlysculptured epidermis, to aid in air flota- In addition to theseangiosperm families, there tion.Such seeds, representingan extremein r-se- are two gymnospermfamilies with single epi- lection and a high risk gamble on chance es- phyticneotropical species in generallyterrestrial tablishment,are foundin thetwo most successful genera and at least 838 epiphyticfern species epiphytegroups, orchids and ferns,as well as in belongingto 32 differentgenera, a common epi- such taxa as Begonia (althoughat least one Af- phyticPsilotum, and some epiphyticspecies of rican epiphyticspecies has a fleshyfruit, pers. Lycopodium. obs.), Utricularia, Rapateaceae, and perhaps capsular melastomes,although seeds of capsular melastomesmay be 1 REPRODUCTIVE BIOLOGY mm longand are not strict- ly comparable (Renner,pers. comm.). In closed- A number of salient characteristicsthat may canopy tropical forestssuch seeds are virtually be criticalto success as epiphytesare shared by unique to epiphytes.While tinysporochore seeds many differentneotropical epiphytic taxa. Mad- are foundin some tropicalweedy herbs, they are ison (1977) nicelysummarized many of the fea- unknown among tropical forestlianas (except tures of epiphytereproductive biology and this mostlyhemiepiphytic Adelobotrys) and trees(ex- discussionis largelybased on his. From theview- cludingtree ferns), although the pterochore seeds point of dispersal biology, thereare threemain of generalike Chimarrhismay not be any larger. 208 ANNALS OF THE MISSOURI BOTANICAL GARDEN [VOL. 74

TABLE 3. Epiphyticneotropical seed plant families(in part fromMadison, 1977).

Number of Neotropical Epiphytic Neotropical Genera Family Species with Epiphytes Distributionof Epiphytes Gnetaceae 1 Gnetum also epiphyticin Malaysia Zamiaceae 1 Zamia Costa Rica, Panama; only in Neotropics Agavaceae 1 Yucca Mexico; only in Neotropics Araceae 1,034 Anthurium,Caladiopsis, Mon- also epiphyticin Africaand stera,Rhodospatha, Philo- Asia dendron,Stenospermation, Syngonium Bromeliaceae 1,144 (fide 18 genera entirelyor predomi- only in Neotropics Kress, 1986) nantlyepiphytic; 5 others with some epiphytes Burmanniaceae 1 Burmannia also epiphyticin New Guinea Commelinaceae 3 Cochliostemon,Campelia only epiphyticin Neotropics Cyclanthaceae 125 Asplundia,Dicranopygium, only in Neotropics Evodianthus,Ludovia, Sphaeradenia, Stelestylis, Thoracocarpus Dioscoreaceae 1 Dioscorea Ecuador Liliaceae 4 Smilacina also epiphyticin Australia, Pacific,and Madagascar Orchidaceae 11,000 (fide ca. 80 genera entirelyor pre- also epiphyticin Africaand Madison) dominantlyepiphytic Australasia Philesiaceae 3 Luzuriaga, Philesia also epiphyticin New Zea- land Rapateaceae 6 Epidryos,Stegolepis only in Neotropics Apocynaceae 1 Mandevilla Costa Rica Alzateaceae 1 Alzatea Costa Rica-Colombia Araliaceae 45 Schefflera,Oreopanax also epiphyticin Africaand Australasia Asclepiadaceae 2 Cynanchum mostlyepiphytic in Malaysia Begoniaceae 25 Begonia also epiphyticin Africaand Asia Bignoniaceae 29 Schlegelia, Gibsoniothamnus only in Neotropics Bombacaceae 4 Spirotheca only in Neotropics; mostly Andean Burseraceae 1 Bursera Costa Rica Cactaceae 133 25 generaentirely or predomi- Rhipsalis also epiphyticin nantlyepiphytic Africaand Ceylon Campanulaceae 7 Burmeistera neotropical;mostly Andean Compositae ca. 30 Mikania, Nelsoniothamnus, CentralAmerica and Ande- Neomirandea, Pseudogynox- an; Senecio also epiphytic ys, Senecio (Pentacalia), Sin- in New Zealand and Mad- clairia, Tuberostylis agascar Crassulaceae 2 Echeverria also few epiphyticin Hima- layas and Madagascar Ericaceae ca. 300 18 genera entirelyor predomi- also epiphyticin Australasia nantlyepiphytic; several with epiphyticspecies Gentianaceae 1 Voyria South America Gesneriaceae 483 12 genera entirelyor predomi- also epiphyticin Africaand nantlyepiphytic; 4 other Australasia 1987] GENTRY & DODSON-NEOTROPICAL VASCULAR EPIPHYTES 209

TABLE 3. Continued.

Number of Neotropical Epiphytic Neotropical Genera Family Species with Epiphytes Distributionof Epiphytes genera with some epiphytic species Griseliniaceae 3 Griselinia Chile and Brazil; also epiphyticin New Zealand Guttiferae ca. 90 Clusia, Clusiella, Havetiopsis, only epiphyticin Neotropics Oedematopus, Quapoya, Renggeria Lentibulariaceae 12 Utricularia also 2 epiphyticin Africa and Australasia Marcgraviaceae 89 all 7 genera entirelyor pre- only in Neotropics dominantlyepiphytic or hemiepiphytic Melastomataceae 227 7 genera entirelyor largelyepi- also epiphyticin Africaand phytic;5 otherswith some Asia epiphyticspecies Moraceae 111 Coussapoa, Ficus subg. Uro- Ficus stranglersalso in Afri- stigma,and 1 Pourouma ca and Australasia Myrsinaceae ca. 12 Cybianthus,Grammadenia, also epiphyticin Africaand Myrsine(Rapanea) Asia Onagraceae 3 Fuchsia only epiphyticin Neotropics Piperaceae ca. 500 Peperomia, Piper also epiphyticin Africaand Asia Rubiaceae ca. 57 Balmea, Coprosma, Cosmi- also epiphyticin Australasia buena, Hillia, Malanea, Ma- nettia,Psychotria, Ravnia, Relbunium,Schradera Sapotaceae 1 Bumelia Costa Rica Saxifragaceae ca. 3 Hydrangea,Phyllonoma only in Neotropics Solanaceae ca. 30 Juanulloa, Lycianthes,Markea also epiphyticin Malaysia (+ segregates),Solanum (Solanum, Lycianthes) Urticaceae ca. 15 Pilea also epiphyticin Indo-Ma- laysia

The second most prevalent dispersal mode in epiphytes is wind-dispersal via winged or among epiphytes is via birds. Most bird-dis- plumed seeds (pterochoryand pogonochory,re- persed epiphyteshave indehiscentberry fruits spectively).Interestingly, plumed seeds as com- but a few,including Drymonia and Clusia, have pared with winged seeds greatlypredominate dehiscentcapsules with arillate seeds. In either amongepiphytes, whereas the opposite holds true case the seeds tend to be smaller and more nu- for trees and lianas, at least in mature forest merous than in relatednonepiphytic taxa (Mad- species.Some ofthe important epiphyte taxa with ison, 1977). In some familiesthere is a marked pogonochore diaspores are Bromeliaceae changein dispersalmode accompanyingthe shift subfamilyTillandsioideae, Asclepiadaceae, Ges- to epiphytism. In Bignoniaceae all epiphytic neriaceae, and Rubiaceae. In Rubiaceae some species (with two probablybat-dispersed excep- epiphyticgenera have small winged seeds while tions) are bird-dispersedbut only one nonepi- othershave truepogonochores; the difference be- phyticspecies (Synapsis ilicifolia)is (see Gentry, tweenthese dispersal modes tendsto breakdown 1983). In Melastomataceae 85% of the epiphytic in such groups,with some species having such species have berryfruits as compared with 60% narrowreduced wings that these effectivelyap- of the nonepiphyticspecies (Renner, 1986). proximatelarge hairs. The thirdmajor diaspore dispersal syndrome Finallythere are a fewepiphyte taxa withsuch 210 ANNALS OF THE MISSOURI BOTANICAL GARDEN [VOL. 74 miscellaneous dispersal syndromes as bat-dis- pollination specializations (e.g., Dodson, 1967; persal (some stranglerfigs), exozoochory via van der Pijl & Dodson, 1966; Williams, this stickydiaspores (some Peperomia), and the not symposium),many aroids have similar Euglos- readily classifiable "sloppy corn-on-the-cob" sine-attractingscent systems.Five of the largest ingestionof some cyclanthfruits by Callicebus neotropicalorchid genera have specificbee pol- and otherprimates (Terborgh, pers. comm.). linatorsattracted by specificscents and Anthur- In general,epiphyte seeds are smallerand more ium is the largest nonorchid genus in the numerousthan those of nonepiphyticrelatives. Neotropics. While Pleurothallis, the largest For example, Renner (1986 and pers. comm.) neotropicalepiphyte genus, does not participate noted thatin Melastomataceae mostlyepiphytic in the presumablyspeciation-promoting Euglos- Blakea and Topobea have ca. 1,000 seeds per sine-pollinationsyndrome, it is pollinatedby the fruitcompared with a fewdozen seeds per fruit large and diverse flygenus Bradesia (CD, pers. in typicalnonepiphytic genera such as Miconia obs.) and similar coevolutionarypatterns may and Clidemia. Madison (1977) estimated that be involved; among othersmall-flowered and in- seeds of epiphyticAnthurium are typicallyca. 2 conspicuous but highlydiverse orchid genera, mm long as compared with4-8 mm long in ter- Stelis and Lepanthes are probablypollinated by restrialAnthurium species. There are also excep- Drosophila and similar flies, and Telipogon is tions to this pattern.For example, Rockwood pollinated by pseudocopulation with tachinid (1985) pointed out that in Gesneriaceae epi- flies,another very large and diversifiedinsect phyticspecies actually have significantlylarger taxon. seeds than do shrubs and herbs. According to Unique to the Neotropics,hummingbird-pol- Rockwood's (1985) analysis, epiphyteseed size linationis also much more prevalentamong epi- tends to be bimodal; those groups with dust- phytes(and terrestrialherbs) than in treesor free- seeds or other wind-dispersedseeds have the climbing lianas. Epiphytic taxa among which smallest seeds of any habit type while taxa not hummingbird-pollinationis prevalent include dispersed by wind have seeds averaging larger Ericaceae, Bromeliaceae, Gesneriaceae (espe- than those of herbs, vines, and shrubs,similar cially Columnea), Marcgraviaceae (Norantea, to lianas, and only smallerthan trees.Neverthe- sensu lato), Rubiaceae (Ravnia, Manettia), and less, since the great majority of epiphytes are Cactaceae (e.g., Schlumbergera).The correlation wind-dispersed,the epiphyte habit class as a between hummingbird-pollinationand epiphy- whole is generallycharacterized by the smallest tism is well shown by Bignoniaceae. Of the two seeds of any habit class. A dispersalstrategy em- epiphyticgenera of Bignoniaceae, one (Gibson- phasizing many small seeds and chance estab- iothamnus) is entirelyhummingbird-pollinated lishmentis typical of the r-selectionsyndrome and the other(Schlegelia) also has several hum- oftenfound in weedy species. Epiphyteswould mingbird-pollinatedspecies; hummingbird-pol- seem to be mostunusual in beingr-selected com- lination is rare elsewhere in the family.Other ponents of matureforest ecosystems. specialized pollination systems shown by epi- phytesinclude hawkmoth-pollinationin Cacta- ceae (e.g., Epiphyllum) and Rubiaceae (e.g., POLLINATION Cosmibuena, Hillia, perhaps Schradera), bat- Madison (1977) emphasized animal-pollina- pollinationin Marcgravia (thoughmost species tion as a characteristictrait shared by all angio- may be autogamous, Bedell, pers. comm.), and sperm epiphytes.While true, this is hardly re- rat-pollinationin Blakea chlorantha (Lumer, markablein a tropicalcontext since, with virtually 1980). Perhaps more strikingthan the diversity no exceptions(Myriocarpon (pers. obs.), Trophis of highlyspecialized pollinationsystems among (Bawa et al., 1985), and just possibly Sorocea neotropicalepiphytes is theirlack of the small, and a few Chamaedorea species (fide Bawa et inconspicuous,generalist-pollinated flowers that al., 1985)), all lowland tropical forest angio- characterizethe great majority of treesin the wet spermsare animal-pollinated.Nevertheless, epi- forestswhere epiphytesare prevalent.The only phytesas a whole surelyhave a more pronounced epiphytictaxa characterizedby such flowersare trendtoward highly specific and specialized pol- Araliaceae, Moraceae, Piperaceae, Myrsinaceae, lination systems than do nonepiphytes,if for and Urticaceae, the lattertwo only marginally no otherreason than thatso many epiphytesare epiphytic.If hummingbirds,well known as spe- orchids. In addition to the well known orchid cialist pollinators(Stiles, 1981), are taken as an 1987] GENTRY & DODSON-NEOTROPICAL VASCULAR EPIPHYTES 211

