ECOTROPICA 17: 103–112, 2011 © Society for Tropical Ecology SAMPLING VASCULAR EPIPHYTE DIVERSITY – SPECIES RICHNESS AND COMMUNITY STRUCTURE Gerhard Zotz1,2 & Maaike Y. Bader1 1 University of Oldenburg, Department of Biology and Environmental Sciences, Functional Ecology of Plants, P.O. Box 2503, D-26111 Oldenburg, Germany 2 Smithsonian Tropical Research Institute, Balboa, Panama Abstract. The minimum requirement for the documentation of the structure of vascular epiphyte communities comprises information on the number of co-occurring species and their abundance and distribution. Using subsampling from a comprehensive census of the epiphytes on more than 1000 trees with a diameter at breast height > 1cm in 0.4 ha of lowland rainforest in Panama as a reference, this study investigates the minimum sampling effort necessary to obtain reli- able information on essential community attributes (species numbers, α-diversity, and evenness). Large trees host the majority of all vascular epiphytes and therefore only the largest trees were sampled. Comparing the resulting community attributes deduced from a sample of epiphytes on 1-12 of the largest trees with those of the entire 0.4-ha study plot indi- cates that a sample size of about 6-8 trees is already sufficient to obtain a representative description of this epiphyte com- munity. Accepted 6 March 2011. Keywords: community structure, evenness, San Lorenzo crane site, patchiness, species coexistence, species-accumulation curves, species richness estimators, substrate preference. INTRODUCTION lems are arguably the major reason for the still lim- ited knowledge of epiphyte community structure and Vascular epiphytes are a strikingly species-rich com- dynamics, but also of regional or biogeographic pat- ponent of many tropical forests, and their occurrence terns (Küper et al. 2004). The current paucity of is associated with an important role in the ecosystem, information still hampers our ability to investigate e.g. in relation to water and nutrient cycles (Benzing possible large-scale changes in species richness, an 1990, Hietz et al. 2002). However, ecological work urgent need in times of dramatic changes in global with epiphytes presents a number of important climate (Benzing 1998). Clearly pure species lists, problems. For example, their distribution is three- which are available for a larger number of sites in the dimensional and access is difficult in all but small- tropics, are of limited value, the study of communi- statured forests (e.g. Zimmerman & Olmsted 1992, ties rather requires an understanding of the abun- Moffett & Lowman 1995, Zotz et al. 1999). Also, dance, distribution, and the number of species epiphytic species are frequently small and rather in- present (Chazdon et al. 1998, Leitner & Turner conspicuous, thus observations from the ground are 2001). bound to be incomplete (Flores-Palacios & García- Franco 2001, Burns & Dawson 2005). The sampling There are quite a few papers dealing with the of epiphyte communities is thus a difficult logistical appropriate methodology to document epiphyte di- task, making the question of the necessary sampling versity and community structure that have been effort an important topic. However, this question is published over the last decades (Hazen 1966, Johans- still unresolved, which is probably not surprising in son 1978, Gradstein et al. 1996, Shaw & Bergstrom view of additional complicating factors, e.g. their 1997, Nieder & Zotz 1998, Flores-Palacios & Gar- highly patchy distribution patterns and the generally cía-Franco 2001, Gradstein et al. 2003, Wolf et al. large proportion of rare species (Nieder et al. 2000, 2009). Those studies including recommendations on Jacquemyn et al. 2005, Zotz 2007). Sampling prob- sample sizes (e.g. number and sizes of trees) invari- ably use data from montane forests (Flores-Palacios & García-Franco 2001, Gradstein et al. 2003, Wolf e-mail: [email protected] et al. 2009). In these papers, minimum sampling 103 Umbruch Ecotropica 17_1.indd 103 07.06.11 14:20 ZOTZ & BADER sizes are generally based on species-accumulation search yielded an additional 6 species of holoepi- curves and are often concluded to coincide with phytes. All of these were orchids, of which only one (Gradstein et al. 2003) or even exceed (Flores-Pala- or very few individuals were found in each case. For cios & García-Franco 2001) the maximum sample the purpose of the present study, 103 and 110 spe- sizes actually studied. However, the estimated species cies are treated as the “true” number of epiphyte richness or more complex community structure species in respectively 0.4 and 0.9 ha of lowland parameters can, in such cases, not really be verified, forest at San Lorenzo. Trees were mostly identified as true values for full communities are usually not to species, and the diameter at breast height (dbh) available (but see Annaselvam & Parthasarathy 2001 of each individual tree was known (R. Condit, STRI, for species numbers). To overcome this