BIOTROPICA 43(3): 324–334 2011 10.1111/j.1744-7429.2010.00696.x

A Dated Phylogeny Complements Macroecological Analysis to Explain the Diversity Patterns in ()

Julissa Roncal1, Anne Blach-Overgaard, Finn Borchsenius, Henrik Balslev, and Jens-Christian Svenning Ecoinformatics & Biodiversity Group, Department of Biological Sciences, Aarhus University, Ny Munkegade 114, DK-8000 Aarhus C, Denmark

ABSTRACT Integrating phylogenetic data into macroecological studies of biodiversity patterns may complement the information provided by present-day spatial patterns. In the present study, we used range map data for all Geonoma (Arecaceae) to assess whether Geonoma species composition forms spatially coherent floristic clusters. We then evaluated the extent to which the spatial variation in species composition reflects present-day environmental variation vs. nonenvironmental spatial effects, as expected if the pattern reflects historical biogeography. We also examined the degree of geographic structure in the Geonoma phylogeny. Finally, we used a dated phylogeny to assess whether species richness within the floristic clusters was constrained by a specific historical biogeographic driver, namely time-for-diversification. A cluster analysis identified six spatially coherent floristic clusters, four of which were used to reveal a significant geographic phylogenetic structure. Variation partitioning analysis showed that 56 percent of the variation in species composition could be explained by spatial variables alone, consistent with historical factors having played a major role in generating the Geonoma diversity pattern. To test for a time-for-diversification effect, we correlated four different species richness measures with the diversification time of the earliest large lineage that is characteristic of each cluster. In support of this hypothesis, we found that geographic areas with higher richness contained older radiations. We conclude that current geographic diversity patterns in Geonoma reflect the present-day climate, but to a larger extent are related to nonenvironmental spatial constraints linked to colonization time, dispersal limitation, and geological history, followed by within-area evolutionary diversification.

Abstract in Spanish is available at http://www.blackwell-synergy.com/loi/btp.

Key words: Amazonia; cenozoic; historical biogeography; neotropics; palms; phylogenetic structure; time-for-diversification; tropical species richness.

EXPLANATIONS FOR THE GEOGRAPHIC PATTERNS in species diver- lineages (Ricklefs & Schluter 1993, Ricklefs 2004, Wiens & Don- sity that we observe today have, in recent decades, focused mostly oghue 2004, Svenning & Skov 2005, Hawkins et al. 2006, on environmental correlations (Gentry 1988, Ruokolainen et al. Donoghue 2008, Svenning et al. 2008). Evolutionary and other 1997, Tuomisto et al. 2003, Jones et al. 2006, Donoghue 2008, historical factors are not necessarily proposed to substitute environ- Sesnie et al. 2009). These are based on the premise that spatial mental explanations, but at least to complement them toward turnover in species composition is driven by the interaction of the a more comprehensive interpretation of biodiversity patterns, as environment with the evolved characteristics of lineages (Gould & species diversity is the product of diversification, dispersal, and Lewontin 1979, Gentry 1988, Donoghue 2008). Environmental extinction of its constituent lineages (Ricklefs 2004, Wiens & characteristics and biological interactions are assumed to mediate Donoghue 2004, Donoghue 2008). With the rapid increase in phylo- plant community composition and are usually components in genetic studies, an increasing understanding of evolutionary niche models of floristic (Pitman et al. 2001, Austin 2002) and species dynamics (Prinzing et al. 2001, Desdevises et al. 2003, Wiens & richness patterns (Hawkins et al. 2003, Currie et al. 2004, Kreft & Donoghue 2004, Warren et al. 2008), and the new development of Jetz 2007). However, alternative explanatory models have also been analytical tools to infer historical biogeography and diversification proposed. Notably, species distributions may result from demo- times (Graham 2003, Ree et al. 2005, Rutschmann 2006, Ricklefs graphic stochasticity and the cumulative effect of dispersal limita- 2007), we anticipate a growing number of studies that integrate tion through evolutionary time (Condit 1996, Bell 2001, Hubbell historical factors into explanatory biodiversity distribution models. 2001, Svenning & Skov 2005), and these distributions could gen- The idea that species richness in an area might be limited by erate richness patterns via their interactions with geographic do- the time since the area was colonized has a long history and is main geometry (Colwell & Lees 2000) or biogeographic history referred to in the literature as the ‘evolutionary time hypothesis’ (Bjorholm et al. 2006). In this scenario, dispersal limitation might (Willis 1922, Stebbins 1974) or the ‘time-for-speciation effect’ be expressed as a spatial pattern in which floristic similarity is ex- (Stephens & Wiens 2003). Evidence for such an effect along lati- pected to decrease with increasing geographic distance (Condit tudinal and elevational gradients has only been found in a few stud- et al. 2002). ies involving well-dated molecular phylogenies (Stephens & Wiens Whether the main drivers of species diversity are environmen- 2003; Wiens et al. 2006, 2007), and very few studies have analyzed tal or not, there is increasing awareness that a comprehensive it at different spatial (local and continental) scales (Stephens & understanding of this issue will require an integration of the under- Wiens 2003). Mechanisms controlling species richness have been lying evolutionary and biogeographic history of the constituent associated with small-scale species interactions within communities (local determinism) and large-scale processes acting over long time Received 8 October 2009; revision accepted 30 May 2010. spans, such as species migration and biogeographical history 1Corresponding author; e-mail: [email protected] (reviewed in Ricklefs 2004). Reconciliation of the local/ecological 324 r 2010 The Author(s) Journal compilation r 2010 by The Association for Tropical Biology and Conservation Diversity Patterns in Geonoma 325