example, this patternmay be clearlyseen at the uals in drierhabitats. Although this pattern seems community level: in lowland tropical forests obvious, it is by no means well-documented.In- hummingbird-pollinationis almost exclusively deed Walter (1985: 57) claimed that epiphytes, confinedto herbs and epiphytes. contraryto such otherhabit groupsas lianas, are Another relevant aspect of epiphytepollina- found in dry as well as wet tropical forests.At tion biologyis thatsuch phenomena as self-com- the other extreme,Schimper (1903) suggested patibilityand autogamy are apparently much thatin areas withmarked dry seasons epiphytes more prevalentthan typicalin tropical lowland are eithercompletely wanting or rare and that taxa. For example seven of seven species of Bla- presence of epiphytesoutside the rain forestis kea, Topobaea, and Adelobotrystested at Mon- always a sign that the dry season is not long or teverde, Costa Rica were self-compatible(Lu- is accompanied by copious dew. Data for 1,000 mer, 1980; Renner, 1986) vs. 34 of 43 tested m2 samples of western Ecuadorian dry forest terrestrialmelastome species in the Manaus area (Capeira, 804 mm per year) and moist forest (Renner, 1984). In Marcgraviaceaethis is carried (Jauneche,1,855 mm) quantifythe extentof this to an extremewith all species tested being au- difference(Gentry & Dodson, 1987; Table 4, Fig. togamouslypollinated in bud despite the elab- 1). If our resultsare indicative,most plants in a orate floraladaptations (Bedell, pers. comm.). wet forestare epiphytes.At Rio Palenque such In summary,epiphyte reproductivebiology a sample included 4,517 epiphyticplants rep- appears to be a unique mix of r-selectionand resenting63% of all individuals sampled. At specialization. Unlike othercomponents of ma- Capeira a mere ten epiphyticplants were includ- ture forestcommunities, epiphytes share many ed in a similarsample of dryforest, representing reputedlyr-selected traits with weedy herbs,es- 0.2% of the sampled individuals. The moist for- peciallyin theirdispersal ecology. Yet at the same est Jaunechesite was intermediatewith 116 epi- time most epiphyteshave highlyspecialized pol- phytesconstituting 4% of the individual plants lination systems,strong niche specificity,and sampled. The differencein epiphytedensity be- many other traits more characteristicof k-se- tweenwet and dryforest is almost 500 fold. lected matureforest species. Moreover the decrease of epiphytedensity in dryforest contrasts greatly with the situationfor DISTRIBUTIONAL PATTERNS other habit groups. The number of herbs more thandoubles fromour moistand wet forestsam- To thispoint, we may conclude that,although ples to our dry one. Contraryto Walter's asser- an epiphytichabit has arisen many times during tion (1985: 57), lianas double fromwet to dry the course of plant evolution, very few higher forest;they are much commoneryet in our moist taxa have been more than marginallysuccessful forestsample, the latter presumably atypical since at speciation and adaptive radiation as epi- Jaunechehappens to be the most liana-richsite phytes. However, the few taxa that have suc- in the Neotropics (of 45 similarsamples). Shrub cessfullyradiated as epiphyteshave done so very densityalso increasessomewhat from wet to dry prolifically.Even though remarkably similar forest;unlike lianas, shrubs are only about half numbers of plant families have achieved epi- as abundant in the intermediatemoist forestas phytismin the Neotropics and Paleotropics,the in dryand wet forest.In contrast,the numberof process of epiphyte speciation would seem to individual trees -10 cm DBH, and thus the ap- have been much accelerated in the former,to parent densityof the forest,changed little (52, judge from the very many more neotropical 64, and 69 trees >-10 cm DBH in wet, moist, species of epiphytes. and dryforest, respectively). We would now like to examine some trends Epiphytesare also importantcontributors to in epiphytedistribution that may help to under- the species richness of neotropical wet forests. stand not only neotropical epiphytebiogeogra- Indeed thereare 35 epiphytespecies in our 0.1 phy but also some of the continentaldifferences ha. wet forestsample, accountingfor over a third in epiphyteoccurrence. of the sampled species. This compares withonly 13 epiphytes(8% of the species) in the compa- MOISTURE rable moist forestsample and three (2% of the One of the most strikingdistributional pat- species) in the dry forestone (Fig. 2). terns shown by epiphytesis a tremendousde- The importanceof the epiphyticcontribution crease in both numbers of species and individ- to species diversityis equally apparentwhen en- 212 ANNALS OF THE MISSOURI BOTANICAL GARDEN [VOL. 74

TABLE 4. Number of species and individuals of differenthabit typesin 1,000 m2 samples of threeforests in westernEcuador. Rio Palenque is wet forest,Jauneche is moist forest,Capeira is dry forest(from Gentry & Dodson, 1987).

Rio Palenque Jauneche Capeira No. No. No. No. No. No. Habit Group Spp. % Ind. % Spp. % Ind. % Spp. % Ind. % Herbs (incl. palmettos) 50 14 1,220 17 18 11 944 34 50 29 2,854 53 Shrubs 39 11 531 7 16 10 279 10 13 8 742 14 Epiphytes(incl. hemiepiphytes) 127 35 4,517 63 13 8 116 4 3 2 10 .2 Climbers (incl. lianas; excl. hemiepiphytes) 36 10 117 2 58 34 484 17 58 34 895 16 Lianas ->2.5 cm (excl. herbaceous + hemi- epiphyt.) 12 3 28 .4 43 25 124 4 19 11 58 1 Total tree spp. (incl. ju- ven.) 114 31 653' 9 64 38 960 34 48 28 927 17 Trees <2.5 cm DBH (= saplings + seedlings) 87 24 5591 8 48 28 672 24 38 22 750 14 Trees -lO cm,-2.5 cm 86 24 217 3 35 21 245 9 32 18 108 2 Trees-1 Ocm DBH 32 9 52 1 30 18 64 2 29 17 69 1 Total herbs (epiphytes+ herbs-woody epiphytes) 162 5,525 31 1,060 53 2,864 Shrub layer (shrubs + saplings) 126 1,090 65 948 51 1,492 Underlayers(< ca. 3 m) (herbs + shrubs + saplings) 176 2,310 83 1,892 101 4,346 Total 365 7,210' 169 2,783 173 5,428 1 Excludingdense patch of 123 Quararibea asterolepisseedlings. tire florulasare compared (Table 5, Fig. 3). Al- changeswith precipitation (Table 6). Many more most one fourthof all the Rio Palenque plant families have epiphyticrepresentatives in wet species are epiphytes(Dodson & Gentry,1978). foreststhan in drierones (Fig. 4), and many epi- Similarly,in another lowland wet forestat La phytictaxa are confinedentirely to wetterforests. Selva, Costa Rica, 25% of the species are epi- In the Neotropics the same familiestend to be phytes(Hammel, pers. comm.). Even in moist representedby epiphytesunder similar climatic forestsites like Barro Colorado Island, Panama conditions. In the driest forests,the only epi- (Croat, 1978) and Jauneche,Ecuador (Dodson phytesare orchids and bromeliads,perhaps the et al., 1985), epiphytesconstitute 12-16% of the two most specialized epiphyticfamilies. Ferns, total flora.Only in dryforests are epiphytesrel- peperomias,and Cactaceae join orchidsand bro- ativelyinsignificant, accounting for 2-4% of the meliads in slightlymoister conditions. The next species of Capeira (Dodson & Gentry,1987) and epiphyticfamilies to appear with increasinghu- Santa Rosa National Park, Costa Rica (Janzen midity are aroids, Moraceae (stranglers),and & Liesner, 1980). Gesneriaceae, joining representativesof all the We conclude, contraryto Walter (1985), that dryforest families, each of whose numberof epi- epiphytesdecrease moredrastically in drierareas phyticspecies is maintained or increased. The than does any otherhabit group,but contraryto two local florula sites with over 2,500 mm of Schimper's(1903: 198) emphasis,a fewvascular precipitation, Barro Colorado Island (Croat, epiphytesare characteristicallypresent in even 1978) and Rio Palenque (Dodson & Gentry, the driestneotropical forests (e.g., Capeira with 1978), have remarkablysimilar epiphytic floras. 804 mm of annual precipitation). The same seven epiphytic families are most Familial makeup of the epiphyticflora also species rich at both sites, in roughlythe same 1987] GENTRY & DODSON-NEOTROPICAL VASCULAR EPIPHYTES 213

100-

_ epiphytes

WET 4 trees 63 4 DRY herbs 60 53 * shrubs

50 00 climbers

MOIST 40 34 34

16 20 17 114 17 17

o0 0 4 4 0 0 9 4 00 ~~~~000 0 ~ 4 | *0 *00 2 0.2 04 0 *00* 0 04

Capeira Jauneche Rio Palenque

FIGURE 1. Percent of individual plants belongingto differenthabit groups in 1,000 m2 samples of three westernEcuadorian forests.

50 - OM ~~~~~~~~~~~~~~~~~~~~~~~epiphytes DRY MOIST WET trees

40 -38 No. ~~~35 herbs 34 I4*shrubs

0 ~ ~ ~ ~ ~031~ ~ 4 0 2899 o04 0 climbers u 3030 ~~~8 004 0 0 4 0 25 o 0~~~~~~~~ 0~~~~40 0 4 0 00 4o0 Bo40 20 o0 Bo. 0

4 o0 00

0 lo ~~~~~0 0 o 4 0 4 No4 8 0 00 0A 0 11 0 4 11 10 00 , 0 ~ ~ ~ 2 N.00, .0, ~ 0 0ol * 0 o 0 10 0aer 4aue *h Ri4 *alnq

FIGUE 2.Percnt o spce 0eogn odfeethbtgop i ,0 2smlso he et Ecuadoria4n*for0est*. 214 ANNALS OF THE MISSOURI BOTANICAL GARDEN [VOL. 74

DRY MOIST WET ? a epiphytic representativesin the Rio Palenque florabut not the Barro Colorado one. Again, the same epiphyte families predomi- nate at otherwet sitesin westernEcuador (Table 7). The exact same seven most species richfam- ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~- ilies at BCI, La Selva, and Rio Palenque are the most species rich at Centinela,again in roughly the same order. Even at Tenafuerste(alt. 1,000 m) the fivefamilies richest in epiphytesare exactlythe same ones thatare most species richat the otherEcuadorian wet forestsites. At the extremewet end of the precipitation gradient there are other, as yet unquantified, changesin the epiphyticflora. In the wetterpart of Choc6 (precipitation > 8,000 mm) berry- CAPEIRA RSR~sA JAUNECHE B. C I PALENOUE SELVA fruited epiphytes like Melastomataceae, Ara- FIGURE 3. Percentof speciesin local florulasbe- ceae, Marcgraviaceae,and Ericaceae, along with longingto differenthabit groups. arillate-seeded Guttiferae,appear to predomi- nate to a much greaterdegree than at Rio Pa- lenque, while ferns(Sota, 1972) and dust-seeded order,just as theyare in Hammel's unpublished orchids are more poorly represented.Possibly La Selva, Costa Rica, species list. Remarkably, thischange is relatedto theinability of dust-seeds thereare 81 epiphyticspecies of orchids at Rio* to establish themselves in the face of such su- Palenque and 82 at Barro Colorado Island (BCI), perabundantrainfall. 35 epiphyticaroids at Rio Palenque vs. 30 at There is also a noticeable habit change in the BCI, 18 species of epiphyticBromeliaceae, and epiphyticflora of the wettestsites. Most of the 14 species of epiphyticor stranglerMoraceae at predominantChoc6 epiphyticfamilies are ac- both sites. Only epiphyticferns and fernallies tually hemiepiphytic,many of them woody as are noticeablybetter represented at BCI than at well. In the same sites with abundant hemiepi- Rio Palenque and only epiphyticGesneriaceae phytes,free-climbing lianas become noticeably and Cyclanthaceaeshow thereverse pattern. Both less prevalent. While the average of climbers siteshave threeepiphytic cacti and one epiphytic - 2.5 cm in diameterfor 0.1 ha. samples at two melastomeplus one or two epiphyticAraliaceae, pluvial forestsites in the Colombian Choc6 was Solanaceae, and begonias. Only Rubiaceae and exactlythe same (68) as that for a series of 20 Saxifragaceae have epiphyticmembers at BCI similarsamples fromneotropical lowland moist but not Rio Palenque; only Urticaceae, Pole- and wet forests,half the sampled pluvial forest moniaceae, Bignoniaceae, and Ericaceae have climbers were hemiepiphyticvs. an average of

TABLE 5. Habit compositions of complete local florulas.Capeira, Ecuador and Santa Rosa National Park, Costa Rica, are dryforest; Jauneche, Ecuador and BarroColorado Island, Panama, are moistforest; Rio Palenque, Ecuador, is wet forest.

