bs_bs_banner Botanical Journal of the Linnean Society, 2014, 174, 141–161. With 3 figures Species limits and diversification in the Madagascar olive (Noronhia, Oleaceae) CYNTHIA HONG-WA1,2* and GUILLAUME BESNARD3 1 Department of Biology, University of Missouri–St. Louis, One University Blvd, St. Louis, MO 63121-4400, USA 2 Missouri Botanical Garden, PO Box 299, St. Louis, MO 63166-0299, USA 3 CNRS, ENFA, Université Paul Sabatier, Laboratoire Evolution et Diversité Biologique, UMR 5174, 31062 Toulouse 4, France Received 10 December 2012; revised 2 July 2013; accepted for publication 17 August 2013 Studies of ecological and phenotypic diversity in adaptive radiations have greatly contributed to our understanding of the patterns and processes of species diversification, whilst also challenging our assessment of the nature of species. Here, analyses of bioclimatic, molecular and morphological data, interpreted in phylogenetic and geographical contexts, were carried out to understand species limits in the Madagascar olive (Noronhia, Oleaceae). Most species hypotheses exhibit clear boundaries and are supported by at least one line of evidence, but a contrasting pattern of high morphological and ecological variation with relatively low nucleotide sequence divergence characterizes the diversification of Noronhia. This diversification was probably driven by fine-scale ecological and evolutionary processes, as suggested by the poor fit with four models of species diversification of the biota of Madagascar and the apparent lack of differentiation detected from large-scale bioclimatic data. Overall, this study offers useful insights into the patterns of plant diversification in Madagascar, the understanding of which requires good circumscription of species, improved knowledge of their distribution and operational models of diversification that take into account the particular biology of plants. © 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161. ADDITIONAL KEYWORDS: adaptive radiations – bioclimatic data – morphology – multivariate analyses – phylogeny – plant diversification. INTRODUCTION The unique and diverse biota of Madagascar testifies to adaptive radiations of a variety of groups of animals and plants (Goodman & Benstead, 2003; Vences, 2005). The island has been isolated from major landmasses since about 90 Mya (de Wit, 2003). Although some of the biota are assumed to be Gondwanan relicts, most are thought to be derived from Tertiary African and Asian colonizers (Yoder & Nowak, 2006; Russell, Goodman & Cox, 2008; Warren et al., 2010; Kuntner & Agnarsson, 2011; Reddy et al., 2012; Buerki et al., 2013; Hong-Wa & Besnard, 2013). In particular, a good proportion of the Malagasy endemic plant genera originated in the Miocene *Corresponding author. E-mail: cynthia.hong-wa@mobot.org onwards and show strong affinities, in decreasing order, with African, South-East Asian and Indian plant taxa (Buerki et al., 2013). Levels of taxonomic endemism and species diversity are high (Goodman & Benstead, 2003), endemism being estimated to be > 90% for non-volant and non-marine vertebrates and > 80% for vascular plants (Goodman & Benstead, 2003; Callmander et al., 2011). The spatial pattern of endemism is even more impressive, with many species having narrow ranges and being known from only one or a few localities (Schatz et al., 2000; Goodman & Benstead, 2003; Vences et al., 2009; Hong-Wa & Arroyo, 2012; Besnard et al., 2013). The plant genus Noronhia Stadman ex Thouars (Oleaceae), recently extended to include AfroMalagasy relatives formerly placed in Chionanthus L. to accommodate phylogenetic relationships (Hong-Wa © 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161 141 142 C. HONG-WA and G. BESNARD & Besnard, 2013), now comprises 54 described species in the Malagasy Floristic Region (MFR, including Madagascar, Comoros and Mascarenes; Takhtajan, 1986) that form a clade (the MFR clade; Hong-Wa & Besnard, 2013). Madagascar is the current centre of diversity of this genus with at least 46 endemic species; eight other species are found in the Comoros and Mascarenes, and a dozen species are found in western and eastern tropical and southern Africa. Colonization of Madagascar, from an African ancestor, may have occurred in the late Cenozoic (c. 23 Mya), and was followed by a burst of diversification starting at c. 15 Mya with subsequent colonization of smaller islands in the Indian Ocean (Hong-Wa & Besnard, 2013). Relationships in this group are currently largely unresolved, and basal divergences seem to have been relatively rapid, leaving little signal of lineage separation (Hong-Wa & Besnard, 2013). In contrast, the extent of morphological and ecological variation in this group is extremely high (Perrier de la Bâthie, 1951, 1952; C. Hong-Wa, pers. observ.). Indeed, there