and continental/historical perspectives has been proposed by sug- for-diversification in shaping Geonoma diversity by comparing clus- gesting that the interactions among locally coexisting individuals ter diversity with the age of the earliest large clade making up its adjust the geographical and ecological species range (Ricklefs flora. We screened for a time-for-diversification effect not only on 2004). Another approach that connects local and continental rich- cluster diversity, represented by cluster-wide species richness (g di- ness is the species-pool hypothesis, which claims that local species versity) and endemic species richness, but also on local (11 square) richness is determined by the number of species available at a larger species richness (a diversity), as predicted by the species-pool hy- scale (Zobel 1997). In support thereof, case studies across a broad pothesis. By exploring whether time-for-diversification on a cluster range of taxa and continents have indeed shown that there can be area level also constrains the diversity of local (11 square) assem- strong continental effects on local diversity beyond those driven by blages, our study goes beyond all previous studies (but see Stephens the local environment (Ricklefs et al. 2004, Freestone & Harrison & Wiens 2003). 2006, Dickson & Foster 2008, Malard et al. 2009). We therefore hypothesize that if there is a time-for-diversification effect on rich- METHODS ness within large continental areas, then this effect should also ap- pear at smaller local community scales. GEOGRAPHIC, ENVIRONMENTAL, AND FLORISTIC DATASET.—We ob- With its approximately 68 species (Henderson et al. 1995, tained species range maps for Geonoma from Henderson et al. Govaerts & Dransfield 2006), Geonoma is widespread throughout (1995), and nomenclature for the macroecological analyses also the American tropics, and its burst of diversification occurred dur- followed this work. We digitized the range maps at a 11 11 res- ing the Miocene, 11.9–19.5 million years ago (mya; Roncal et al. olution (Bjorholm et al. 2005). The study area thus consisted of 2010). This estimate is broadly contemporaneous with the diversi- 1010 11 11 grid cells where Geonoma was present, after excluding fication timing estimated for other plant lineages in tropical Amer- 53 cells of coastal and small island cells with incomplete environ- ica; e.g., large genera like Ocotea (Lauraceae), Inga (Fabaceae), mental data. We used 12 environmental factors as explanatory vari- Annona and Guateria (Annonaceae), Costus (Costaceae), Renealmia ables. Elevation (ELV), mean annual temperature (MAT), (Zingiberaceae), Chamaedorea (Arecaceae), and the tribes Isertieae temperature of the coldest month (MTCO), mean annual precip- and Cinchoneae (Rubiaceae) all radiated in the late Tertiary itation (MAP), minimum monthly precipitation (MINMPR), (Chanderbali et al. 2001; Richardson et al. 2001, 2004; Kay et al. and water balance (WATBAL) were taken from the high-resolution 2005; Sarkinen et al. 2007; Cuenca et al. 2008; Antonelli et al. climate grid CRU CL 2.0 (New et al. 2002). WATBAL was com- 2009). Hence, an improved understanding of the drivers of the puted following Ahn & Tateishi (1994). Potential evapotranspira- current diversity patterns in Geonoma would contribute toward tion (PET) and actual evapotranspiration (AET) were taken from an improved understanding of the environmental, geographic, the 300 resolution United Nations Environment Programme and evolutionary-historical factors that have generated the enor- GNV183 dataset (http://www.grid.unep.ch). Percentage sand mous plant diversity in tropical America and driven its spatial (sand), percentage CaCO3 (Ca), base saturation (BS), and pH were arrangement. obtained from FAO Digital Soil Map of the World, version 3.5, In this study, we investigated the extent to which major spatial November 1995. We computed means per grid cell for each envi- patterns in species richness and composition in Geonoma can be ronmental variable. We generated nine spatial variables from the X linked to time constraints on its colonization and diversification and Y coordinates using the terms of a cubic polynomial trend sur- across its geographic range. The focus is often on species richness, face: centered longitude (X = Xi Xmean), centered latitude (Y), but because species richness is the product of overlapping species and X 2, XY, Y 2, X 3, X 2Y, XY 2, Y 3, used to detect broad- ranges, we believe that understanding the patterns and drivers of scale spatial trends (Legendre & Legendre 1998). The floristic species composition offers a crucial insight into the determinants of dataset consisted of presence/absence data for 50 Geonoma species, species richness (e.g., Rangel & Diniz-Filho 2005). Here, we refer after eliminating Geonoma scoparia from the data set because of its to species composition as the list of species in a given area without narrow distribution in one grid cell for which environmental data considering any measure of relative abundance. Our approach had were missing (Puntarenas, Costa Rica). the following steps: (1) We assessed whether grid cells with a similar composition of Geonoma species form spatially coherent clusters. A DATED PHYLOGENY FOR TRIBE GEONOMATEAE.—DNA sequences (2) We then quantified the extent to which the compositional vari- from two low-copy nuclear regions, intron four of the phosphor- ation reflected in these clusters could be explained by variation in ibulokinase gene and intron 23 of the RNA polymerase II gene, the present-day environment or could be attributed to nonenviron- have been used to reconstruct the molecular phylogeny of the tribe mental spatial patterns, which would be expected if the geographic and its divergence times (Roncal et al. 2005, 2010). We extended origin of lineages and subsequent dispersal limitation has played a the taxonomic sampling of these studies to include Geonoma taxa significant role in shaping the spatial diversity pattern of the . from underrepresented geographical areas. Therefore, we se- (3) By comparing the floristic composition of the clusters with a quenced the same nuclear regions for 11 taxa from Central Amer- molecular phylogeny, we examined the degree of geographic phylo- ica, four species from French Guiana, one from Bolivia, and one genetic structure in Geonoma, and the extent to which the commu- collected in the Peruvian Andes (Table S1). Despite this increase in nity of a cluster reflects one or multiple colonization and speciation the sample, species endemic to Venezuela and the Pacific coast and events. (4) Finally, we assessed the potential importance of time- inter-Andean valleys of Colombia are missing in our dated 326 Roncal, Blach-Overgaard, Borchsenius, Balslev, and Svenning

phylogeny. The total number of Geonomateae samples was 54, in- with down-weighting of the rare species, to synthesize the main cluding 35 Geonoma species, which represent 51.5 percent of the patterns of variation in species composition and to evaluate the genus (Table S1). Outgroup selection, sequencing, and alignment length of the species turnover gradient. The axes of a DCA are protocols followed those in Roncal et al. (2005, 2010). To estimate scaled in units of average standard deviation (SD) of species turn- the divergence times for lineages in the tribe, we conducted a over (Kent & Coker 1992). A complete turnover in species com- Bayesian relaxed molecular clock analysis using BEAST v1.4.8 position occurs in about 4 SD, indicating unimodal responses to (Drummond & Rambaut 2007). We used the uncorrelated log- the underlying environmental gradient (McCune & Grace 2002). normal model of rate variation with the same parameter settings Species turnover was thus estimated in SD. We compared the re- and calibration points as in Roncal et al. (2010). We chose the sults of the DCA with those obtained under a Nonmetric Multidi- Bayesian implementation of BEAST because it accounts for phylo- mensional Scaling (NMS; Kruskal 1964). For the NMS, we used genetic uncertainty, it allows the definition of calibration distribu- the Srensen (Bray–Curtis) distance measure, a random starting tions, and it does not assume that substitution rates are configuration and ten runs with real data. A Monte Carlo test was autocorrelated among branches. Empirical evidence does not al- performed with 20 randomized runs. The final solution was for ways support the autocorrelation assumption (Ho 2009), and in three dimensions, based on preliminary NMS runs. We assessed the this study, we used two metrics to evaluate the appropriateness of extent to which the main compositional gradients were related to the uncorrelated model. We rejected the rate change autocorrela- environmental or spatial factors by correlating the DCA and NMS tion as evidenced by the covariance mean (r = 0.003, 95% HPD = axes scores to the environmental and spatial explanatory variables. 0.145 to 0.137), and the substitution rate heterogeneity along the We performed a constrained ordination using Canonical Cor- tree was high as judged by the coefficient of variation respondence Analysis (CCA; ter Braak 1986) to directly evaluate (mean = 0.316, 95% HPD = 0.212–0.424). Nomenclature in the the main environmental and spatial correlates of compositional dated phylogeny followed the recent taxonomic accounts of Gov- turnover in Geonoma. We partitioned the variation in species com- aerts & Dransfield (2006) and Henderson’s Geonoma home page position into four independent components: pure spatial (RPS), (http://www.nybg.org/botany/geonoma/). pure environmental (RPE), mixed spatial–environmental (RMX), and undetermined (RUN) fractions (Borcard et al. 1992). The total 2 DATA ANALYSES.—We performed a hierarchical cluster analysis us- explained variation (R = RT), the environmentally explained varia- ing the Srensen (Bray–Curtis) distance measure and group average tion (RE), and the spatially explained variation (RS) were deter- as the group linkage method to derive the main Geonoma species mined by performing the CCA analysis on the total set of variables, associations and their compositional relationships (McCune & the environmental variables alone, and the spatial variables alone,