Santa Barro Rio Capeira Rosa Jauneche Colorado Palenque Habit Category No. % No. % No. % No. % No. % Trees-1O cm DBH 69 15 141 21 112 18 290 22 165 16 Small trees + large shrubs 28 6 64 10 60 10 151 11 99 9 Herbs + subshrubs 242 52 317 48 224 37 389 30 376 36 Epiphytes(including stranglers) 9 2 24 4 72 12 216 16 238 23 Parasites and saprophytes 4 1 6 1 4 1 12 1 6 1 Lianas 46 10 52 8 81 13 149 11 87 8 Small vines 66 14 63 9 55 9 109 8 84 8 Total species 464 667 608 1,316 1,055 1987] GENTRY & DODSON-NEOTROPICAL VASCULAR EPIPHYTES 215

TABLE 6. Familial composition of epiphytefloras in lowland forestswith differentprecipitations.

Santa La Selva, Rio Barro Jauneche, Rosa, Costa Palenque, Colorado, Ecuador4 Costa Capeira, Makokou, Rica' Ecuador2 Panama3 1,855 Rica5 Ecuador6 Gabon7 Family 4,000 mm 2,980 mm 2,750 mm mm 1,550 mm 804 mm 1,755 mm Orchidaceae 109 81 82 33 8 5 21 Araceae 76 35 30 10 - - 10 Ferns and allies 59 28 43 5 7 - 26 Piperaceae 12 19 10 4 1 1 - Bromeliaceae 29 18 18 6 3 2 - Moraceae 13 14 14 9 2 - 7+ Gesneriaceae 16 12 4 2 - - Cyclanthaceae 11 8 1 - Marcgraviaceae 8 5 2 Guttiferae 11 4 2 - - - - Cactaceae 6 3 3 3 3 1 1 Ericaceae 2 3 - - - - - Araliaceae 2 2 1 - - - 1 Bignoniaceae 3 2 - Melastomataceae 2 1 1 Polemoniaceae - 1 - Solanaceae 1 1 2 Urticaceae 1 1 - Begoniaceae 2 1 1 Rubiaceae 4 - 1 Saxifragaceae - - 1 Commelinaceae 1 - - Total epiphytes 368 238 216 72 24 9 66+ Percentof flora 24 23 16 12 4 2 6+ No. familieswith epiphytes 20 18 17 8 6 4 6 ' B. Hammel, pers. comm. 2 Dodson & Gentry,1978. 3Croat, 1978. 4 Dodson et al., 1985. 5 Janzen & Liesner, 1980. 6 Dodson & Gentry,1987. 7 Hladik & Gentry,in prep.; Florence & Hladik, 1980 and cited references.

2.7 hemiepiphyticclimbers sampled at the moist Ecuadorian sites (Table 7) documentsthis trend and wet sites (Gentry,1986). In a sense, hemi- forthe lower part of the gradient.The sites for epiphyticclimbers seem somehowto replacefree- which relevantdata are available are Centinela climbinglianas in thewettest lowland forestsand (600 m) and Tenafuerste(1,000 m), both on the also in middle elevation cloud forests. westernslopes of the CentralEcuadorian Andes, and Mera (1,000 m) on the easternslope. Of the well-documentedsites, Centinela has the most ALTITUDINAL species of epiphytes,337, or 35% of the flora. The epiphyticflora also changes in both di- This compares with 238 epiphyte species ac- versityand composition on an altitudinalgra- countingfor 23% of the floraat nearbyRio Pa- dient. The general tendencyis for epiphytesto lenque (alt. 200 m). The data forthe two 1,000 be betterrepresented in intermediateelevation m sites are less complete,with many species re- cloud forests.In the Andes the peak in epiphyte mainingto be discovered.Tenafuerste, with much diversityappears to be between 1,000 m and less cloud foresteffect than Centinela or Mera, 2,000 m, but it lies somewhat lower in Costa has thepoorest epiphyte representation, only 31% Rica and Panama. Few data are available, but a of the flora.The extremeis pluvial Mera where comparison of incomplete data sets for several few collections have been made as yet. Three 216 ANNALS OF THE MISSOURI BOTANICAL GARDEN [VOL. 74

NUMBER OF EPIPHYTE FAMILIES AND SPECIES VS. PRECIPITATION a)~~~~~~~~~~~~~~~~~~~~~~~ a 400- .20 e

* number of families * number of species

300- 15 E

0 E.0~~~~~~~~~~~~~~~~~~~~c z .-~ 10 200- E coCZ0:

100~ ~ ~ ~ ~ ~~~~~~~~~~

1000 2000 3000 4000

Annual precipitation (mm) FIGURE4. Increase in numbersof epiphytespecies and familiesin local florulasas a functionof precipitation. hundred twenty-twoorchid species are already gradient,familial composition of the epiphyte knownand we expectthat there may be as many florachanges verylittle along an altitudinalgra- as 200 additional epiphytesas well. dient,or at least on that part of it forwhich we In contrast,Peru's Huascardn National Park, have data. The same families are importantin 3,500-5,000 m in altitude,has only seven epi- roughlythe same order.The seven familieswith phytic species, constitutinga mere 1% of the the most epiphytespecies at Centinela (600 m) park's flora(Smith, pers. comm.). are the same seven thathave the most species at The differencesbetween Mera and Tenafuerste Rio Palenque (200 m). The fivefamilies with the point out thataltitudinal and moistureeffects on most species at Tenafuerste(1,000 m) are the epiphytesare complexly interrelated.Similarly same fivethat have the most species at Rio Pa- Gilmartin(1973) showed that the 17 species of lenque and Centinela.The most noteworthydif- Bromeliaceae thatoccur on both sides of the Ec- ferencebetween these sites is the absence of Mo- uadorian Andes occur at lower altitudeson the raceae stranglersat Tenafuerste;Gesneriaceae are moistereastern slopes. also conspicuously less diverse at the 1,000 m Density of epiphytes,although also greatestat Tenafuerstesite but Ericaceae are more diverse intermediatealtitudes, does not closely parallel (nine species vs. threeat Rio Palenque). Melas- diversity;although unquantified, we feelthat epi- tomataceae is another familythat has notably phytedensity in the Andes tends to be greatest moreepiphyte diversity at intermediatealtitudes around 2,000-2,500 m, again occurringat some- (four spp. at Centinela and two at Tenafuerste what lower elevations in Costa Rica and Pana- but only one at Rio Palenque; see also Renner, ma. Due to thehigh densities, epiphytes are often 1986). A few families (e.g., Bignoniaceae) dis- most conspicuous at these relativelyhigh alti- appear fromthe middle elevation epiphyteflora tudes,even thoughrelatively few species may be and most other families have decreasing num- present.In middle elevation cloud forests,epi- bers of epiphytesat higherelevations. phytesmay make up as much as 30% ofthe foliar Our only high altitude data set is for Huas- biomass and 45% of the foliarmineral capital of caran National Park, Peru (Smith,pers. comm.), a forest(Nadkarni, 1984). where the seven epiphytespecies, all restricted Contraryto what happens along the moisture to the lower part of the park between 3,500 and 1987] GENTRY & DODSON-NEOTROPICAL VASCULAR EPIPHYTES 217

4,000 m, belong to fourfamilies. The fourfam- TABLE 7. Familial composition of epiphytefloras ilies with epiphytes-Piperaceae, Bromeliaceae, in wet forestsat differentaltitudes in Ecuador. Orchidaceae, and ferns-are all in the top five epiphyticfamilies in the wet Ecuadorian sites. Rio Pa- Centi- Tena- lenque nela fuerste Of the usually prevalentepiphyte families, only Family 200 m 600 m 1,000 m Araceae is lacking.Perhaps more interesting,the epiphytefamilies at Huascardn are exactly the Orchidaceae 81 133 68 Araceae 35 52 same ones that are representedat Santa Rosa, 26 Ferns 28 38 28 Costa Rica, exceptthat Cactaceae is missing.Ap- Piperaceae 19 19 11 parentlyat environmentalextremes, either alti- Bromeliaceae 18 23 18 tudinal or precipitational,only these same fam- Moraceae 14 10 - ilies that are otherwise most successful as Gesneriaceae 12 16 8 epiphytesare able to survive. Cyclanthaceae 8 5 3 The veryinteresting but controversialsugges- Marcgraviaceae 5 3 2 tion has been made that in the tropicsdiversity Guttiferae 4 9 3 is generallygreatest at middle elevations along Cactaceae 3 2 1 an altitudinalgradient. This has been shown for Ericaceae 3 9 9 Araliaceae 2 4 - leaf litterherps (Scott, 1976), insects (Janzen, Bignoniaceae 2 2 - 1973; Janzen et al., 1976), and suggested for Melastomataceae 1 4 2 plants. Greaterequability is a likelycontrolling Polemoniaceae 1 1 - factorfor this putative "mid-elevation bulge" in Solanaceae 1 2 1 species diversity.However, data for0.1 ha. sam- Urticaceae 1 1 1 ples of plants -2.5 cm DBH suggestthat plant Acanthaceae - 1 - species diversitydecreases more or less uniform- Rubiaceae - 1 - ly fromthe most diverselowland wet forestsites Total 238 337 181 to the least diverse high altitude ones (Gentry, Percentof flora 23 35 31 1982a, 1987c). If a mid-altitudebulge in plant species richnessreally does occur,it mustbe due largelyto epiphytes.Unfortunately our data sets habitats.Whitmore (1984) also emphasized the frommiddle and upper elevation forestsare too frequencyof epiphytesin these forests. incomplete to be definitive.Indeed one of us It is increasinglywell-documented that major (CD) thinksthat because of the increase in epi- changesin thediversity and floristic phytes there are more plant species at middle composition of other components of elevations than in lowland tropical forestwhile tropical plant commu- nitiesare associated withchanges in one of us (AG) thinksthat the decrease in species soil fertility (e.g., Ashton, numbersof such otherhabit groups as lianas and 1976, 1977, 1978; Huston, 1979, 1980; Gentry,1987b; Gentry& treeswith altitude outweighs the increasednum- Emmons, 1987). For example,there are generallyfewer tree, ber of epiphytes.In eithercase the role of epi- liana, and terrestrialherb in phytes in the plant communityis presumably species neotropicalforests on greatestin middle elevation forests. poorer soils (Gentry, 1981; Gentry& Em- mons, 1987). We have few data with which to relate epiphyte SOIL FERTILITY communitycomposition to soil fertility.One of us (AG) has compiled species To our knowledge no attempt to relate epi- listsfor a seriesof sites on differentsubstrates in phyte diversityto soil fertilityhas been made theIquitos, Peru,area whichshare a similarrain- previously.Indeed one mightsuppose thatsince fall and climatic regime. Of these, the site with epiphytesare intrinsically"insulated" fromdi- the poorest soil (Mishana, on almost pure white rectdependence on soil nutrientsthey would be sand) has the fewest epiphytes (31 epiphyte relativelyunaffected by changes in soil fertility. species plus a few"indets." in a relativelyinten- For example,Janzen (1 974a) discussed the sym- sivelyinventoried area vs. 38 identifiedand many bioticrelationships between several epiphytes and unidentifiedat less intensivelystudied better-soil ants in low-diversitypoor soil "kerangas" hab- Yanamono), suggestingthat epiphytediversity itats in Borneo, with the implication that epi- varies with soil fertilityas does the diversityof phytes are unusually well representedin such other habit groups. However, to date the sam- 218 ANNALS OF THE MISSOURI BOTANICAL GARDEN [VOL. 74 pling of epiphytesat these sites has been much to the effectsof relativelyslight decreases in en- less intensivethan at our Ecuadorian local florula vironmentalfavorability. sites and is probablytoo haphazard and incom- LATITUDE plete to make these data verymeaningful. Data are also available for understorycom- It is well known that the presenceof vascular position and levels of floweringand fruitingfor epiphytesis a characteristicof tropicalforests as a broad arrayof neotropical(and paleotropical) comparedwith temperate ones. We knowof only sites (Gentry& Emmons, 1987). If the data for fourvascular epiphytesthat occur northof Flor- epiphytesare extractedfrom that data set,a very ida in the temperateUnited States- Tillandsia strongreduction in numbersof fertilespecies of usneoides, T. recurvata(only in southernmost understoryepiphytes on poorersoils is apparent, Arizona), Epidendrumconopseum, and Polypo- parallelingthe overall trend of decreased num- dium polypodioides.Even Tillandsia usneoides, bers of floweringand fruitingunderstory species the northernmostvascular epiphyte,does not on the same gradient.Our data for habit com- reach the Mason-Dixon line. The only continen- positionof the understory suggest that epiphytes, tal United States epiphyte species that is not like terrestrialherbs, are especially sensitive to widespread in the tropicsis Tillandsia simulata loss of soil fertility:as soil fertilitydecreases, Sm., endemic to Central Florida but sometimes terrestrialherbs, epiphytes,understory shrubs, lumpedwith T. bartramiiEll. (The Mexican range and lianas disappear fromthe understoryin that ofEpidendrum conopseum is also somewhatlim- sequence,leaving virtually only tree saplings and ited.) Even in very wet areas that would be full seedlings in the most severely stressed forests ofepiphytes in thetropics, only lower plants have (Gentry& Emmons, 1987). adopted theepiphytic habit. In intermediatesub- Anecdotal evidence also indicates that epi- tropicalareas a gradientof increasing epiphytism phytesare much less diverse and abundant on at loweraltitudes is evident.The decrease in vas- poor soils. We have observed many fewerepi- cular epiphyteswith increasinglatitude can be phytesin poor soil parts of Central Amazonian clearlyseen in Florida where subtropicalSouth Brazil, southernVenezuela, and elsewherein the Florida (latitude 250N) has 46+ (= 2.8%) epi- Guiana shield area than in partsof the Neotrop- phytic species (Long & Lakela, 1971), Central ics withricher soils. Large-scalebiogeographical Florida has 41 (= 1.9%) (Wunderlin,1982), and analysis also indicates that epiphytictaxa are Florida Caverns State Park in northernFlorida poorlyrepresented in these areas compared with (30'50'N latitude) has only two (= 0.4%), Til- richersoil areas nearerthe Andes and in Central landsia usneoidesand Polypodiumpolypodioides America (Gentry,1 982b). In a somewhatdiffer- (Mitchell, 1963). ent context,Janzen (1977) has suggestedthat the Curiously,the decrease in vascular epiphytes paucityof epiphytes in dipterocarpforests results with increasinglatitude is not symmetricalon fromthe generally nutrient-poor Southeast Asian both sides of the equator. A numberof epiphytic soils. Among the evidence for this hypothesis species and even a few endemic genera of epi- cited by Janzen (1977) is the observation that phytesoccur in southtemperate forests. Endemic trees cultivated in Malesia along roads, where temperateSouth American epiphyticgenera in- dust stirredup by passing vehicles increases the clude the monotypicfern Synammia, the mono- nutrientsavailable to epiphytesin an otherwise typic cactus Pfeiffera,three monotypic Gesne- unusuallypoor-soil situation, have relativelylarge riaceae genera (Asteranthera,Sarmienta, and epiphyteloads compared withthe native forests. facultativelyepiphytic Mitraria), and the Lili- On balance it seems clear that the epiphytic aceae (or Philesiaceae) Luzuriaga and Philesia. plantcommunity is verysensitive to soil fertility, In the Australasian region,New Zealand is es- withfewer epiphytes and fewerepiphytic species pecially noteworthyfor its autochthonous epi- in forestson poorer soils. Indeed epiphytedi- phytesincluding genera like the monotypicfern versitymay be even more sensitiveto change in Anarthropteris,the liliaceous Collospermum(also soil fertilitythan is treeor liana diversity,a sug- reaching Fiji and Samoa), the only epiphytic gestionthat would accord with the idea (Gentry species of families like Cunoniaceae (with two & Emmons, 1987) that plants (presumablyin- differentgenera having epiphytic members), and cludingepiphytes despite their lack ofdirect con- generalike Microlaena (Gramineae) and Metro- tact withthe soil) thatare barelyable to eke out sideros (Myrtaceae). There is even a largelyepi- a marginalexistence should be more susceptible phytic south temperatefamily shared by New 1987] GENTRY & DODSON-NEOTROPICAL VASCULAR EPIPHYTES 219