is a great variation in traits, such as leaf shape, size, texture and venation pattern, flower shape, size, colour and arrangement, and fruit shape, size and ornamentation. Noronhia also grows in arid to humid habitats from sea level to > 2000 m, specializes in various substrates, such as limestone and quartzite, and is a food source for various species of lemur and probably also for birds. Furthermore, Noronhia spp. often grow in sympatry, but show considerable fine-scale ecological differentiation related to pollination and dispersal, edaphic properties and topoclimatic conditions (Hong-Wa, 2012). Similar contrasting patterns have been observed in other Malagasy plant groups, such as Gaertnera Lam., Adenia Forssk., Cyathea Sm., Coffea L. subgenus Coffea and Pandanus Parkinson (Malcomber, 2002; Hearn, 2006; Janssen et al., 2008; Anthony et al., 2010; Davis et al., 2011; Buerki et al., 2012), which all exhibit wide variation in morphology, but low genetic divergence. Although the molecular study of the Malagasy flora has yet to be intensified, large morphological variation in many plant groups has been well documented and appears to be the signature of diversification processes operating on the island. Ecological variation, molecular evolution and morphological differentiation show contrasting patterns in Noronhia, and this raises questions about the boundaries of the species it contains. Morphology was initially used to recognize species in this group, which resulted in a classification (Perrier de la Bâthie, 1952) that is now outdated and unsatisfactory. Indeed, in spite of the extent of morphological variation in Noronhia, species recognition in this genus has been based mainly on the shape and size of flowers, the length of the corolla tube, the presence or absence of a corona, the shape of the corona lobe, the shape of the anthers and some vegetative features, such as leaf arrangement, shape, size and apex (Perrier de la Bâthie, 1952). This focus on a few floral characters, even though flowers of nine species were unknown at the time of their description (Perrier de la Bâthie, 1949, 1952), has made the identification of specimens of Noronhia challenging, especially when flowers are missing. Hence, many specimens of the genus, intensively collected in the last three decades, could not be assigned to any of the currently described species and were left as a huge pile of ‘Noronhia indet.’ in various herbaria (C. HongWa, pers. observ.). This has also been observed in other plant groups, especially large genera, such as Memecylon L. (Melastomataceae) (Stone, 2012), and represents a real impediment to the study of the Malagasy flora. Morphology remains important in species delimitation and most morphological plant species may correspond to reproductively independent lineages and represent biologically real entities (Rieseberg, Wood & Baack, 2006). However, boundaries of morphological species may be obscure, for instance, in the case of cryptic, plastic or polymorphic phenotypes (Duminil & Di Michele, 2009). As such, an integrative approach using multiple criteria (e.g. morphology, ecology and genetics) has been increasingly applied to the species delineation problem (Dayrat, 2005; Meudt, Lockhart & Bryant, 2009; Zapata, 2010; Barrett & Freudenstein, 2011). Certainly, meaningful recognition of species, considered as separately evolving lineages, can be based on multiple lines of evidence under the general lineage concept of species, but a single form of evidence can be sufficient to delimit species in any one case (de Queiroz, 2007). Each line of evidence, resulting from an evolutionary process affecting lineage splitting and divergence, may or may not appear at the same level, in the same order or with the same magnitude (de Queiroz, 2007; Padial & De la Riva, 2010). Therefore, congruence between lines of evidence, although desirable, is not necessary to recognize species and incongruence is even expected (de Queiroz, 2007; Padial & De la Riva, 2010). The recognition of species that follows an integrative framework without the necessity of congruence has been referred to as ‘integration by cumulation’ by Padial et al. (2010), an approach that considers the additive value of each line of evidence in distinguishing species. This approach is most appropriate in cases in which diversification occurred recently and/or over a relatively short period of time (Sites & Marshall, 2003; Padial et al., 2010). It is opposed to the ‘integration by congruence’ approach, in which agreement between at least two sources of evidence is necessary to recognize species (Padial et al., 2010). The two approaches are subject to overestimation and under-estimation of species number, respectively. © 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161 