Grace 2002), as well as to assess whether they form spatially coher- respectively. Following Borcard et al. (1992), RT, RE, and RS were ent clusters (Heikinheimo et al. 2007). We performed the analysis used to calculate the pure environmental (RPE = RT RS), pure spa- with ten clusters using the higher level grouping option (McCune tial (RPS = RT RE), mixed environmental–spatial (RMX = RS1RE & Grace 2002). In this way, it was possible to evaluate the appro- RT), and undetermined fractions (RUN =1 RT). To avoid the priate number of clusters and percentage information retained inclusion of redundant environmental variables in the CCA, we ex- (McCune & Grace 2002). The ecologically and biogeographically cluded one variable from each pair of environmental variables meaningful number of clusters was chosen based on the ten-group that were highly correlated (r 4 0.80). Thereby, ELV, MAT, dendrogram and visual evaluation in ArcGIS 9.2 (ESRI, Redlands, WATBAL, and BS were removed. California, U.S.A.). Subsequently, we performed an Indicator Spe- We quantified the geographical phylogenetic structure using cies Analysis (ISA; Dufreˆne & Legendre 1997) on the major clusters Webb’s (2000) approach for quantifying the community phyloge- in order to evaluate possible indicator species for the main species netic structure (Lavin 2006). Using PHYLOCOM 4.1 (Webb et al. associations. Exclusion of four minor clusters in the ISA eliminated 2008), we estimated the net relatedness index (NRI) and the nearest Geonoma paraguanensis from these analyses. This species was exclu- taxa index (NTI) for four geographical areas: Amazon, Andes, Cen- sively present in a cluster, which consisted of a single grid cell. The tral America, and Cerrado/Mata Atlantica. Positive values indicate ISA produces an indicator value ranging from 0 (no indication) to that species in a geographical area are likely to come from the same 100 (perfect indication) and is computed by combining a measure phylogenetic lineage (clustering), values not significantly different of the specificity and fidelity of species j in cluster k (Dufreˆne & from 0 indicate that the species in an area likely come from any Legendre 1997). A perfect indicator species of a cluster is always lineage (random), while negative values indicate that species in an present in this cluster and does not occur in other clusters (McCune area preferentially come from different lineages (overdispersion, & Grace 2002). Significance was tested through 999 Monte Carlo Lavin 2006). To evaluate the significance of the observed NRI and permutations. In addition, we compared the average species rich- NTI values, we used two null models to generate 99,999 null index ness per cell and residual richness per cell after controlling for values: (i) the phylogeny shuffle, which randomizes species names environmental factors among the six clusters by conducting a across the phylogeny, and (ii) species in each geographical area be- one-way ANOVA, including a Tukey–Kramer HSD test for com- come random draws from the phylogeny pool (Webb et al. 2008). parison of means. Finally, we explored the potential importance of time-for- We then used the unconstrained ordination technique diversification in shaping Geonoma richness patterns at cluster area Detrended Correspondence Analysis (DCA; Hill & Gauch 1980), (g) and local (a) scales by performing Pearson’s correlations. We Diversity Patterns in Geonoma 327

correlated four species richness measures for each cluster: total rich- and a broad-scale environmental wet/warm-to-cold/dry gradient ness, average species richness per cell, average residual richness per (Fig. 3). The spatial variables, X and Y, were correlated with the cell after controlling for environmental factors, and species endemic DCA first axis (r = 0.739 and 0.666, respectively, P o 0.0001) to a cluster, with the crown node age of the earliest large clade rep- and the NMS second axis (r = 0.598 and 0.649, respectively, resenting each cluster. All tests of significance are one-tailed in the P o 0.0001). The environmental variable with the strongest corre- direction of the observed effect, and P-values represent the proba- lation with the ordination scores was AET for both the DCA bility that the true value of the effect is of a sign opposite to the (r = 0.495, P o 0.0001, second axis) and the NMS (r = 0.627, observed values. All multivariate analyses were conducted in P o 0.0001, first axis). In addition, MAP, MTCO, PET, WATBAL, PC-ORD 4.41 (MjM Software, Gleneden Beach, Oregon, U.S.A.), and MAT all correlated negatively with the NMS first axis scores while the correlation analyses and one-way ANOVA were computed (all r o 0.500, P o 0.0001), while MINMPR was correlated in JMP 7.0.0 (SAS Institute, Cary, North Carolina, U.S.A). with the second axis (r = 0.454, P o 0.0001) (Fig. 3). For the DCA, PET correlated with the first axis (r = 0.437, P o 0.0001) and RESULTS MTCO correlated with the second axis (r = 0.443, P o 0.0001). The CCA-based variation partitioning confirmed this interpreta- The ten-group cluster analysis accounted for 65 percent of the spe- tion, showing that 55.7 percent of the total explained variation in cies compositional variation and produced six large, geographically species composition reflected purely a spatial variation, while delimited clusters and four minor clusters (Fig. 1). The six large 41.1 percent reflected a mixed spatial and environmental variation clusters corresponded to Central America (including 45 cells in (Table S2). northernmost South America), the Andes, the Amazon, the Cerr- The relationships and clades recovered in the phylogeny sup- ado of Central Brazil, the Mata Atlantica (the Atlantic coastal forest ported the ordination result that geography is an important deter- of Brazil), and the southern extreme of this area (Southern Mata minant of Geonoma species richness and composition patterns. The Atlantica; Fig. 1). Each cluster had 25–100 percent of its species as phylogeny showed four Geonoma clades that corresponded closely significant indicator species of the area (Table 1), and except for the to the compositional clusters. The radiation time for the earliest Cerrado and Southern Mata Atlantica, all clusters had species that clade of Amazonian species was estimated at ca 16.8 mya (Fig. 4). exclusively occurred in them, i.e., endemic species (Fig. 2). The The three other clades had more recent diversification times. A Amazon, Andes, and Central America clusters formed one large clade formed by taxa adapted to high elevations (1000–3000 m) in group, while the Cerrado, Mata Atlantica, and Southern Mata Atl- the Andes and Central American mountains radiated ca 7.7 mya; antica clusters formed another (Figs. 1 and 2). There was little spe- another clade, which included species from the Brazilian Cerrado cies overlap between the broadly adjoining Amazon and Cerrado and coastal Atlantic forest, was estimated at ca 6.4 mya; and the last clusters (Fig. 2). The most species-rich clusters were the Amazon clade, which grouped species growing exclusively in Central Amer- and the Andes, which also had the highest average species richness ica, had an age of ca 4.6 mya (Fig. 4). These four clades represented per grid cell (Table 1). The Southern Mata Atlantica cluster had the the earliest large radiations that occurred in the Amazon, Andes, smallest area (14 grid cells), the lowest total species richness, and the Mata Atlantica, and Central American cluster areas, respectively. In second lowest average species richness per cell (Table 1). addition, we estimated the age of the Cerrado at 5.2 mya based on The DCA estimate of species turnover was 12.4 SD, indicating the divergence time of Geonoma brevispatha, the only species in the a strong turnover in species composition across the study area. Cerrado with an indicator value; the other three species occurred at The DCA and NMS ordinations showed that the main variation in very low relative frequencies and abundances and received indicator species turnover was a geographic northwest-to-southeast gradient values of 0. We estimated the age of the Southern Mata Atlantica at

FIGURE 1. (A) Species richness. (B) Simplified cluster dendrogram showing the ten basal clusters. (C) Geographic location of the ten basal clusters. 328 Roncal, Blach-Overgaard, Borchsenius, Balslev, and Svenning

TABLE 1. Indicator species and three different variables of species diversity for the six major geographical clusters derived from the cluster analysis of Geonoma species composition across 1010 11 11 grid cells. Richness refers to the total number of species in either the whole cluster or per grid cell. N, number of grid cells per cluster; IV, indicator values (only species with IVZ20 are listed). Clusters marked with different letters (A–D) were significantly different.