Zealand and SouthernArgentina-Chile-Grise- Rey, L. Malmierca, pers. comm.) as compared liniaceae (sometimesincluded in Cornaceae). In with250N (46 epiphytesconstituting ca. 2.8% of the north the only noteworthytemperate epi- the South Florida flora,Long & Lakela, 1971). phytesare in the Himalayas whereaberrant epi- At leastin SouthAmerica, the prevalence of many phyticspecies of otherwiseterrestrial genera like speciesof the genus Tillandsia in southernforests Rex, Tripogon,Euonymous, Sedum, and Tha- (e.g., 14 at Parque El Rey), perhaps due purely lictrumoccur. This latitudinalasymmetry was to biohistorical reasons, is another important already noted by Schimper(1903), who pointed factor.A similarpattern occurs with lianas, which out that north temperateepiphytes are merely are betterrepresented in New Zealand than in range extensionsof widespread tropicalspecies, northtemperate forests (Dawson, 1980). whereasmany unusual and distinctiveepiphytic taxa occur in the South Temperate region. CONTINENTAL TRENDS At the communitylevel the same trendis apparent.For example, Parque Nacional El Rey in There are several conspicuous differencesbe- Argentina,at 24045'S latitude,has a species list tweenthe epiphytic floras of differentcontinents. (L. Malmierca, pers. comm.) of well over 500 Obviously, predominantlyextratropical conti- vascular plantsincluding 47 species ofepiphytes: nents have few vascular epiphytes. However, 20 ferns,four orchids, three species of Rhipsalis thereare also strikingdifferences within the trop- and Peperomia,and no fewerthan 17 bromeliads ics betweenthe Neotropics,tropical Africa, and including14 tillandsias.In contrast,Florida Cav- tropicalAustralasia. The Africanepiphytic flora erns State Park at 30'50'N latitudehas only two has been widely noted to be very impoverished epiphytesin its similar-sizedflora of 485 native compared with the other two regions,presum- species (Mitchell,1963). Even in ratherdry south ably reflectinga loss of mesic-adapted species temperatevegetations vascular epiphytescan be duringthe dryperiods associated withthe Pleis- extremelyprevalent, a situationapparently with- toceneglacial advances at higherlatitudes (Rich- out parallel in the North Temperate region.For ards, 1973; Madison, 1977). Accordingto Mad- example, in the Valley of Lerma near Salta, Ar- ison thereare only ca. 2,400 epiphyticspecies in gentina(1,200 m, ca. 700 mm ppt.), thereare at Africa,less than a sixthas many as in the Neo- least 14 angiospermepiphytes including at least tropicsand a quarteras many as in tropicalAus- ten Tillandsiaspecies in a floraof over 750 species tralasia.Even thoughseveral familiesand genera (Novara, 1984). withepiphytes in Madagascar wereomitted from Farthersouth in the Valdivian regionof Chile Madison's (1977) epiphyte summary (see Ap- vascular epiphytes,mostly belonging to endemic pendix), theirinclusion does not appreciablyin- genera,are conspicuous elementsof local floras, crease the numberof Africanepiphyte species. rangingfrom six species (3% of the native flora) Curiously,the depauperate nature of the Af- at relatively dry Parque Nacional Tolhuaca rican epiphyticflora is not obvious at the com- (38015'S) (Ramirez, 1978) to 17 species (17% of munity level. For example, the 59 epiphyte thenative flora) at verywet Fundo de San Martin species at Makokou, Gabon, constitute5% of the (39'30'S) (Cdrdenas, 1976; Riveros & Ramirez, total Makokou flora (Table 8; compiled from 1978); even at 41PS there are 15 vascular epi- Hladik & Gentry,in prep.; Florence & Hladik, phytespecies in PuyehueNational Park (Muiioz, 1980, and included references).While 5% of a 1980). New Zealand forestshave even more epi- moist forestflora might seem fewerepiphytes phytesthan the Chilean ones; even well south of than would be expected in the Neotropics, Ma- 40'S, about 30 vascular epiphytesare typically kokou is quite dry,with only 1,785 mm ofannual included on local species lists (Dawson, 1980). rainfall,and its 66 (6%) epiphyticspecies (in- However, at comparable latitudes in North cluding stranglers)are quite in line with the 72 America thereare no vascular epiphytes. (12%) epiphyticspecies at Jaunecheand 24 (4%) Whythere are more,and moredistinctive, epi- at Santa Rosa (Table 9). Johansson(1974) stud- phytes in south temperatethan in north tem- ied a relativelymoist region in theNimba moun- perate forestsis unclear but presumablyrelates tains of northernLiberia and reported 153 vas- to the relativelymesic, more or less oceanic cli- cular epiphytespecies (excludingsix filmyferns mates thatprevail in the SouthernHemisphere. and 23 "facultative"epiphytes) in his studyarea, There are more epiphytesat 250S (47 epiphytes up to 44 species in a single 750 m plot, and up constitutingca. 8.5% of the flora of Parque El to 22 species on a singletree. In even wetterareas 220 ANNALS OF THE MISSOURI BOTANICAL GARDEN [VOL. 74

TABLE 8. Habit distributionsof Makokou, Gabon plant species.

Gymno- Habit Ferns sperms Monocots Dicots Total Epiphytes 26 0 31 2 59 Parasites + saprophytes 0 0 3 6 9 Climbers 1 2 8 248 259 Trees 0 0 5 390 395 Herbs, shrubs,and treelets 42 0 125 251 418 Total species 69 2 172 897 1,140

like southwesternCameroon epiphytesare al- but thatwould hardlyexplain the lowernumbers most as prevalentas in many similarareas of the of individuals in comparable lowland forests. Neotropics (AG, pers. obs.). Janzen (1974b, 1977) emphasized that many That tropical Australasia is also floristically tropicalAsian foreststend to have relativelynu- impoverishedwith respect to the Neotropicshas trient-poorsoils, and, if so, our suggestionof a only recentlybeen realized (Raven, 1976; Gen- positive correlationbetween epiphytesand soil try,1982b). Epiphytesaccount for much of the fertilitymight help explain therelatively low epi- overall differencebetween the two regionswith phytediversity and biomass in tropical Asia. half again as many epiphytesin the Neotropics Anotherimportant continental level difference as in Australasia (15,500 vs. 10,200) according in epiphytesis taxonomic. Orchids and fernsare to Madison's (1977) figures.Moreover, at least the predominant vascular epiphytes nearly in those lowland Asian forestswe have visited, everywhere,but the other elements of the epi- therealso seem to be many fewerepiphytic in- phyticflora are oftenvery differenton different dividuals thanin comparableneotropical forests. continents(Table 10). Johansson (1974) gener- Richards's (1936) remarkis typical:"One of the alized that the Africanepiphytic flora is made most strikingfeatures of the Sarawak rain forest, up almost entirelyof pteridophytesand orchids, especially when compared with that of tropical whereas these groups are joined by bromeliads South America, is the povertyof the epiphytic and Cactaceae as importantepiphytic taxa in vegetation both in species and individuals." South America and by Asclepiadaceae and Ru- Madison suggestedthat the fewertropical Asian biaceae in SoutheastAsia. Althoughthere are 18 epiphytesmight stem simply from lack of the seed plant families with epiphyticspecies only extensivecloud foresthabitats of the Neotropics, in Australasiaand 15 withepiphytes only in the

TABLE 9. Representationof differenthabits in local florulas.

Barro Santa Jau- Colo- Rio Pa- La Capeira Rosa neche rado lenque Selval Makokou Habit No. % No. % No. % No. % No. % No. % No. % Epiphyte(incl. stranglers) 8 2 19 3 72 12 216 16 238 23 368 25 66+ 6+ Parasites + saprophytes 4 1 6 1 4 1 12 1 6 1 8 1 9 1 Climbers 112 24 115 18 136 22 258 20 171 16 182 12 259 23 Trees -10 cm DBH 69 15 142 21 112 19 290 22 165 16 310 21 389 34 Terrestrial herbs, shrubs,treelets 270 58 381 58 280 47 540 41 475 45 622 42 418 37 Total species 463 667 604 1,316 1,055 1,490 1,140

l Data fromB. Hammel (pers. comm.). 1987] GENTRY & DODSON-NEOTROPICAL VASCULAR EPIPHYTES 221

TABLE 10. Taxonomic distributionof epiphytictaxa in some lowland florulas.