DIVERSIFICATION OF NORONHIA Malagasy Noronhia, being a monophyletic radiation and morphologically and ecologically diverse, represents an ideal system to study species diversification. The mechanisms by which diversity arose in Noronhia, and in fact most groups of organisms in Madagascar, are largely unknown. Patterns of diversification and endemism, mainly faunal, on Madagascar have been explained by various hypotheses. In particular, Yoder & Heckman (2006) proposed the ecogeographical constraint hypothesis to explain the east–west vicariance that follows the sharp bioclimatic division of Madagascar into a humid east and a dry west. Raxworthy & Nussbaum (1995) found the mountain massifs of northern Madagascar to be centres of endemism, and claimed that they have a large role in the generation and maintenance of diversity in this region. Wilmé, Goodman & Ganzhorn (2006) suggested that watershed contractions during past climatic oscillations led to zones of isolation, thus promoting microendemism. Angel (1942) and Martin (1972) proposed a zoogeographical zonation of Madagascar based on the distributions of its reptiles and lemurs, respectively, and the role of rivers as barriers to dispersal. Recent population genetic analyses confirm that rivers and geographical distance are primary factors structuring some rodent and lemur populations, although they may not act as strict barriers to dispersal (e.g. Quéméré et al., 2010; Rakotoarisoa, Raheriarisena & Goodman, 2010). All of these hypotheses emphasize the importance of physical barriers to gene flow in species divergence. By contrast, the current climate hypothesis (Pearson & Raxworthy, 2009) invokes a strong influence of environmental gradients in driving species divergence and in generating local endemism. Overall, these hypotheses suggest a major role of ecological diversification in lineage separation, and assume a predominantly adaptive speciation. Moreover, the environment of Madagascar is particularly heterogeneous (Dewar & Richards, 2007), and past changes may have led to successive population fragmentations and reconnections for many taxa (allopatry–sympatry oscillations). Such habitat dynamics may promote speciation events by reinforcement during secondary contacts (e.g. Aguilée, Lambert & Claessen, 2011). Although these various hypotheses were formulated to explain mainly animal species diversity, endemism and diversification on Madagascar, they may also be broadly applicable to the flora. In fact, in their paper advocating Madagascar as a model region for species diversification study, Vences et al. (2009) included examples from plant taxa, albeit scarce, that were consistent with the predictions of the various diversification mechanisms or diversity models [e.g. Coffea (the Baracoffea alliance), Beccariophoenix; Maurin et al., 2007; Shapcott et al., 2007; Davis & 143 Rakotonasolo, 2008]. These hypotheses, in particular the ecogeographical constraint (ECH), riverine barrier (RBH), watershed contraction (WCH) and current climate (CCH) hypotheses, can also shed light on the diversification of Noronhia. Indeed, phylogenetic analyses could clarify whether: (1) species or clades of the humid east separate from species or clades of the dry west according to the ECH; (2) species or clades are separated by major rivers according to the RBH and WCH; or (3) they diversify along an environmental gradient according to the CCH. The RBH and WCH, although developed from lemur studies, may be particularly relevant for Noronhia, as this genus is predominantly dispersed by various kinds of lemur (e.g. Eulemur, Lemur, Microcebus, Varecia etc.; Birkinshaw, 1999, 2001; Donati, Lunardini & Kappeler, 1999; Simmen et al., 2006; Radespiel, 2007; Andriamaharoa, Birkinshaw & Ludovic, 2010; Martinez, 2010; Thorén, 2011) and, perhaps, also by some birds, but little is known about the latter. Thus, the distribution of Noronhia, which covers the entire island, although as many as 19 of the 46 currently described Malagasy species have narrow ranges, may mirror that of its lemur dispersers. Specimens accumulated since the last taxonomic treatment of Noronhia (Perrier de la Bâthie, 1952) have greatly improved our understanding of the distribution patterns of species, which fall into three range categories: narrow, moderate and wide. Widely distributed species are mostly in the humid east, whereas those with moderate and narrow ranges are in the dry west and topographically complex areas of the north and south, respectively (C. Hong-Wa, unpubl. data; but see also Perrier de la Bâthie, 1951). Narrow endemism may reflect sampling bias owing to the difficulty of access to some areas, but may also reflect the role of topographical complexity in the diversification of this genus, as richness and endemism are highest in mountainous