Total Average (maximum) SD Average (maximum) SD Cluster N Indicator species (IV) richness species richness per grid cella residual species richness per grid cellb

Andes 91 Geonoma weberbaueri (85.2) 29 5.8 (13) 3.1A 1.0 (6.3) 2.6A Geonoma orbignyana (74.1) Geonoma jussieuana (63.1) Geonoma densa (41.8) Geonoma longepedunculata (25.2) (21.8) Amazon 527 Geonoma macrostachys (72.2) 24 6.1 (12) 2.4A 0.7 (5.1) 1.9A (58.9) Geonoma stricta (58.7) Geonoma leptospadix (57.2) Geonoma deversa (46.9) Geonoma baculifera (38.9) Central America 90 Geonoma interrupta (55.3) 21 3.4 (10) 2.5B 0.9 (5.4) 2.0B, C Geonoma congesta (26.8) Geonoma ferruginea (20.0) Mata Atlantica (MA) 67 Geonoma pauciflora (83.6) 7 2.5 (5) 1.3B,C 0.4 (2.2) 1.1B Geonoma pohliana (62.7) Geonoma gastoniana (27.0) Geonoma rubescens (25.4) Southern MA 14 Geonoma gamiova (61.4) 2 1.4 (2) 0.5C,D 2.9 ( 1.1) 1.2D Geonoma schottiana (48.2) Cerrado 209 Geonoma brevispatha (87.0) 4 1.1 (3) 0.4D 1.4 (0.6) 0.9C aWelch ANOVA for average species richness per grid cell (F ratio = 464.3, df = 5, P o 0.0001). bWelch ANOVA for average residual species richness per grid cell (F ratio = 99.1, df = 5, P o 0.0001). Residuals were obtained from a multiple linear regression of 2 richness vs. elevation, actual evapotranspiration, mean annual precipitation, and minimum monthly precipitation (Radj = 0.550, P o 0.001, AICc weight = 0.983 compared with all nested models). All P o 0.001, obtained by 999 permutations.

3.9 mya based on the divergence time of Geonoma schottiana, one Geonomas in these clusters and proportionally we are not missing of the two species in this cluster. Eight Amazonian species appeared more species from them than we are from the other clusters. In ad- outside the 16.8 mya clade without resolution and with younger dition, merging the Cerrado and Southern Mata Atlantica for the divergence times (o 15 mya). However, because our aim was to time-for-diversification test would be undesirable because these clus- determine the relationship between the age of the first species-rich ters are very different climatically and floristically. Diversification radiation in a cluster area and the species richness of the cluster, we ages are underestimated with a sparse taxon sampling and the degree did not consider the age of these Amazonian species in our corre- of age underestimation increases logarithmically with the proportion lations. Likewise, 15 species occurring in the Andean slopes below of undersampling (Lindler et al. 2005); thus, our age estimates would 1000 m of elevation appeared outside the high-elevation clade, be improved by a more complete taxon sampling. It is also possible widely dispersed in the phylogeny. We did not consider the age of that the missing taxa will fall outside the geographically structured these species to represent the first Andean species-rich lineage be- clades, forming new lineages and compromising our age estimates cause the majority of them occur mostly in the adjacent Amazon and time-for-diversification test. However, with the current sam- and Central American areas and may represent post-speciation dis- pling, there was a clear geographic structure in the Geonoma phylog- persal events from these areas into the Andes. eny as evidenced by positive and significant NTI and NRI values for Our dated phylogeny included 18 of 24 (75%) Amazonian spe- all four areas (Table S4). We also obtained evidence for a time-for- cies, 19 of 29 (66%) Andean species, 13 of 21 (62%) Central Amer- diversification effect in species richness. Scatter diagrams (Fig. 5) ican species, 4 of 7 (57%) Mata Atlantica species, 3 of 4 (75%) showed that cluster areas with more endemic species and with higher Cerrado species, and 1 of 2 (50%) Southern Mata Atlantica species average species richness per cell tend to have older radiation times for (Table S3). Age estimates for the latter two clusters might seem weak their earliest large lineages (strongest correlations: r = 0.76–0.8; Figs. because they are based on one species each; however, there are few 5B and D; Fig. S1). Diversity Patterns in Geonoma 329

Andes DISCUSSION

4 Central Patterns of Geonoma composition depended mostly on nonenvi- America Amazon ronmental spatial factors, as expected if they are controlled by his- 10 8 torical determinants such as geologic history and dispersal 5 6 8 limitation and only to a lesser extent on present-day broad-scale environmental variation. It has been proposed that geographic 1 position may have a greater explanatory power when data for an important environmental factor are missing from the analyses or when the length and steepness of the environmental gradient is short 1 (Ruokolainen et al. 1997, Duque et al. 2002, Jones et al. 2006, Sesnie et al. 2009). For Geonoma, we cannot reject the influence of unmea- 1 sured variables, such as physiographic heterogeneity (habitat and topography), which is an important determinant of di- 1 1 Southern versity worldwide (Kreft & Jetz 2007). The most important environ- Mata Atlantica Cerrado mental drivers of the geographical composition in Geonoma included 3 water-related variables, followed by temperature. However, soil phys- ical and chemical properties have been found to be important deter- Mata Atlantica minants of species occurrence or abundance in several palm studies (Kahn & de Granville 1992, Clark et al. 1995, Poulsen et al. 2006, FIGURE 2. Venn diagram of species associations (49) exclusive to an area or Sesnie et al. 2009). Empirical evidence for the relative importance of shared by two or more areas between six geographically distinct clusters derived environmental and nonenvironmental factors in controlling plant from the cluster analysis and indicator species analysis. Empty fields should be distributions varies depending on geographical locations, spatial scale, interpreted as absence of species unique to the area. and the plant groups investigated (Condit et al. 2002, Normand et al. 2006, Sesnie et al. 2009), making generalizations on this topic challenging. The use of a molecular phylogeny and associated diversifica- tion times in this study provided an evolutionary time perspective to the assembly of Geonoma diversity and distribution. Geological and climatic changes in tropical America throughout the Cenozoic are likely to have influenced the migration and diversification of lineages across areas. For example, the onset of radiation of the Ge- onoma clade endemic to Central America coincides with the late Miocene–Pliocene orogenic events that gave rise to the Central America archipelago before the closure of the Panamanian isthmus (Coates & Obando 1996, Coates et al. 2004). Similarly, diversifi- cation of the high-elevation species is contemporaneous with the Miocene Andean uplift (Gregory-Wodzicki 2000), and the Brazil- ian Cerrado/Mata Atlantica clade has a radiation time that coin- cides with the second South American marine incursion (Hernandez et al. 2005). The age of G. brevispatha from the Cerr- ado is also in line with estimated ages of other Cerrado clades that diversified less than 10 mya (i.e., Mimosa, Andira, Lupinus, Micro- licieae), and with the observation that the Cerrado flora constitute crown lineages of plant phylogenies (Simon et al. 2009). Therefore, the geological and climatic history of tropical America provides an explanatory framework for the geotemporal patterns of coloniza- tion and the subsequent diversification of Geonoma. The present study assessed the role played by a specific histor- FIGURE 3. Biplot of the first and second axes of the Nonmetric Multidimen- ical constraint, namely time-for-diversification, in generating the sional Scaling (NMS) on 1010 cells. The six major clusters are depicted using geographical diversity patterns of Geonoma. Our data showed that different symbols as indicated in the legend. The remaining four minor clusters time-for-diversification was a constraint on within-cluster species are lumped into one (Others). Arrows indicate the correlations of the explana- richness, in particular for endemic species. This time effect is not tory variables with the first and second NMS ordination axes. Only variables due to environmental differences, as shown by the positive correla- with a Pearson r2 4 0.200 with at least one of the axes are shown. tion of the residual species richness and age (Fig. 5C). We note that 330 Roncal, Blach-Overgaard, Borchsenius, Balslev, and Svenning