Ferns Monocots Dicots Total No. No. No. No. No. No. Site Spp. % Fam. Spp. % Fam. Spp. % Spp. % Neotropics Capeira, Ecu. - - 2 7 88 2 2 22 9 2 Santa Rosa, C.R. 7 29 2 11 46 3 6 25 24 4 Jauneche,Ecu. 5 7 3 49 68 4 18 25 72 12 Barro Colorado, Pan. 43 20 4 131 61 12 42 19 216 16 Rio Palenque, Ecu. 28 12 4 142 60 14 69 29 238 23 La Selva, C.R. 59 16 5 226 61 15 83 23 368 25 Africa Makokou, Gabon 26 44 2 31 53 3 9+ 3 66+ 6+ Asia Flora of Java 200 24 3 520 63 11 109 13 829 -

Neotropics (in the case of Campanulaceae there plant familieswith epiphytes(Table 3) but epi- are also 11 Hawaiian species), Africa has only phytismis minimal in 19 of these (one to four one familyuniquely epiphytic. That family,Cos- epiphyticspecies in the Neotropics). Of the fam- taceae, is a dubious segregateof Zingiberaceae, ilies with at least five epiphyticspecies in the the latterwith epiphytesin Asia. At least nine Neotropics, seven-Bromeliaceae, Cyclantha- seed plantfamilies have epiphyticspecies in both ceae, Rapateaceae, Campanulaceae (also in Ha- the Neotropics and Australasia (but not Africa), waii), Bignoniaceae, Marcgraviaceae, Guttifer- but not a single one has epiphytesin both the ae-are epiphyticexclusively in the Neotropics, Neotropicsand Africa-Madagascarbut not Aus- and another,Cactaceae, has only one epiphytic tralasia. There are 14 families with epiphytic species widespreadin the Paleotropics.Fourteen species in all threeof the world's main tropical of the familieswith some neotropicalepiphytic regions.In total thereare 33 (or 34 if Costaceae species have epiphytesin all three tropical re- is segregated)families with epiphytes on onlyone gions-Araceae, Orchidaceae, Liliaceae, Aralia- continentas compared with 23 with epiphytes ceae, Begoniaceae, Compositae, Crassulaceae, on more than one continent:clearly most seed Gesneriaceae, Lentibulariaceae, Melastomata- plant families have epiphyteson only a single ceae, Moraceae, Myrsinaceae,Piperaceae (plus, continent. marginally,Cactaceae). In addition to exclusive- Many of the "epiphytic" familiesincluded in ly south temperateGriseliniaceae and Philesi- theabove analysisare actuallyterrestrial families aceae, seven families-Gnetaceae, Burmanni- withone or two aberrantspecies adapted to epi- aceae, Asclepiadaceae, Ericaceae, Rubiaceae, phytism.There are only 32 seed plant families Solanaceae, Urticaceae-have epiphytesonly in with five or more epiphyticspecies. If the con- the Neotropics and Australasia. tinentalrepresentations of these 32 familiesare The sharingof epiphytic taxa betweenAustral- compared, seven (Bromeliaceae, Cyclanthaceae, asia and the Neotropics but not withAfrica is a Rapateaceae, Bignoniaceae, Campanulaceae, verydifferent pattern from that normallyfound Marcgraviaceae,and Guttiferae)have epiphytes in angiosperms,where close floristicrelation- only in the Neotropics; six (Myrtaceae,Nepen- ships betweenAfrica and the Neotropics(reflect- thaceae, Pittosporaceae,Loganiaceae, Balsami- ing a sharedearly angiosperm Gondwanan flora, naceae, and Zingiberaceae, sensu stricto)have Raven & Axelrod, 1974) or betweenAfrica and epiphytesonly in Australasia; 14 have epiphytes tropicalAsia (reflectingthe relativelydirect mi- in all threetropical regions;and five (Asclepia- grationroute provided by today's geography)are daceae, Ericaceae, Rubiaceae, Solanaceae, and the generalrule. Urticaceae) have epiphytesonly in the Neotrop- Even withinthe same familypaleotropical and ics and Australasia. neotropicalepiphytes are oftennot closely relat- Put anotherway, there are 42 neotropicalseed ed. For example, the paleotropical epiphytic 222 ANNALS OF THE MISSOURI BOTANICAL GARDEN [VOL. 74

Gesneriaceae belong to subfamilyCyrtandroi- clearestcases of a shared epiphyticancestor are deae while neotropical ones are mostly in the Cactaceae where long distance dispersal of one endemic subfamily Gesnerioideae (Wiehler, species of Rhipsalis is responsiblefor the pattern 1983). Madison (1977) noted that epiphytism seen today (Barthlott,1983), and the peculiar has arisen independentlyin at least threediffer- case of Ficus wherethe large successfulpantrop- ent groupsof aroids. Most epiphyticneotropical ical subgenus (Urostigma) has specialized as orchids belong to subtribes Pleurothallidinae, stranglers.The fivepantropical epiphytic orchid Maxillarinae,and Oncidinae, while most paleo- generalikely achieved theirpresent distributions tropical ones belong to Dendrobiinae and Bul- via long-distancedispersal of their dustlike seeds. bophyllinae.The few epiphyticCentral Ameri- Predominantlyepiphytic Peperomia may have can species of Cynanchumare quite unrelatedto been originallyepiphytic; it is probably ancient the many paleotropicalepiphytic species of As- (cf. Burger,1977) and in many aspects of its bi- clepiadaceae. Most epiphyticneotropical Erica- ology,in addition to its pantropicaldistribution, ceae belong to subfamilyVaccinioideae, halfthe it is anomalous among epiphytes. paleotropical ones to Rhododendroideae. Most Schefflerais predominantlyepiphytic in the epiphyticneotropical Rubiaceae belong to tribe Neotropics,predominantly terrestrial in the Pa- Cinchoneae (subfamily Cinchonioideae), most leotropics; according to Madison's (1977) esti- paleotropical ones to Psychotrieae (subfamily mate only65 species are epiphytic.Many species Rubioideae). growboth as epiphytesand terrestrials;epiphyt- Even whenthe same genushas epiphyticspecies ism, though widespread, does not seem funda- in all threecontinental regions, these may not be mentallyintrinsic to Scheffleraand may have closely relatedto each other.Altogether only 30 arisen independentlyin all three regions and epiphyte-containingseed plant generaare found probablywithin a given regionas well. in more than one of the threemain tropical re- Ferns, notoriousfor the ease of long-distance gions, and only 14-Liparis, Malaxis, Vanilla, dispersalof their dustlike diaspores, contrast with Polystachya,Bulbophyllum, Schefflera, Begonia, the angiospermsin having most large epiphytic Rhipsalis, Vaccinium, Utricularia,Ficus, Myr- genera preponderatelypantropical. There are sine, Peperomia, Senecio -are pantropical. even exclusivelyepiphytic pantropical fern gen- Twelve of the generathat occur as epiphyteson era includingPleopeltis, Platycerium, Ctenopte- more than one continentare large,diverse, ter- ris,Xiphopteris, Psilotum, Vittaria,and (almost) restrialgenera in which epiphytismhas arisen Polypodium;there is not a singleexclusively epi- occasionally. For example, thereare at least two phyticpantropical seed plantgenus. Even in ferns independentorigins of epiphytismin Utricularia themajority of the epiphytic genera are restricted (P. Taylor, fide Madison, 1977). Genera like to one geographicalregion. Only one epiphytic Gnetum, Myrsine, Burmannia, Schefflera,Be- ferngenus is disjunctbetween the Neotropics and gonia, Senecio, Gaultheria, Vaccinium, Piper, Asia (Ophioglossum,a largegenus withonly two Psychotria,Solanum, and Pilea have indepen- epiphyticspecies separatelyarisen on the two dentlyderived and mostlyquite unrelated epi- continents).Five epiphyticfern genera occur in phyticspecies in Old and New Worlds. There both tropical Asia and Africabut not the Neo- are fiveexclusively paleotropical genera with epi- tropics,three of them on Madagascar but not phyticspecies in Africaand Asia: predominantly continentalAfrica; again thisis a patternwithout epiphyticMedinilla (Melastomataceae) and the parallel in the seed plants. orchid genera Acampe, Oberonia, and Taenio- We may conclude that, except for the ferns, phyllum, and occasionally epiphytic Embelia stranglerfigs, Peperomia, Rhipsalis, and a few (Myrsinaceae). We are leftwith only Schefflera, orchidgenera, the epiphytic floras of the different Rhipsalis, Ficus, Peperomia, Liparis, Malaxis, tropicalregions are independentlyderived, even Vanilla, Polystachya,and Bulbophyllumas gen- in most cases wherethe same familyor genus is era in which epiphytismis widely prevalenton involved in differentregions. all three continents.Indeed the only genera in While certain taxa are preadapted to an epi- which epiphytismin both the New and Old phyticlifestyle, evolution of an extensive array Worlds seems to representa truesynapomorphy of epiphyticspecies in any given taxon depends are three south temperate genera (Luzuriaga, largelyon the peculiaritiesof that region's evo- Griselinia, and Coprosma), Ficus, Rhipsalis, lutionary milieu. Epiphytism has arisen very probablyPeperomia, and possiblySchefflera. The manytimes in verymany groups. However, cer- 1987] GENTRY & DODSON-NEOTROPICAL VASCULAR EPIPHYTES 223 tain regionshave given rise to many epiphytes; lesser extentin southernCentral America. For othershave not. In this context,the partial ex- epiphytesthis concentrationof species diversity planation offeredby Madison (1977) for the could have been predictedfrom the trendsout- greaterrepresentation of epiphytesin the Neo- lined above. But whyare epiphytes(and some tropics-that historical accident in the distri- otherplants) so much more diverse in these re- bution of familieslike Bromeliaceae, Cactaceae, gions? Marcgraviaceae, and Cyclanthaceae has led to One reason that epiphytesare especially di- the neotropical predominance in epiphytic versein wetaseasonal forestsis thatthey are able species-seems largely irrelevant. Indeed, as to achieve a much finerniche partitioning,and manyfamilies (and othersupraspecific taxa) have thusa higheralpha diversity,there. Western Ec- evolved epiphytismin the Paleotropicsas in the uador provides a good example of how this phe- Neotropics; the differenceis that in the Neo- nomenon operates. In the evergreen Rio Pa- tropics evolutionaryexperiments with an epi- lenque wet forestnearly all of the epiphyteshave phyticlife-style have subsequentlyled to much a characteristicand usually very restrictedhab- more profuse speciation. Madison (1977) also itat,occurring only in the understory,the middle thoughtthat one elementin explainingthe con- story,or the canopy. Altogether41 species of tinentaldifference of epiphytediversity is a pau- vascular epiphytesat Rio Palenque are under- city of paleotropical nonorchid monocot epi- storyspecialists: 19 species of Araceae, one of phytes. However, our analysis emphasizes that Begoniaceae, one of Bignoniaceae, five of Cy- it is not theevolution of epiphytismitself in such clanthaceae, eightof Piperaceae, one of Solana- taxa that is the critical factor,but ratherthat ceae, and six ferns.However, there is not a single therehas been littlesubsequent radiation. Why understory-specialistepiphyte species in the have epiphytictaxa of Zingiberaceae,Costaceae, highlyseasonal semideciduous moist forestat Pandanaceae, or Liliaceae not evolved into pa- Jauneche,only a few tens of kilometersaway. leotropicalversions of Bromeliaceae or Cyclan- The presence of 41 species of understoryspe- thaceae? We will tryto analyze why this should cialist epiphytesat Rio Palenque accounts for be so in the next section. much of the differencebetween its diverse epiphyticflora and the relativelydepauperate one at Jauneche. Nonstranglingepiphytic trees like EVOLUTION OF EPIPHYTE SPECIES Clusia, another specialized habit not occupied DIVERSITY by Jaunecheepiphytes, account foranother five Why are epiphytesso much betterrepresented species of the difference.Presumably the more in some habitatsthan in otherswithin the Neo- constant environmentat Rio Palenque favors tropics?Part of the answer to that question can within-communitymicrohabitat specialization be adduced from the diversitygradients dis- by epiphytes. Thus classical ideas about the cussed above. From the patternsdocumented greaterspatial heterogeneityof everwettropical above we can generalizethat epiphytes are most forests(e.g., Baker, 1970) are certainlyapplicable diverse in wet, middle elevation,rich-soil, trop- to epiphytediversity patterns both within the ical American forests.With the exceptionof the tropicsand on a latitudinalgradient. somewhat tenuously established trend toward To some extentniche fine-tuningin constant greaterdiversity at middle elevations, these are environmentsalso occurs in nonepiphytes.A exactlythe trendsshown by angiospermsin gen- good example is provided by Gasteranthusat eral (Gentry,1982a, 1982b, 1987c). In a previous Centinela, Ecuador (see Gentry, 1987b). Six paper, based on data extrapolatedfrom a large species occur togethersympatrically. All are ter- array of published monographs of neotropical restrialherbs and fiveare strictlyendemic. The taxa, Gentry(1982b) concluded that plant fam- nonendemic species, G. oncogastris,occurs on ilies belongingto differenthabit groups have fun- the lower slopes away fromthe 600 m ridgetop. damentally differentdistributional patterns. Anotherspecies, G. crispus,grows only in sandy Families composed mostly of canopy trees or creek beds of the northpart of the ridge and is lianas have theirgreatest diversity in Amazonia not strictlysympatric with the otherfour species. whereas families made up mostlyof epiphytes, Of the four strictlysympatric species, one, G. shrubs,or palmetto-typeherbs are largelyextra- atratus,has switchedfrom hummingbird- to bee- Amazonian and are especiallyconcentrated along pollination and has yellow flowerscompletely the lower slopes of the northernAndes and to a distinct from the orange flowersof the other 224 ANNALS OF THE MISSOURI BOTANICAL GARDEN [VOL. 74 species.Two ofthe strictlysympatric species both group to show such patterns,but they do seem have large flowersand grow in the deep shade especiallyprone to this kind of microgeographic along creek beds, but G. macrocalyxblooms in speciation. Probably the association of accen- the wet season and G. pubescens in the dry sea- tuated microgeographywith mountainous ter- son. The finalspecies, G. carinatus,is morpho- rain,which is much more prevalentin the Neo- logicallydifferentiated by slightlysmaller flowers tropics, explains some of the intercontinental and ecologicallyby growingonly along the ridge diversitydifferences. top. In Gasteranthus,as in the Rio Palenque epi- A third, not necessarily mutually exclusive, phyticflora, very fine niche partitioningin a rel- potentialexplanation for high epiphyte diversity ativelyconstant climate seems an importantkey in the northernAndean and southernCentral to maintenanceof high species diversity.To the American cloud forestsis the "evolutionaryex- extentthat such niche partitioningis related to plosion" hypothesisadvanced by Gentry(1 982b) equable montanecloud forestconditions, it might in an attemptto explain why thereare so many be expected to be relativelyfavored in the Neo- moreneotropical than paleotropical plant species. tropics. A relativelysmall numberof genera of epiphytes, A second explanation for the great epiphyte understoryshrubs, and palmetto-typeherbs have diversityin the Andean area focuses on :-di- speciated profuselyin the northernAndean re- versityresulting from the greatermicrosite dif- gion,in each case givingrise to verymany locally ferentiationtypical of mountainous regions. A endemic and often ratherpoorly differentiated test of this hypothesismight come from com- species. Gentry(1 982b: 587) interpretedthe high paringepiphyte diversity in areas withhigh and local endemism(cf. Gentry, 1 987b) and apparent low micrositedifferentiation. The two "nudos" "species swarms" in such large evolutionarily where the eastern and westernAndean cordil- plastic genera as Anthurium,Piper, and Caven- leras come togetherbriefly at opposite ends of dishia as reflecting"shifting balance" founder Ecuador provide such a test.The southernNudo effectphenomena (Wright, 1977; Templeton, de Loja (or Sabanilla) marksthe beginning of the 1980) with major geneticreorganizations or ge- Huancabamba biogeographicdiscontinuity (cf. netic transilience(Templeton, 1980) optimized Berry,1982); the northernNudo de Pasto, most- by small and localized populations and by the ly across theborder in Colombia, marksthe point need forconstant recolonization in a habitatpar- where the threeColombian cordillerasdiverge. titionedby mountains,local rainshadows,ver- In both of these areas a tremendousvariety of ticallyshifting cyclically coalescing vegetational habitats spanningthe gauntletfrom very wet to zones, and frequentlandslides. If genetic founder verydry occur togetherin very close proximity effectsassociated withrecolonization of the open due to the unusually broken terrainand very areas resultingfrom landslides or unusuallyfre- differentorientations of adjacent slopes. TelM- quent treefalls in these geologicallyand ecolog- pogon (Orchidaceae) providesa good example of ically dynamic regions are major determinants how such micrositeheterogeneity can multiply of speciation events, then much speciation in epiphytespecies diversity.There are 37 Telipo- Andean-centeredtaxa could well be essentially gon species in Ecuador and five more on the sympatricand largelyrandom. This model thus Colombian side of the Nudo de Pasto. There are differsfrom the "microgeographicspeciation" 14 species on the Nudo de Loja and 16 species model in that speciation could take place at a on the Nudo de Pasto, as compared with only much finer"sympatric" scale, e.g., as coloniza- 14 speciesin theentire intervening 600 km. More tion of a specific landslide, with many of the instructive,nine of the Telipogonspecies on Nudo resultantspecies ecologically indistinguishable de Loja and five on Nudo de Pasto are locally fromeach other,rather than each adapted to a endemic,each foundin a singlevalley or a single specificmicrosite. Since this scenario is largely slope. In the much largerintervening area there dependent upon the extreme ecological dyna- are only nine endemic species, each restrictedto mism imputed to be associated with tropical a small area of a few hectares,in several cases mountainsit would be expected to be most ap- representingunusual and restrictedmicrohabi- plicable in the Andean region,by far the most tats similarto those around the Nudos. It seems extensivemountain systemin the world's trop- clear that in Telipogon diversityis associated ics. with local speciation and microgeographicspe- We have indirect but highlysuggestive evi- cialization. Again, epiphytes are not the only dence forspeciation associated withgenetic tran- 1987] GENTRY & DODSON-NEOTROPICAL VASCULAR EPIPHYS 225