problem, the unpubl. data). present study took advantage of a complete inven- Data analysis. Data analysis was conducted with Es- tory of vascular epiphytes in 0.4 ha of a species-rich timateS (Version 7.5, Colwell 2005) and R (Version lowland forest in Panama, yielding more than 13 000 2.5.1, R Development Core Team 2007). The fol- epiphytes (Zotz 2007). This unique database was lowing rationale guided the “virtual sampling” of used as a reference to deduce the minimum sampling trees from the database. Large trees usually harbor by effort needed to allow a representative description of far the majority of epiphytes (Zotz & Vollrath 2003, an epiphyte community in terms of species diversity Flores-Palacios & García-Franco 2006, Wolf et al. and community structure (Colwell & Coddington 2009) and this was also the case in San Lorenzo (Zotz 1994, Leitner & Turner 2001). 2007). We therefore “sampled” the largest trees, as The present study addresses two major questions: also recommended by Gradstein et al. (2003), start- 1) how many trees have to be searched for epiphytes ing with the one with the greatest dbh. As one would to obtain a reliable estimate of the local species pool, and should do in field sampling, we excluded two and 2) how does the community structure deduced individuals that hosted an extraordinarily small num- from these samples differ from the entire inventory? ber of epiphytes: 3 and 14 individuals (3 species The first question was addressed by using species- each), respectively. The 12 selected trees, which be- accumulation curves and species richness estimators, longed to 11 different species (Table 1), hosted 6638 and the second by additionally calculating indices of epiphyte individuals of 83 species, which represents α-diversity and evenness and by a comparison of > 80% of the total observed on the epiphyte-bearing species rankings. 389 trees in the 0.4-ha plot. Treating each of the 12 METHODS TABLE 1. List of trees sampled to compute species- Field site and data collection. The empirical data used accumulation curves, species richness estimators, and in this study were collected at the San Lorenzo evenness and diversity indices. Species names follow Canopy Crane site, located within the former Fort Correa A. et al. (2004). Sherman area near the Atlantic coast of the Republic of Panama (Wright et al. 2003). The average annual rainfall is ca. 3500 mm. Canopy height of this pri- Species name Tree Epiphytes mary rain forest is quite variable, a few emergents dbh Species Individuals reaching a maximum of ca. 40 m. Brosimum utile 85 45 1978 The use of a small gondola allowed access to all Apeiba membranacea 74 48 732 strata of the forest. Each tree in a roughly square Calophyllum longifolium 73 19 259 area of ca. 0.4 ha with a diameter at breast height Manilkara bidentata 66 30 1170 (dbh) ≥ 1cm was inspected for the occurrence of Tapirira guianensis 66 31 873 vascular epiphytes, and all but the smallest seedlings Humiriastrum diguense 58 22 76 were included in the census. With the exception of Poulsenia armata 55 50 960 a few taxa that could not be distinguished reliably Brosimum utile 51 20 139 in the field, all individuals were identified to species. Jacaranda copaia 51 14 71 In the 0.4 ha that were studied in detail, we found Virola surinamensis 49 37 336 a total of 13 099 individuals of holoepiphytes, which Carapa guianensis 48 5 16 belonged to 103 species. In an additional 0.5 ha that Lonchocarpus heptaphyllus 47 6 28 was also accessible by gondola, a less comprehensive 104 Umbruch Ecotropica 17_1.indd 104 07.06.11 14:20 SAMPLING EPIPHYTE DIVERSITY trees as a sampling unit, species-accumulation curves Simpson’s D in turn is defined as were obtained. Trees were randomized 50 times S 2 without replacement to obtain a stable mean curve. D = ps (2) Σs=1 A range of estimators of total species richness were evaluated to determine which would be most stable where ps = χs/Σχ and xs is the abundance of the sth at the lowest number of sampled trees. The following species. richness estimators as suggested by Chazdon et al. The popular index J, which is based on the (1998) were chosen: five non-parametric statistics Shannon-Wiener diversity index, H´, as (Abundance-based Coverage Estimator (ACE), J = H´ (3), Chao 1, Chao 2, jackknife 1, jackknife 2) and an 1n(S) extrapolation of mean species-accumulation curves depends on species richness and is only given for (Michaelis-Menten-Means; MMM). The alternative comparison with literature values. Similar evenness Michaelis Menten statistic (MMruns) and the Inci- may be observed in spite of a substantial shift in the dence-based Coverage Estimator (ICE) are not ap- ranks of individual species. A particularly strong shift propriate for small sample sizes and were therefore would be expected if there were epiphyte species that not used here (for a full description of these statistics prefer smaller trees as hosts (Johansson 1974, Krömer see Colwell & Coddington 1994 and Colwell 2005).