FIGURE 4. Maximum clade credibility chronogram resulting from the Bayesian dating analysis in BEAST based on PRK and RPB2 nuclear DNA regions. Numbers are node ages in million years for those nodes within the Geonomateae supported with PP 4 0.9 in the phylogenetic reconstruction of BEAST. The ruler at the bottom of the figure indicates ages in million years and the geological time scale. The six calibration points are indicated: (1) split of Howea and Laccospadix at 5.5 2.9 mya (mean SD), (2) crown age of tribe Iriarteeae at 25 1 mya (zero offset SD), (3) crown age of tribe Chamaedoreeae at 50 6 mya (mean SD), (4) crown age of tribe Cocoseae at 50 1 mya (zero offset SD), (5) crown age of subfamily Arecoideae at 64 2.2 mya (mean SD), and (6) age of the root node at 83.5 1 mya (zero offset SD). The four highlighted clades are discussed in the text. Diversity Patterns in Geonoma 331

Andes Central America Amazon ABSouthern Mata Atlantica 7 35 X Cerrado 6 30 † Mata Atlantica 25 5 20 4 15 3 10 2 each cluster

5 Average species richness per cell 1 Total richness in 0 0 0 5 10 15 20 0 5 10 15 20 Age (Mya) Age (Mya)

CD2 9 1 8 7 0 6 0 5 10 15 20 5 –1 4 –2 3 2 in each cluster

–3 Species endemic richness per cell 1 Average residual 0 –4 0 5 10 15 20 Age (Mya) Age (Mya)

FIGURE 5. Pearson correlations between four species richness variables vs. crown node age for the earliest large radiation in each of the six geographical cluster areas. (A) Total richness in the cluster area (r = 0.54, P = 0.135, 95% CI = 0.483 to 0.94). (B) Average species richness per cell (r = 0.76, P = 0.04, 95% CI = 0.131 to 0.972). (C) Average residual richness per cell after controlling for environmental variables (r = 0.67, p = 0.073, 95% CI = 0.311 to 0.96). (D) Species endemic to each cluster area (r = 0.80, p = 0.029, 95% CI = 0.042 to 0.977).

the correlation between endemic species richness and age was Nissolia, Poissonia, and Ruprechtia (Lavin et al. 2004, Pennington stronger than the overall species richness–age correlation, probably et al. 2004). Dispersal limitation of species confined to these highly because overall species richness also reflects immigration into the fragmented and small dry forest patches was hypothesized to be the cluster, while endemic species richness more purely represents cause for this structure (Lavin 2006). In contrast, wet forest genera within-cluster diversification. The time-for-diversification effect such as Clusia and Inga lack a high level of geographical structuring also appeared at the local scale (11 grid cell), supporting the spe- (Gustafsson & Bittrich 2002, Lavin 2006) presumably due to the cies-pool hypothesis in which the species richness of local assem- high dispersal rates of these species among large and continuous wet blages is constrained by the size of the larger cluster species pool forest areas. As a wet forest genus, Geonoma contradicts this gener- (Zobel 1997). Hence, the species richness in Geonoma at the 11 alization, although the inclusion of some Central American species scale is directly constrained by the time-for-diversification of the in the Amazon and high-elevation clades, and some Andean species cluster species pool. Similarly, Stephens & Wiens (2003) found in the Amazonian clade indicates that this geographic structure is that the local species richness of emydid turtles increased with the actually partial and potentially confounded by undersampling of richness of the continental area, which again strongly increased with taxa. More studies of this kind shall elucidate the predominant de- the age of the oldest emydid lineage in the area. These studies gree of geographic phylogenetic structure for tropical plant lineages exemplify how large-scale historical factors can strongly affect the according to their preferential habitat type or geographic distribu- assembly of local assemblages (Ricklefs 2004, Ricklefs et al. 2004). tion, as well as the relative contributions of migration vs. speciation Hubbell (2001), in his unified neutral theory, argues that rates in determining phylogenetic tree structure. of extinction, speciation, and dispersal influence the shape of a There is little empirical evidence for long-distance palm seed phylogeny. Thus, tree topology and diversification times can pro- dispersal among suitable adjacent areas in tropical America, and we vide an insight into vegetation dynamics over evolutionary time. only know of three studies that have documented it. The first in- The geographic phylogenetic structure found for Geonoma can be directly showed a long-distance dispersal of Raphia taedigera from explained by low dispersal rates among geographic clusters. These Africa to Central America (Urquhart 1999). The second was an dispersal rates must not surpass the within-cluster speciation rates elevational range expansion of eight Andean species likely due to to avoid endemic lineages being replaced by immigrants (Hubbell animal activities (Kessler 2000), and the third showed a maximum 2001). The geographic phylogenetic structure of Geonoma was sim- seed dispersal of 875 m in Iriartea deloidea (Sezen et al. 2005). Two ilar to that found in Coursetia as revealed by the NRI and NTI Geonoma species occur in the Antilles (Henderson et al. 1995), pre- indexes (Lavin 2006) and to other dry tropical forest genera such as sumably as a result of two independent colonization events from 332 Roncal, Blach-Overgaard, Borchsenius, Balslev, and Svenning