I I

FiGupE 5. Stanhopeajenishiana Reichb. f. siliencein founderpopulations in several genera ulation near Pifias in southernEcuador. There of epiphytes.Embrya rodigasiana (Cogn.) Dod- are two othersmall geographicallyisolated pop- son providesa good example of a founderevent. ulationsof thiscomplex in westernEcuador, each This species is distributedmostly in westernCo- of them specificallydistinct but clearlyderived lombia, between 1,000 m and 1,500 m, ranging fromS. jenishiana and each likelyresulting from fromwest of Medellin to near Buenaventura. a single long-distancedispersal event-Stanho- There is also a disjunct population in a small pea frymireiDodson is endemic to a range of area of southernEcuador near Pangui, at 1,500 relativelymoist hilltops near thecoast in Manabi m above Bomboiza and Gualaquiza; priorto re- and Guayas provinces and S. embreel Dodson centwidespread habitat destruction,the species occursonly around 1,000 m in theBolivar-Caiiar was verycommon in the disjunctand geograph- borderarea (Fig. 6). icallyvery limited Ecuadorian part of its range, The herbaceous and shrubbytaxa that show whereit was surelyintroduced by a long-distance such patternshave relativelyshort generation dispersal founder event. Masdevallia chonta- times,providing conditions appropriate for rap- knsis is a similarexample. Well known and oc- id evolutionarydiversification. Under eitherthe cunfingin a well-knownhabitat, it ranges from microgeographicor founder effecthypothesis, Guatemalato Panama witha small disjunctpop- epiphytes,as the major herbaceous component ulation near Pifias in southernEcuador. There ofwet tropical forests, might be expectedto show are severalsimilar cases of unsuccessfulfounder unusually rapid "evolutionary explosion" type eventsby dust-seeded tropical epiphytes that have speciation. Relatively specific pollination sys- temporarilyestablished disjunct populations in tems constitute a second factor suggested by Florida, been duly recorded as members of the Gentry(1 982b) as potentiallypromoting unusu- Florida flora, and then subsequently disap- ally rapid speciationin "Andean-centered"taxa, peared. An example is Leochilus labiatus which with shiftsin specificpollinators accompanied grewfor a while near Fackahatchee. by coevolutionaryfine-tuning of precise plant- The Stanhopeajenishiana complex,well char- pollinator systems apparentlya common evo- acterizedby a suite of distinctivemorphological lutionarytheme. Again epiphytes,often char- characters,provides an example of the next step acterizedby highpollinator specificity, should be in such a process. Stanhopea jenishiana Reichb. prime candidates forrapid speciation. f. (Fig 5) ranges from Cali to Popayan in the The Stanhopeajenishiana complex (see above) Cauca Valley of Colombia, with a disjunctpop- is a good example of how founderevents and 226 ANNALSOF THE MISSOURI BOTANICALGARDEN [Vot. 74

FIGuRE 6. Two localized derivatives of Stanhopea jenishiana. - A S. embreei Dodson.-B. S. frvmirei Dodson. shiftsin pollination syndromescan combine to how rapid speciation mightbe in such taxa. Sce- give rise to new species. The presumedancestral lochilus is a genus of 34 species found mostly species. S. jenishiana, has orange flowersand is epiphyticin guava trees:sparsely represented in pollinated by bees of the genus Euglossa while naturalvegetations, it was apparentlyideally pre- both of the locally endemic derivativespecies in adapted for the special conditions provided by centralEcuador have white(E. embreei)or straw- guava plantations.Fifteen species of Scelochilus colored (E. firmirei) flowersand have switched occur in Ecuador. In 1957 one of us (CD) made to pollinationby Eulaema bomboides.All three an intensivestudy of populations of Scelochilus specieshave methylcinnamate as themajor scent in an extensiveguava grove near the edge of wet attractantbut each has a differentset of scent- forestat 1,000 m altitude at km. 94 of the old modifyingcompounds. Eulaema bomboides, a Guayaquil-Cuenca road in Ecuador. Two un- veryeffective pollinator that specializes on flow- described species of Scelochilus were present, ers producingmethyl cinnamate. is endemic to growingintermixed, both verycommon, and to- centralEcuador. We interpretthis situation as getheraveraging about 30 floweringplants per reflectingthree differentinstances of long-dis- host tree. Vegetativelythe two species were in- tance dispersal. In the southernEcuador Pifias distinguishable,but theirflowers, though some- population, outside the range of E. bomboides, what variable in each species, were verydistinc- pollinationby Euglossa was maintainedand no tive withno overlap whatsoeverbetween the two speciation occurred.In the two isolated central species. Indeed the original study, intended to Ecuadorian populations speciation to take ad- focus on hybridintrogression, had to be aban- vantage of Eulaema bomboides as a pollinator doned because the two taxa proved to be so con- occurredwith essentially the same selectionop- sistentlydifferentiated. These species were de- eratingin these two differentfounder popula- scribed as S. frymireiand S. embreei (Fig. 7). tions. Differentevolutionary solutions reflected Fifteenyears later, in 1982, a returnvisit was by the two distinctderivative species resulted. made to thesame guava grove.In theintervening Scelochilus,another epiphytic genus of orchid, yearsmost of the nearbyforest had been cut and provides an even more intriguingindication of convertedto pasture.As a resultthe habitatwas 1987] GENTRY & DODSON-NEOTROPICAL VASCULAR EPIPHYTES 227 verydifferent, with a much dryeraspect and the lists the 47 largestgenera of vascular epiphytes remainingguava trees ratherold and decrepit. worldwide (those with 90 or more epiphytic Scelochiluswas much rarerbut ca. 50 plantswere species) and provides the data forsuch an anal- located in the remainingguavas. Incredibly,nei- ysis. Again orchids are preeminent.Half (22) of ther S. embreei nor S. frymireiwas presentin the 47 largestepiphyte genera are orchids.While 1982, but rathertwo differentnew species were orchids mightappear to be exceptional in their found, later described as S. gentryiand S. ro- unusual geneticplasticity and highlyspecific pol- mansii. As in 1957, both of these species, veg- lination systems,they must unavoidably be re- etativelyindistinguishable, were clearly differ- gardedas the most successfulpractitioners of the entiatedfrom each otherby floralcharacters (Fig. art of being epiphytic.Biogeographical analysis 7). Both ofthe species presentin 1982 are closely of the large orchid genera is instructive.Three relatedto S. frymirei.We suspect that theymay are pantropical,nine exclusivelyneotropical, and representin situ speciationevents, at least in the ten exclusivelypaleotropical. Even though the case ofS. romansii(S. gentryihas also been found number of large neotropical and paleotropical in several otherlocalities in westernEcuador). If orchid genera is the same, there is a dramatic so, naturalspeciation in Scelochilus can occur in differencein thenumber of species thatthey con- as littleas 15 years! tain. The nine neotropicalgenera are farlarger, An obvious alternativeinterpretation is that accountingfor 5,240 species (average 582 spp. specificlimits in Scelochilusare too finelydrawn, per genus) vs. 2,626 species (average 263) forthe withS. gentryiand S. romansilrepresenting part ten paleotropicalones. of the intraspecificvariability within polytypic Ferns account foreight of the largestepiphyte S. frymirei.However, this possibilityseems ob- genera. Not surprisingly,in view of their dia- viated by the factthat the two co-occurringSce- spore vagility,nearly all of the large epiphytic lochilusspecies of 1982 pass the testof sympatry ferngenera are pantropical,showing little evi- as biologically differentiatedpopulations, even dence of unusuallyrapid differentiationor spe- thoughthey are more similarto each otherthan ciation in the Neotropics. Only one ferngenus, to theirputative ancestorS. frymirei.We favor Pyrrhosia,is restrictedto the Paleotropics,none the interpretationthat the kind of genetic reor- to the Neotropics. ganizationthat reflects speciation in Scelochilus, The remaining15 of the largestepiphytic gen- and presumablyother orchids and even some era are split evenly between dicots and mono- nonorchid epiphytes,is so labile that it can be cots. The two largest dicot genera, Peperomia effectedin incrediblyshort times. If this inter- and Ficus, are pantropical,although the former, pretationis correct,then it is no wonder that at least, is betterrepresented in the Neotropics; many epiphytictaxa have undergonewhat ap- the other largestepiphytic dicot genera are two pears to be trulyexplosive speciationin theNeo- neotropical gesneriads (Drymonia and Colum- tropics.Obviously, it is also possible thatS. gen- nea, sensu lato), two ericads (Rhododendronand tryiand S. romansii immigratedto the guava Cavendishia),one each epiphyticin Old and New grove in question sometime between 1957 and Worlds, and two melastome genera (Medinilla 1982 with S. frymireiand S. embryicoinciden- and Blakea), restrictedrespectively to the Old tally becoming locally extinctduring the same and New Worlds (and extremelysimilar to each time interval.But the mere factthat natural spe- otheras wellas to anotherlarge neotropical genus ciation in fifteenyears seems an equally plausible Topobea). In general,there seems a reasonable explanation forthese observationsis surelysig- numericalbalance betweenOld and New World nificant. representationin the largestdicot epiphytegen- Another line of reasoning also supports the era, exceptin Gesneriaceae. The situationis very idea that certain neotropical epiphytes have differentamong the nonorchid monocot epi- undergoneexplosive evolution. It is probablysafe phytes.Six of the seven largestgenera-Anthur- to assume that,as a generalrule and despitemany ium, Tillandsia, Philodendron, Vriesia, Guz- potential exceptions,genera with many species mania, and Aechmea -are exclusively are those whichhave, on the average,undergone neotropical;only Rhaphidophora is paleotropi- the most rapid speciation.Thus an examination cal. Thus thevery many species that have evolved ofthe largest epiphyte genera might indicate some among certainmonocot epiphytegenera seem a of the trends,both geographic and ecological, peculiarityof the Neotropics. thatcharacterize successful epiphytism. Table 1 1 Why have epiphytesbeen especially suscep- 74 228 ANNALS OF THE MISSOURI BOTANICAL GARDEN [VOL.