regions (Hong-Wa, 2012). Although taxon sampling remains challenging, we estimated that ours was a fair representation of the current state of knowledge on this group. Moreover, sympatric species are often more distinct from each other than they are from allopatric species, thus leading to potential underestimation of the actual diversity. Indeed, under the various scenarios of diversification discussed here, we can largely assume some cryptic diversity that is certain to be a taxonomic challenge. Thus, far from being exhaustive, our current estimate of Noronhia spp. provides a first approximation of the diversity in this genus that can be subsequently refined with further botanical, geographical and molecular sampling. However, the radiation of Noronhia also provides an opportunity to examine the aforementioned diversification processes from a plant perspective and to appreciate the taxo- © 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161 144 C. HONG-WA and G. BESNARD nomic complexity of a morphologically heterogeneous but genetically less diverse group. Our approach to understanding diversification in Noronhia focused on phylogeny, taxonomy and biogeography. In particular, we used a multifaceted, integrated approach to: (1) reassess the phylogenetic relationships among Noronhia spp. distributed in the MFR with a denser taxon sampling than in our previous study (Hong-Wa & Besnard, 2013); (2) examine patterns of morphological variation and species limits in a phylogenetic context and across geographical scales; and (3) evaluate the congruence between the predictions of the four hypotheses of diversification mentioned above with phylogenetic patterns in Noronhia. MATERIAL AND METHODS STUDY GROUP The basic taxonomy of Noronhia was confirmed using the morphological characters employed by Perrier de la Bâthie (1949, 1952). We then recorded additional qualitative diagnostic features [e.g. plant habit, presence of indumentum, colour of stem, leaf, flower and fruit, texture of leaf and fruit, leaf venation pattern, flower arrangement, inflorescence structure (diffuse vs. compact) and fruit shape, ornamentation, apex; Supporting Information Table S1] that allowed a better understanding of species limits and a refined circumscription of species in Noronhia. They are also suggestive of the extent of evolutionary changes in this group that perhaps correlate with the colonization of novel environments. We used these additional features, together with some previously recognized characters, to sort available specimens from the MFR into narrowly defined groups or ‘hypotheses of species’ (abbreviated from here onwards as species). Some named species that could not be distinguished from others on the basis of these qualitative characters were combined when they were also not differentiated in discriminant function analyses of quantitative variables (data not shown). These included, for instance, N. ecoronulata H.Perrier vs. N. alleizettei Dubard, and N. cruciata H.Perrier and N. verticilliflora H.Perrier vs. N. verticillata H.Perrier. In total, 87 qualitatively defined groups were distinguished {excluding two from the Mascarenes [N. ayresii (A.J.Scott) Hong-Wa & Besnard and N. boutonii (A.J.Scott) Hong-Wa & Besnard] for which highresolution images were seen, but no specimens were available for study} and were named when they corresponded to any of the 54 currently described species in the MFR; potentially new species are referred to as ‘Noronhia sp.’ and are followed by a number. These 87 putative species represent the operational taxonomic units (OTUs) used in the analyses of patterns of variation of bioclimatic, molecular and morphological variables in this study (see below). Twelve OTUs included only one or two specimens in our analyses. Thirty-three morphologically different specimens could not be assigned to any OTU. Some of these probably represent new OTUs, but the absence of either reproductive or highly distinctive characters prevented us from confidently assigning them to any group or considering them as distinct groups at the time this study was conducted. They were initially included in the multivariate analyses, but were subsequently removed for being uninformative. Additional specimens and molecular data will help to clarify their identity. PHYLOGENETIC ANALYSES Plastid (trnL-F, trnT-L, trnS-G, trnK-matK) and nuclear (internal transcribed spacer, ITS) DNA sequences were obtained from 68 of the 87 OTUs in the MFR, 42 of which had been included in an earlier study (Hong-Wa & Besnard, 2013). African relatives were also included and other members of Oleaceae were represented as outgroup taxa; specimens of subtribe Schreberineae (Comoranthus Knobl. and Schrebera Roxb.) were used to root the tree. In total, this molecular dataset included 157 