continental America. Given the clear importance of generally lim- phylogeny shuffle, and ii) species in each geographical area become ran- ited, but sometimes far-reaching, dispersal for the geographic struc- dom draws from the phylogeny pool. P-values are one-tailed and were turing of species distribution in the New World palms generally calculated by dividing the number of null trees or communities with and Geonoma specifically, a better understanding of the factors that index values greater than or equal to the observed by 100,000. promote or constrain propagule dispersal and range expansions FIGURE S1. We recalculated the four species richness values for could contribute significantly to our understanding of the processes a combined cluster of the three Brazilian areas (Cerrado, Mata Atl- driving diversity patterns across scales. For example, stem height antica, and Southern Mata Atlantica), and assigned it the age of correlates positively with range size in Amazonian palms, perhaps 6.4 mya. reflecting greater dispersal ability of taller species, due to greater seed-release heights and greater fecundity (Kristiansen et al. 2009). Please note: Wiley-Blackwell is not responsible for the content Our study exemplifies how the geographic structure of species or functionality of any supporting materials supplied by the au- diversity can be attributed to time- and dispersal-limited diversifi- thors. Any queries (other than missing material) should be directed cation within the New World tropics. The geological and climatic to the corresponding author for the article. processes that occurred during the Cenozoic in tropical America (especially during the last 30 million years) provide the framework that has both constrained and facilitated the dispersal and diversi- LITERATURE CITED fication of Geonoma. Future research should attempt to estimate the diversification rates for each of the lineages described here, because AHN, C.-H., AND R. TATEISHI. 1994. Development of global land surface evapo- it appears to be an important explanatory factor of richness distri- transpiration and water balance data sets. J. Jpn. Soc. Photogram. Re- bution in palms (Svenning et al. 2008) and other Andean taxa mote Sens. 33: 48–61. (Hughes & Eastwood 2006). Importantly, as predicted by the spe- ANTONELLI, A., J. A. A. NYLANDER,C.PERSSON, AND I. SANMARTIN. 2009. Trac- cies-pool hypothesis (Zobel 1997), we found that the time-for- ing the impact of the Andean uplift on Neotropical plant evolution. Proc. Natl. Acad. Sci. USA. 106: 9749–9754. diversification effect on cluster species richness also constrains the AUSTIN, M. P. 2002. Spatial prediction of species distribution: An interface be- richness of local (11) assemblages. However, it remains to be inves- tween ecological theory and statistical modelling. Ecol. Model. 157: tigated whether the time-for-diversification effect translates to even 101–118. finer scales as well, where intra-cluster historical factors may addi- BELL, G. 2001. Neutral macroecology. Science 293: 2413–2418. tionally drive patterns of species richness and composition (e.g., BJORHOLM, S., J.-C. SVENNING,W.J.BAKER,F.SKOV, AND H. BALSLEV. 2006. Historical legacies in the geographical diversity patterns of New World Normand et al. 2006). palm (Arecaceae) subfamilies. Bot. J. Lin. Soc. 151: 113–125. BJORHOLM, S., J.-C. SVENNING,F.SKOV, AND H. BALSLEV. 2005. Environmental ACKNOWLEDGMENTS and spatial controls of palm (Arecaceae) species richness across the Americas. Global Ecol. Biogeogr. 14: 423–429. This work was financed by the Danish Natural Science Research BORCARD, P., P. LEGENDRE, AND P. DRAPEAU. 1992. Partialling out the spatial component of ecological variation. Ecology 73: 1045–1055. Council through grants #272-07-0242 to JCS and #272-06-0476 CHANDERBALI, A. S., H. VAN DER WERFF, AND S. S. RENNER. 2001. Phylogeny to HB and by the Faculty of Science at Aarhus University (ABO). and historical biogeography of Lauraceae: evidence from the chloroplast We are grateful to A. Sloth for assistance in the laboratory, A. Hen- and nuclear genomes. Ann. Mo. Bot. Gard. 88: 104–134. derson for taxonomic identifications, and J.-J. de Granville, J. Mc- CLARK, D. A., D. B. CLARK,R.SANDOVAL, AND M. V. CASTRO. 1995. Edaphic Coy, C. Guerra, C. Villaorduna, and C. Girod for assistance in the and human effects on landscape-scale distributions of tropical rain forest palms. Ecology 76: 2581–2595. field. COATES, A. G., AND J. A. OBANDO. 1996. The geological evolution of the central American Isthmus. In J. B. C. Jackson, A. F. Budd, and A. G. Coates SUPPORTING INFORMATION (Eds.). Evolution and environment in tropical America. pp. 21–56. University of Chicago Press, Chicago, Illinois. Additional Supporting Information may be found in the online COATES, A. G., L. S. COLLINS,M.P.AUBRY, AND W. A. BERGGREN. 2004. The geology of the Darien, panama, and the Miocene–Pliocene collision of version of this article. the panama arc with northwestern South America. Geol. Soc. Am. Bull. 116: 1327–1344. TABLE S1. Voucher information and GenBank accession numbers COLWELL, R., AND D. LEES. 2000. The mid-domain effect: Geometric con- for the taxa sampled in the Bayesian divergence time estimation of the straints on the geography of species richness. Trends Ecol. Evol. 15: Geonomateae. 70–76. CONDIT, R. 1996. Defining and mapping vegetation types in mega-diverse trop- TABLE S2. Variation partitioning using CCA to divide species ical forests. Trends Ecol. Evol. 11: 4–5. composition for 1010 cells into its pure environmental, pure spatial, CONDIT, R., N. PITMAN,E.G.LEIGH,J.CHAVE,J.TERBORGH,R.B.FOSTER,V. mixed spatial-environmental, and undetermined fractions. P. NUNEZ,S.AGUILAR,R.VALENCIA,G.VILLA,H.C.MULLER-LANDAU, TABLE S3. Presence (1) or absence (0), and indicator values for E. LOSOS, AND S. P. HUBBELL. 2002. Beta-diversity in tropical forest each Geonoma species across the six geographic clusters. trees. Science 295: 666–669. CUENCA, A., C. B. ASMUSSEN-LANGE, AND F. BORCHSENIUS. 2008. A dated phy- TABLE S4. Net relatedness index (NRI) and nearest taxa index logeny of the palm tribe Chamaedoreeae supports Eocene dispersal be- (NTI, Webb 2000) calculated from a maximum parsimony tree of tween Africa, North and South America. Mol. Phylogenet. Evol. 46: Geonoma using Phylocom (Webb et al. 2008) and two null models: i) 760–775. Diversity Patterns in Geonoma 333