L~~

2mm

2cm

2mm ~~~~~~~~~~~~2mm

SCELOCHILUS EMBREEI Dodson SCELOCHILUS FRYMIREI Dodson

3mm

1cm

5mm

lmm 3mm

5mm

SCELOCHILUS GENTRYI Dodson SCELOCHILUS RONIANSII Dodson & Gary 1987] GENTRY & DODSON-NEOTROPICAL VASCULAR EPIPHYTES 229 tibleto the rapid,even explosive,speciation that chidaceae, is preciselythat familythat has both seems to have characterizeda large element of the most specific pollination systems and the the neotropicalflora? At least in the case of or- tiniest,most r-selecteddiaspores of any angio- chids,unusually specific pollination systems have sperm. Moreover, fromthe viewpoint of "acci- clearly played a major role in making possible dental" speciation via foundereffect phenome- veryactive speciation.Like the orchids,the non- na, the unusual genetic similarityamong the orchid epiphytefamilies that have speciated the potential-colonizerorchid propagules that arrive most profuselyin thenorthern Andes and south- togetherat a specificsite-due to orchid seeds' ern Central America-Gesneriaceae, Bromeli- unique sharing of fathersas well as mothers, aceae, Ericaceae, and Araceae-all share, to a thanks to their pollinia- mightbe expected to greateror lesserextent, relatively specialized pol- accentuatethe potentialfor genetic drift in col- linatorsand specificpollination systems as com- onizing populations. Interestingly,the closest pared with the "average" tree, shrub, or free- parallelto thepostulated rapid speciationamong climbingliana (e.g., prevalenthummingbird or Andean cloud forestepiphytes is found in her- euglossine pollination). On the other hand the baceous and palmettotaxa thatmost closely ap- epiphytictaxa thathave speciatedmost profusely proximatethe unique epiphytecombination of are characterized,as a group,by more general- specific pollination systems and an r-selected ized, higherrisk dispersal strategiesthan typical high-riskdispersal mode. of othermature forest habit groups.If speciation along thelower slopes of the Andes and in south- CONCLUSION ern CentralAmerica typicallyderives fromsub- optimal genetictransilience related to multiply Althoughmany unrelated kinds of plantshave replicatedfounder events in a dynamicand kal- evolved epiphytichabits, most of theserepresent eidoscopicallychanging habitat as Gentry(1 982b) isolated species or generain otherwiseterrestrial suggested,then epiphytes, characterized by their families. Only three nonfernfamilies-Orchi- unique combination of r- and k-selectedrepro- daceae, Cyclanthaceae, Marcgraviaceae (and ductive traits,might be uniquely equipped to possiblyalso Bromeliaceae)-are predominantly react to this adaptive milieu. Like weeds, they epiphytic. Eighty percent of all vascular epi- have diaspores intrinsicallyadapted for quick phytesare concentratedin only fourfamilies- colonization of newlyavailable habitats. In epi- Orchidaceae, Bromeliaceae, Polypodiaceae, and phytes,like weeds, the primaryneed for such Araceae-and 89% in eight families. Very few adaptation is presumably a response to the familieshave been successfulat undergoingex- short-livedand unstable nature of theirnormal tensiveradiation as epiphytes.Indeed over two- substrate,a reproductivestrategy ideally pre- thirdsof all epiphytespecies belong to the single adapting epiphytes for rapid speciation in an familyOrchidaceae, and to a large degree un- environmentin which the need to recolonize derstandingepiphytic diversity is synonymous dynamicallychanging microhabits is unusually with understandingorchids. frequent. However, unlike weeds whose di- Althoughnotably few families have been very versificationis generallyconstrained by overly successfulas epiphytes,these few taxa have made generalized pollination syndromes (frequently a tremendouscontribution to thediversity of the even withloss of sexual recombination),the rich world's flora.At least 10% of all vascular plant array of specialized pollination systems that species are epiphytes and in some places epi- characterizesthe successfulepiphytic taxa would phytesmay constitutea thirdof all plant species seem to preadapt them forrapid, pollinator-re- in a local floraor 63% of the individual plants lated evolutionarydiversification. In this con- in a given sample area. text,it is not likely to be an accident that the Epiphyteshave speciatedmost profusely in the preeminentlysuccessful epiphytic family, Or- Neotropics,especially in northwestSouth Amer-

FIGURE 7. Postulated rapid speciation in Scelochilus. Top two species were described as new from 1957 collectionsfrom guava grove at km. 94 of Guayaquil-Cuenca road. In 1982 theyhad been replaced by bottom two species, both new and both closely related to S. frymirei.(Illustrations from Orchids of Ecuador, Icones PlantarumTropicarum.) 230 ANNALS OF THE MISSOURI BOTANICAL GARDEN [VOL. 74

TABLE 1 1. Major epiphytegenera (in part afterMadison, 1977; Dressler, 1981; Kress, 1986).

Genus No. EpiphyticSpp. Total No. Spp. Distribution Pleurothallis 1,520 1,600 Neotropical Bulbophyllum 1,000 1,000 Pantropical Dendrobium 900 900 Australasian Epidendrum 720 800-t Neotropical Peperomia 700 1,000 Pantropical Anthurium' 600 850 Neotropical Lepanthes 600 600 Neotropical Stelis 540 600 Neotropical Ficus 500 800 Pantropical Maxillaria 570 600 Neotropical Eria 500 550 Australasian Oncidium 475 500 Neotropical Asplenium 400 650 Pantropical Tillandsia 400 450 Neotropical Grammitis 400 400 Pantropical Philodendron' 350 475 Neotropical Masdevallia 400 400 Neotropical Medinilla 300i+ 400-F Paleotropical Liparis 300 350 Cosmopolitan Oberonia 300 300 Paleotropical Odontoglossum 285 300 Neotropical Columnea (sensu lato) 262 265 Neotropical Elaphoglossum 250 500 Pantropical Hymenophyllum 250 300 Pantropical Angraecum 206 206 Trop. Africa Lycopodium 200 400 Cosmopolitan Polystachya 200 210 Pantropical Ctenopteris 200 200 Pantropical Vriesia 200 260 Neotropical Phreatia 190 190 Australasian Trichomanes 150 300 Pantropical Polypodium 140 150 Pantropical Encyclia 130 130 Neotropical Octomeria 130 130 Neotropical Taeniophyllum 120 120 Paleotropical, esp. Afr. Guzmania 120 140 Neotropical Dendrochilum 120 120 Australasia Aechmea 120 150 Neotropical Rhododendron 112 850 Mostly Asian Coelogyne 100 100 Trop. Asia Drymonia 100 110 Neotropical Pyrrosia 100 100 Paleotropical Rhaphidophora 100 100 Paleotropical Appendicula 100 100 Australasia Blakea2 98 100 Neotropical Thrixspermum 90 100 Australasia Cavendishia3 90 100 Neotropical Croat, pers. comm. 2 Renner, 1986 and pers. comm. I Luteyn,pers. comm. 1987] GENTRY & DODSON-NEOTROPICAL VASCULAR EPIPHYTES 231 ica and southernCentral America. The explosive Gabon (Sixieme liste).Adansonia, ser. 2., 20: 235- speciation of relativelyfew epiphytegenera in 253. GENTRY, A. H. 1981. Inventariofloristico de Ama- thisregion has been responsibleto a largeextent zonia Peruana: estado y perspectivasde conser- forthe excess floristicdiversity of theNeotropics vaci6n. Pp. 36-44 in T. GutierrezG. (editor),Se- as compared with the Paleotropics. minariosobre Proyectos de Investigaci6nEcologica para el manejo de los Recursos Naturales Reno- vables del Bosque Tropical Humedo. Direcci6n General Forestal y de Ministeriode LITERATURECITED Fauna, Agri- cultura,Lima. AIRY SHAW,H. K. 1973. A Dictionaryof the Flow- 1982a. Patternsof neotropicalplant species ering Plants and Ferns, 8th edition. Cambridge diversity.Evol. Biol. 15: 1-84. Univ. Press, Cambridge. . 1982b. Neotropical floristicdiversity: phy- ASHTON,P. S. 1976. Mixed dipterocarpforest and togeographicalconnections between Central and itsvariation with habitat in the Malayan lowlands: South America, Pleistocene climatic fluctuations, a re-evaluationat Pasoh. Malaysian Forester39: or an accident of the Andean orogeny?Ann. Mis- 56-72. souri Bot. Gard. 69: 557-593. * 1977. A contributionof rain forestresearch 1983. Dispersal and distributionin Bigno- to evolutionarytheory. Ann. Missouri Bot. Gard. niaceae. Sonderbd. Naturwiss. Ver. Hamburg 7: 64: 694-705. 187-199. * 1978. Vegetationand soil association in trop- * 1986. Species richnessand floristiccompo- ical forests.Malay. Nat. J. 30: 225-228. sition of Choc6 region plant communities. Cal- BAKER, H. G. 1970. Evolution in the tropics.Biotro- dasia 15: 71-91. pica 2: 101-111. * 1987a. An overview of neotropical phyto- BARTHLOTT,W. 1983. Biogeographyand evolution geographicpatterns with an emphasis on Ama- in neo- and paleotropicalRhipsalinae (Cactaceae). zonia. Proc. 1st Simposio do Tropico Humedo, Sonderb. Naturwiss.Ver. Hamburg 7: 241-248. Belem, Brazil (in press). BAWA, K. S., S. H. BULLOCK, D. R. PERRY, R. E. COVILLE . 1987b. Endemism in tropical versus tem- & M. H. GRAYUM. 1985. Reproductivebiology perate plant communities. In M. Soule (editor), of tropicallowland rain foresttrees. II. Pollination ConservationBiology Revisited. Sinauer Press (in systems.Amer. J. Bot. 72: 346-356. press). BERRY, P. E. 1982. The systematicsand evolution of 1987c. Changesin plantcommunity diversity Fuchsia sect. Fuchsia (Onagraceae). Ann. Mis- and floristiccomposition on environmentaland souri Bot. Gard. 69: 1-198. geographicalgradients. Ann. Missouri Bot. Gard. BURGER,W. 1977. The Piperales and the monocots: (in press). alternatehypotheses for the originof monocoty- & C. H. DODSON. 1987. Contributionof non- ledonous flowers.Bot. Rev. 43: 345-393. trees to species richness of tropical rain forest. 1985. Why are thereso many kinds of flow- Biotropica 18 (in press). ering plants in Costa Rica? Pp. 125-136 in W. & L. EMMONS. 1987. Geographical variation D'Arcy & M. Correa (editors), The Botany and in fertilityand composition of the understoryof Natural History of Panama: La Botanica e His- neotropicalforests. Biotropica 18 (in press). toriaNatural de Panama. Missouri Botanical Gar- GILMARTIN, A. J. 1973. Transandean distributionsof den, St. Louis. Bromeliaceae in Ecuador. Ecology 54: 1389-1393. CARDENAS, R. 1976. Flora y Vegetaci6n del Fundo HUSTON, M. 1979. A general hypothesisof species "San Martin" Valdivia, Chile. Tesis Lic. Fac. diversity.Am. Nat. 113: 81- 101. Ciencias, Univ. Australde Chile, Valdivia. 1980. Soil nutrientsand treespecies richness CROAT, T. B. 1978. Flora of Barro Colorado Island. in Costa Rican forests.J. Biogeogr.7: 147-157. StanfordUniv. Press, Stanford,California. JANZEN, D. 1973. Sweep samples of tropical foliage DAWSON,J. W. 1980. Middle-latituderainforests in insects: effectsof seasons, vegetation types, ele- the southernhemisphere. Biotropica 12: 159-160. vation, time of day, and insularity.Ecology 54: DODSON, C. H. 1967. Relationships between polli- 687-708. natorsand orchidflowers. Atas do Simposia sobre . 1974a. Epiphytic myrmecophytesin Sara- a Biota Amazonica 5: 1-72. wak: mutualismthrough the feedingof plants by & A. H. GENTRY. 1978. Flora of the Rio Pa- ants. Biotropica 6: 237-259. lenque Science Center. Selbyana 4: 1-628. 1974b. Tropical blackwaterrivers, animals, & . 1987. Flora of Capeira and the and mast fruitingby the Dipterocarpaceae. Bio- Region of Guayaquil. Banco Nacional de Ecuador tropica 6: 69-103. (in press). 1977. Promisingdirections of studyin trop- I~ & F. M. VALVERDE. 1985. Pp. 1-512 ical animal-plantinteractions. Ann. Missouri Bot. in Flora of Jauneche,Los Rios, Ecuador. Banco Gard. 64: 706-736. Nacional de Ecuador. & R. LIESNER. 1980. Annotated checklistof DRESSLER, R. L. 1981. The Orchids,Natural History lowland Guanacaste Province,Costa Rica, exclu- and Classification. Harvard Univ. Press, Cam- sive of grasses and nonvascular cryptogams.Bre- bridge,Massachusetts. nesia 18: 15-90. FLORENCE, J. & A. HLADIK. 1980. Catalogue des pha- , M. ATAROFF, M. FARINAS, S. REYES, N. RINCON, n6rogames et des pteridophytesdu Nord-est du A. SOLER, P. SORIANO & M. VERA. 1976. Changes 232 ANNALS OF THE MISSOURI BOTANICAL GARDEN [VOL. 74

in the arthropodcommunity along an elevational the rain forestof Mt. Dulit, Sarawak. J. Ecol. 24: transectin the Venezuelan Andes. Biotropica 8: Part 1: 1-37; Part II: 340-360. 193-203. * 1957. The Tropical Rain Forest. Cambridge JOHANSSON,D. 1974. Ecology of vascular epiphytes Univ. Press (reprint). in West Africanrain forest.Acta Phytogeogr.Sue- * 1973. Africa,the "Odd Man Out." Pp. 21- cica 59: 1-129. 26 in B. Meggars et al. (editors),Tropical Forest KRESS, W. J. 1986. The systematicdistribution of Ecosystemsin Africaand South America: A Com- vascular epiphytes:an update. Selbyana 9: 2-22. parative Review. SmithsonianInst. Press, Wash- LONG, R. W. & 0. LAKELA. 1971. A Flora of Tropical ington,D.C. Florida. Univ. Miami Press, Coral Gables, Flor- RIVEROS,M. & C. RAMiREZ. 1978. Fitocenosis epi- ida. fitasde la asociaci6n Lapagerio-Aextoxiconetum LUMER, C. 1980. Rodent pollinationof Blakea (Me- en el fundoSan Martin(Valdivia-Chile). Act. Cient. lastomataceae) in a Costa Rican cloud forest.Brit- Venezolana 29: 163-169. tonia 32: 512-517. ROCKWOOD,L. L. 1985. Seed size and plant habit in . 1983. Blakea (San Miguel). Pp. 194-195 in seven familiesfrom Panama and Costa Rica. Pp. D. Janzen(editor), Costa Rican Natural History. 197-205 in W. D'Arcy & M. Correa (editors),The Univ. Chicago Press, Chicago. Botanyand Natural Historyof Panama. Missouri MADISON,M. 1977. Vascular epiphytes:their system- Botanical Garden, St. Louis. atic occurrenceand salient features.Selbyana 2: SCHIMPER, A. F. W. 1888. Die epiphytischeVegeta- 1-13. tion Amerikas. Bot. Mitt. Trop. 2. MITCHELL, R. S. 1963. Phytogeographyand floristic 1903. Plant-Geographyupon a Physiological surveyof a relic area in the Marianna Lowlands, Basis. Clarendon Press, Oxford. [Englishtransla- Florida. Am. Midl. Nat. 69: 328-366. tion, revised edition.] MuNoz S., M. 1980. Flora del Parque Nacional Puye- SCOTT, N. J., JR. 1976. The abundance and diversity hue. Editorial Universitaria,Santiago. of the herpetofaunaof tropical forestlitter. Bio- NADKARNI, N. M. 1984. Epiphytebiomass and nu- tropica 8: 41-58. trientcapital of a neotropicalelfin forest. Biotro- SOTA,E. DE LA. 1972. Las pterid6fitasy el epifitismo pica 16: 249-256. en el Departamento del Choc6 (Colombia). Ann. NOVARA, L. 1984. Las utilidades de los generos de Soc. Cient. Arg. 194: 245-278. antofitasdel nordestedel Valle de Lerma (Salta, STILES, G. 1981. Geographicalaspects of bird-flower Republica Argentina).Facultad de Ciencias Na- coevolution, with special referenceto Central turales,Universidad Nacional de Salta, Salta. America. Ann. Missouri Bot. Gard. 68: 323-351. RAMiREZ, C. 1978. Estudio floristicoy vegetacional TEMPLETON, A. R. 1980. The theoryof speciationvia del Parque Nacional Tolhuaca (Malleco-Chile). the founderprinciple. Genetics 94: 1011-1038. Publ. Occ. Mus. Nac. Hist. Nat. Santiago de Chile VAN DER PIJL, L. & C. DODSON. 1966. OrchidFlowers: 24: 1-23. Their Pollination and Evolution. Univ. Miami RAVEN, P. H. 1976. Ethics and attitudes.Pp. 155- Press, Coral Gables, Florida. 179 in J. Simmons et al. (editors), Conservation WALTER, H. 1985. Vegetationof the Earth, 3rd edi- of Threatened Plants. Plenum, New York, Lon- tion. Springer-Verlag,Berlin. don. WHITMORE, T. C. 1984. Tropical Rain Forestsof the & D. AXELROD. 1974. Angiospermbiogeog- Far East, 2nd edition. Clarendon Press, Oxford. raphyand past continentalmovements. Ann. Mis- WIEHLER, H. 1983. A synopsisof the neotropical Ges- souri Bot. Gard. 61: 539-673. neriaceae. Selbyana 6: 1-219. RENNER, S. 1984. Phaenologie, Bluetenbiologieund WRIGHT, S. 1977. Evolutionand theGenetics of Pop- Rekombinationssysteme einiger zentralamazo- ulations. Volume 3. Experimental Results and nischer Melastomataceen. Ph.D. Dissertation. Evolutionary Deductions. Univ. Chicago Press, Universityof Hamburg,West Germany. Chicago. . 1986. The neotropical epiphytic Melasto- WUNDERLIN, R. 1982. Guide to the Vascular Plants mataceae: phytogeographicpatterns, fruit types, of Central Florida. Univ. Florida Press, Gaines- and floralbiology. Selbyana 9: 104-111. ville. RICHARDS, P. W. 1936. Ecological observations on 1987] GENTRY & DODSON-NEOTROPICAL VASCULAR EPIPHYTES 233

APPENDIX I. Changes in data base fromepiphyte list in Madison (1977).

Additions Lycopodiaceae Crassulaceae Lycopodium(200/450) Kalanchoe epiphyticin Madagascar, pers. obs. Araceae (1,200 species, not 850) Marcgraviaceae (89 epiphyticspecies, not 94) Anthurium(600/850) Marcgraviastrum(10/ 15) Philodendron(350/475) Marcgravia (ca. 50-55/55) Cyclanthaceae(125 species, not 31) Norantea (1/2) Dicranopygium(few/44) Ruyschia (7/7) Evodianthus(1/ 1) Sarcopera (4/10) Stelestylis(4/4) Schwartzia (8/14) Dioscoreaceae Souroubea (9/19) Dioscorea (1/600) (Ecuador, pers. obs.) Melastomataceae (567 epiphyticspecies, not 483) Liliaceae The paleotropical genera Baekeria, Dalenia, Dic- Rhodocodon (1/8) (Madagascar, pers. obs.) cochaeta, Neodissochaeta, Omphalopus, and Philesiaceae Plechiandraeach has at least one hemiepiphytic Philesia (1/1) (Chile, pers. obs.) species fideRenner, 1986. Luzuriaga (3/3) (Chile, pers. obs.) Myricaceae Alzateaceae Myrica (1/35) (SE Asia, fideKress, pers. comm.) Alzatea (1/2) Sapotaceae Anacardiaceae Bumelia (1 /60)(Costa Rica, fideKress, pers. comm.) Spondias (1/10) (SE Asia, fideKress) Saxifragaceae Apocynaceae Hydrangea (2/80) (epiphyticin Neotropics) Mandevilla (1/114) (M. pittieri,Costa Rica) Solanaceae Begoniaceae Lycianthes(also epiphyticin Neotropics) Begonia (40/1,000) Deletions Bignoniaceae Apostasiaceae Orchidaceae Gibsoniothamnus( 11/11) Dulongiaceae = Saxifragaceae Schlegelia (18/18) Bromeliaceae-Ananas Bombacaceae Spirotheca(4/4) Cyclanthaceae- Carludovica Burseraceae Balsaminaceae-Impatiens not epiphytic in Neo- Bursera (1/80) (B. standleyana,Costa Rica) tropics Compositae (ca. 30 epiphyticspecies, not 3) Bignoniaceae-Radermachera Liabum (Sinclairia) (2/90) Guttiferae-Clusia only occurs in Neotropics Mikania (1/300) Marcgraviaceae- Caracasia Neomirandea (20/24) Araliaceae-Sciadophyllum = Schefflera Nelsoniothamnus(1/1) Note: Additional occasional or sporadic epiphytesin- Pseudogynoxys(1/21) cluded by Kress (e.g.,Myrrhidendron donnellsmithii, Senecio (Pentacalia) (5/1,500) (epiphyticin Neo- Cyperus,Pseudoeverardia (= Everardia), Arenaria, tropics,Madagascar, N.Z.) Stellaria, Maranta, Pourouma) are also excluded. Tuberostylis(2/2) 本文献由“学霸图书馆-文献云下载”收集自网络，仅供学习交流使用。

学霸图书馆（www.xuebalib.com）是一个“整合众多图书馆数据库资源，

提供一站式文献检索和下载服务”的24 小时在线不限IP 图书馆。图书馆致力于便利、促进学习与科研，提供最强文献下载服务。

图书馆导航：

图书馆首页文献云下载图书馆入口外文数据库大全疑难文献辅助工具