CURRIE,D.J.,G.G.MITTELBACH,H.V.CORNELL,R.FIELD,J.F.GUEGAN,B.A. JONES, M. M., H. TUOMISTO,D.B.CLARK, AND P. OLIVAS. 2006. Effects of me- HAWKINS,D.M.KAUFMAN,J.T.KERR,T.OBERDORFF,E.O’BRIEN, AND J. soscale environmental heterogeneity and dispersal limitation on floristic R. G. TURNER. 2004. Predictions and tests of climate-based hypotheses of variation in rain forest ferns. J. Ecol. 94: 181–195. broad-scale variation in taxonomic richness. Ecol. Lett. 7: 1121–1134. KAHN, F., AND J.-J. DE GRANVILLE. 1992. Palms in forest ecosystems of Amazon- DICKSON, T. L., AND B. L. FOSTER. 2008. The relative importance of the species ia. Ecological studies, 95. Springer-Verlag, Berlin, Germany. pool, productivity and disturbance in regulating grassland plant species KAY, K. M., P. A. REEVES,R.G.OLMSTEAD, AND D. W. SCHEMSKE. 2005. Rapid richness: A field experiment. J. Ecol. 96: 937–946. speciation and the evolution of hummingbird-pollination in Neotropi- DESDEVISES, Y., P. LEGENDRE,L.AZOUZI, AND S. MORAND. 2003. Quantifying cal Costus subgenus Costus (Costaceae): Evidence from nrDNA ITS and phylogenetically structured environmental variation. Evolution 57: ETS sequences. Am. J. Bot. 92: 1899–1910. 2647–2652. KENT, M., AND P. COKER. 1992. Vegetation description and analysis. A practical DONOGHUE, M. J. 2008. A phylogenetic perspective in the distribution of plant approach. John Wiley & Sons, Chichester, U.K. diversity. Proc. Natl. Acad. Sci. USA. 105: 11549–11555. KESSLER, M. 2000. Upslope-directed mass effect in palms along an Andean ele- DRUMMOND,A.J.,AND A. RAMBAUT. 2007. BEAST: Bayesian evolutionary analysis vational gradient: A cause for high diversity at mid elevations? Biotropica by sampling trees. BMC Evol. Biol. 7: 214, doi: 10.1186/1471-2148-7-214. 32: 756–759. DUFREˆNE, M., AND P. LEGENDRE. 1997. Species assemblages and indicator spe- KREFT, H., AND W. JETZ. 2007. Global patterns and determinants cies: The need for a flexible asymmetrical approach. Ecol. Monogr. 67: of vascular plant diversity. Proc. Natl. Acad. Sci. USA 104: 5925– 345–366. 5930. DUQUE, A., M. SANCHEZ,J.CAVELIER, AND J. DUIVENVOORDEN. 2002. Different KRISTIANSEN, T., J. C. SVENNING,C.GRANDEZ,J.SALO, AND H. BALSLEV. 2009. floristic patterns of woody understorey and canopy in Colombian Commonness of Amazonian palm (Arecaceae) species: Cross scale links Amazonia. J. Trop. Ecol. 18: 499–525. and potential determinants. Acta Oecol. 35: 554–562. FREESTONE, A. L., AND S. HARRISON. 2006. Regional enrichment of local assem- KRUSKAL, J. B. 1964. Nonmetric multidimentional scaling: A numerical blages is robust to variation in local productivity, abiotic gradients, and method. Psychometrika 29: 115–129. heterogeneity. Ecol. Lett. 9: 95–102. LAVIN, M. 2006. Floristic and geographical stability of discontinuous season- GENTRY, A. H. 1988. Changes in plant community diversity and floristic com- ally dry tropical forests explains patterns of plant phylogeny and position on environmental and geological gradients. Ann. Missouri Bot. endemism. In R. T. Pennington, G. P. Lewis, and J. A. Ratter (Eds.). Gard. 75: 1–34. Neotropical savannas and seasonally dry forests: Plant diversity, GOULD, S. J., AND R. C. LEWONTIN. 1979. The spandrels of san Marco and the biogeography, and conservation. pp. 433–447. CRC Press, Boca Raton, panglossian paradigm: A critique of the adaptationist program. Proc. R. Florida. Soc. Lond. Ser. B. 205: 581–598. LAVIN, M., B. P. SCHRIRE,G.LEWIS,R.T.PENNINGTON,A.DELGADO-SALINAS, GOVAERTS, R., AND J. DRANSFIELD. 2006 World checklist of Arecaceae. The M. THULIN,C.E.HUGHES,A.BEYRA MATOS, AND M. F. WOJCIECHOWSKI. Board of Trustees of the Royal Botanic Gardens, Kew. Available at 2004. Metacommunity process rather than continental tectonic history http://www.kew.org/wcsp/ (accessed 4 May 2009). better explains geographically structured phylogenies in legumes. Phil. GRAHAM, A. 2003. Geohistory models and Cenozoic paleoenvironments of the Trans. R. Soc. Lond. B 359: 1509–1522. Caribbean region. Syst. Bot. 28: 378–386. LEGENDRE, P., AND L. LEGENDRE. 1998. Numerical ecology (2nd edition). Else- GREGORY-WODZICKI, K. M. 2000. Uplift history of the central and Northern vier, Amsterdam, The Netherlands. Andes: A review. Geol. Soc. Am. Bull. 112: 1091–1105. LINDLER, H. P., C. R. HARDY, AND F. RUTSCHMAN. 2005. Taxon sampling effects GUSTAFSSON, M., AND V. BITTRICH. 2002. Evolution of morphological diversity in molecular clock dating: An example from the African Restoniaceae. and resin secretion in flowers of Clusia L. (Clusiaceae): Insights from Mol. Phylogenet. Evol. 35: 569–582. ITS sequence variation. Nord. J. Bot. 22: 183–203. MALARD, F., C. BOUTIN,A.I.CAMACHO,D.FERREIRA,M.GEORGES,B.SKET, HAWKINS, B. A., R. FIELD,H.V.CORNELL,D.J.CURRIE,J.F.GUEGAN,D.M. AND F. STOCH. 2009. Diversity patterns of stygobiotic crustaceans across KAUFMAN,J.T.KERR,G.G.MITTELBACH,T.OBERDORFF,E.M. multiple spatial scales in Europe. Freshwat. Biol. 54: 756–776. O’BRIEN,E.E.PORTER, AND J. R. G. TURNER. 2003. Energy, water, and MCCUNE, B., AND J. B. GRACE. 2002. Analysis of ecological communities. MjM broad-scale geographic patterns of species richness. Ecology 84: Software Design, Gleneden Beach, Oregon. 3105–3117. NEW, M., D. LISTER,M.HULME, AND I. MAKIN. 2002. A high-resolution data set HAWKINS, B. A., J. A. F. DINIZ,C.A.JARAMILLO, AND S. A. SOELLER. 2006. Post- of surface climate over global land areas. Clim. Res. 21: 1–25. Eocene climate change, niche conservatism, and the latitudinal diversity NORMAND,S.,J.VORMISTO,J.-C.SVENNING,C.GRANDEZ, AND H. BALSLEV. 2006. gradient of New world birds. J. Biogeogr. 33: 770–780. Geographical and environmental controls of palm beta diversity in paleo- HEIKINHEIMO, H., M. FORTELIUS,J.ERONEN, AND H. MANNILA. 2007. Biogeog- riverine terrace forests in Amazonian Peru. Plant Ecol. 186: 161–176. raphy of European land mammals shows environmentally distinct and PENNINGTON, R. T., C. A. PENDRY,W.GOODALL-COPESTAKE, AND S. O’SULLI- spatially coherent clusters. J. Biogeogr. 34: 1053–1064. VAN. 2004. Phylogenetic analysis of Rupechtia (Polygonaceae). In C. A. HENDERSON, A., G. GALEANO, AND R. BERNAL. 1995. A field guide to the palms Pendry. A monograph of Rupechtia (Polygonaceae). Syst. Bot. Monogr. of the Americas. Princeton University Press, Princeton, New Jersey. 67: 12–17. HERNANDEZ, R. M., T. E. JORDAN,A.D.FARJAT,L.ECHAVARRIA,B.D.IDLEMAN, PITMAN, N. C. A., J. W. TERBORGH,M.R.SILMAN,P.NUNEZ,D.A.NEILL,C.E. AND J. H. REYNOLDS. 2005. Age, distribution, tectonics, and eustatic CERON,W.A.PALACIOS, AND M. AULESTIA. 2001. Dominance and dis- controls of the paranense and Caribbean marine transgressions in south- tribution of tree species in upper Amazonian terra firme forests. Ecology ern Bolivia and Argentina. J. S. Am. Earth Sci. 19: 495–512. 82: 2101–2117. HILL, M. O., AND H. G. GAUCH. 1980. Detrended correspondence analysis: An POULSEN, A. D., H. TUOMISTO, AND H. BALSLEV. 2006. Edaphic and floristic improved ordination technique. Vegetatio 42: 47–58. variation within a 1-ha plot of lowland Amazonian rain forest. Biotro- HO, S. Y. W. 2009. An examination of phylogenetic models of substitution rate pica 38: 468–478. variation among lineages. Biol. Lett. 5: 421–424. PRINZING, A., W. DURKA,S.KLOTZ, AND R. BRANDL. 2001. The niche of higher HUBBELL, S. P. 2001. The unified neutral theory of biodiversity and biogeogra- plants: Evidence for phylogenetic conservatism. Proc. R. Soc. Lond. B. phy. Princeton University Press, Princeton, New Jersey. 268: 2383–2389. HUGHES, C., AND R. EASTWOOD. 2006. Island radiation on a continental scale: RANGEL, T. F., AND J. A. F. DINIZ-FILHO. 2005. An evolutionary tolerance model Exceptional rates of plant diversification after uplift of the Andes. Proc. explaining spatial patterns in species richness under environmental gra- Natl. Acad. Sci. USA. 103: 10334–10339. dients and geometric constraints. Ecography 28: 253–263. 334 Roncal, Blach-Overgaard, Borchsenius, Balslev, and Svenning

REE, H. R., B. R. MOORE,C.WEBB, AND M. J. DONOGHUE. 2005. A likelihood SIMON, M. F., R. GRETHER,L.P.DE QUEIROZ,C.SKEMA,R.T.PENNINGTON, framework for inferring the evolution of geographic range on phyloge- AND C. E. HUGHES. 2009. Recent assembly of the cerrado, a neotropical netic trees. Evolution 59: 2299–2311. plant diversity hotspot, by in situ evolution of adaptations to fire. Proc. RICHARDSON, J. E., L. W. CHATROU,J.B.MOLS,R.H.J.ERKENS, AND M. D. Natl. Acad. Sci. USA. 106: 20359–20364. PIRIE. 2004. Historical biogeography of two cosmopolitan families of STEBBINS, G. L. 1974. Flowering plants: Evolution above the species level. Har- flowering plants: Annonaceae and rhamnaceae. Phil. Trans. R. Soc. vard University Press, Cambridge, Massachusetts. Lond. B 359: 1495–1508. STEPHENS, P. R., AND J. J. WIENS. 2003. Explaining species richness from con- RICHARDSON, J. E., R. T. PENNINGTON,T.D.PENNINGTON, AND P. M. HOL- tinents to communities: The time-for-speciation effect in emydid turtles. LINGSWORTH. 2001. Rapid diversification of a species-rich genus of Neo- Am. Nat. 161: 112–128. tropical rain forest trees. Science 293: 2242–2245. SVENNING, J.-C., F. BORCHSENIUS,S.BJORHOLM, AND H. BALSLEV. 2008. High RICKLEFS, R. E. 2004. A comprehensive framework for global patterns in biodi- tropical net diversification drives the new world latitudinal gradient in versity. Ecol. Lett. 7: 1–15. palm (Arecaceae) species richness. J. Biogeogr. 35: 394–406. RICKLEFS, R. E. 2007. Estimating diversification rates from phylogenetic infor- SVENNING, J.-C., AND F. SKOV. 2005. The relative roles of environment and his- mation. Trends Ecol. Evol. 22: 601–6210. tory as controls of tree species composition and richness in Europe. J. RICKLEFS, R. E., AND D. SCHLUTER. 1993. Species diversity in ecological com- Biogeogr. 32: 1019–1033. munities: Historical and geographical perspectives. The University of TER BRAAK, C. J. F. 1986. Canonical correspondence analysis: A new eigenvector Chicago Press, Chicago, Illinois. technique for multivariate direct gradient analysis. Ecology 67: RICKLEFS, R. E., H. QIAN, AND P. S. WHITE. 2004. The region effect on meso- 1167–1179. scale plant species richness between eastern Asia and eastern North TUOMISTO, H., K. RUOKOLAINEN, AND M. YLI-HALLA. 2003. Dispersal, environ- America. Ecography 27: 129–136. ment, and floristic variation of western Amazonian forests. Science 299: RONCAL, J., F. BORCHSENIUS,C.B.ASMUSSEN-LANGE, AND H. BALSLEV. 2010. 241–244. Divergence times in tribe geonomateae (Arecaceae) coincide with ter- URQUHART, G. R. 1999. Long-term persistence of Raphia taedigera Mart. tiary geological events. In O. Seberg, G. Pedersen, A. S. Barfod, and J. I. swamps in Nicaragua. Biotropica 31: 565–569. Davis (Eds.). Diversity, phylogeny, and evolution of the monocotyle- WARREN, D. L., R. E. GLOR, AND M. TURELLI. 2008. Environmental niche dons. pp. 245–265. Aarhus University Press, Aarhus, Denmark. equivalency versus conservatism: Quantitative approaches to niche evo- RONCAL, J., J. FRANCISCO-ORTEGA,C.B.ASMUSSEN, AND C. E. LEWIS. 2005. lution. Evolution 62: 2868–2883. Molecular phylogenetics of tribe (Arecaceae) using nuclear WEBB, C. O. 2000. Exploring the phylogenetic structure of ecolo- DNA sequences of phosphoribulokinase and RNA polymerase II. Syst. gical communities: An example for rain forest trees. Am. Nat. 156: Bot. 30: 275–283. 145–155. RUOKOLAINEN, K., A. LINNA, AND H. TUOMISTO. 1997. Use of melastomataceae WEBB, C. O., D. D. ACKERLY, AND S. W. KEMBELL. 2008. Phylocom: Software and pteridophytes for revealing phytogeographical patterns in Amazon- for the analysis of phylogenetic community structure and character evo- ian rain forest. J. Trop. Ecol. 13: 243–256. lution. Bioinformatics 24: 2098–2100. RUTSCHMANN, F. 2006. Molecular dating of phylogenetic trees: A brief review of WIENS, J. J., AND M. J. DONOGHUE. 2004. Historical biogeography, ecology, and current methods that estimate divergence times. Divers. Distrib. 12: 35–48. species richness. Trends Ecol. Evol. 19: 639–644. SARKINEN, T. E., M. F. NEWMAN,P.J.M.MAAS,H.MAAS,A.D.POULSEN,D.J. WIENS, J. J., C. H. GRAHAM,D.S.MOEN,S.A.SMITH, AND T. W. REEDER. 2006. HARRIS,J.E.RICHARDSON,A.CLARK,M.HOLLINGSWORTH, AND R. T. Evolutionary and ecological causes of the latitudinal diversity gradient in PENNINGTON. 2007. Recent oceanic long-distance dispersal and diver- hylid frogs: Treefrog trees unearth the roots of high tropical diversity. gence in the amphi-Atlantic rain forest genus Renealmia L.f. (Zing- Am. Nat. 168: 579–596. iberaceae). Mol. Phylogenet. Evol. 44: 968–980. WIENS, J. J., G. PARRA-OLEA,M.GARCIA-PARIS, AND D. B. WAKE. 2007. Phylo- SESNIE, S. E., B. FINEGAN,P.E.GESSLER, AND Z. RAMOS. 2009. Landscape-scale genetic history underlies elevational biodiversity patterns in tropical sal- environmental and floristic variation in Costa Rican old-growth rain amanders. Proc. R. Soc. B. 274: 919–928. forest remnants. Biotropica 41: 16–26. WILLIS, J. C. 1922. Age and area. Cambridge University Press, Cambridge, U.K. SEZEN, U. U., R. L. CHAZDON, AND K. E. HOLSINGER. 2005. Genetic con- ZOBEL, M. 1997. The relative role of species pools in determining plant species sequences of tropical second-growth forest regeneration. Science 307: richness: An alternative explanation of species coexistence? Trends Ecol. 891. Evol. 12: 266–269.