accessions, with 40 of the 68 OTUs of Noronhia being represented by multiple individuals (Supporting Information Table S2). Voucher specimens are listed in Table S2 and sampling localities are shown in Figure 1. Thirteen potentially new species remain to be sampled for DNA. Other OTUs corresponding to the described species of Noronhia and lacking molecular data are N. crassinodis H.Perrier, N. jeremii Hong-Wa & Callm., N. leandriana H.Perrier, N. populifolia H.Perrier, N. urceolata H.Perrier and N. verrucosa H.Perrier, all from Madagascar; N. populifolia is only known from the type specimen that is 106 years old. Laboratory protocols, primers used and data preparation are described in Hong-Wa & Besnard (2013). All sequences have been submitted to GenBank (accession numbers are given in Table S2). Despite inconsistencies between the plastid and ITS datasets [P < 0.05 from the partition homogeneity test in PAUP* 4.0b10 (Swofford, 2002)], topological discordances among MFR OTUs were considered to be soft incongruences, as they were either weakly supported in both trees (bootstrap < 50%) or were part of polytomies in one of them (data not shown, but see also Hong-Wa & Besnard, 2013). The use of the incongruence length difference (ILD) test (Farris et al., 1995), as implemented by the partition homogeneity test, in the assessment of data partition combinability has been discouraged because of its sensitivity (Yoder, © 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161 DIVERSIFICATION OF NORONHIA 145 Figure 1. Geographical locations of the samples used for molecular analysis superimposed on different biogeographical zonations of Madagascar. A, Sampling localities. B, Bioclimatic zones of Schatz (2000). Letters refer to different zones: D, dry; H, humid; SA, subarid; SH, subhumid. C, Zoogeographical zonations/riverine barriers of Martin (1972); CP, central plateau; E1 and E2, east 1 and east 2; N, north; NW, north-west; W1 and W2, west 1 and west 2, Sb, Sambirano. D, Centres of endemism of Wilmé et al. (2006); numbers and letters indicate centres of endemism and zones of retreat– dispersion (also hatched), respectively. E, Climate clusters of Pearson & Raxworthy (2009); the different clusters are shown with different symbols in the legend. © 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161 146 C. HONG-WA and G. BESNARD Irwin & Payseur, 2001; Barker & Lutzoni, 2002; Darlu & Lecointre, 2002). The incongruence detected in our dataset concerns the outgroups more strongly than the MFR clade (see Hong-Wa & Besnard, 2013). Therefore, we chose to follow a total evidence approach and combined the two datasets into a single matrix for subsequent analyses, as our focus here was only on the MFR clade, especially the Malagasy species. Statistics of the different alignments are given in Supporting Information Table S3. Phylogenetic analyses of the combined dataset used maximum likelihood (ML) conducted on RAxML v7.2.6 (Stamatakis, 2006) and Bayesian inference (BI) carried out on MrBayes v.3.2.1 (Ronquist & Huelsenbeck, 2003) on the CIPRES portal (Miller, Pfeiffer & Schwartz, 2010) using the same program settings as those described in Hong-Wa & Besnard (2013) and GTR+I+G as the best-fit model of nucleotide substitution for each partition (i.e. plastid and ITS). Exceptions to the BI approach, which consisted of two parallel runs of four chains (one cold and three hot), included 75 million generations sampled every 7500th generation and a Markov chain Monte Carlo (MCMC) that used an exponential prior of unity for the shape parameter. Temperature was also reduced to 0.02 to allow optimal chain swapping. Model parameters were unlinked across partitions. Convergence of the two runs was checked with the online application AWTY (Nylander et al., 2008). A 50% majority-rule consensus tree was generated after discarding the first 20% of trees as a burn-in. Bootstrap support (BS) values and posterior probabilities (PPs) obtained from the ML and BI analyses, respectively, are shown below (Fig. 2). PATTERNS OF MORPHOLOGICAL AND NICHE VARIATION For this study, we measured 973 herbarium specimens from the MFR deposited at G, MO, P, TAN and TEF (abbreviated according to the Index Herbariorum; Holmgren, Holmgren & Barnett, 1990) to evaluate patterns of morphological variation among the 87 OTUs across their geographical ranges. Thirty-six quantitative characters (14 leaf, 11 flower and 11 fruit) were recorded (Supporting Information Table S4). A ruler was used to measure the length and width of leaf and fruit parts and a digital caliper for diameter and thickness. All flower parts were measured under a dissecting microscope. For each character, three independent measurements from organs of different sizes were taken from each specimen to account for within-individual variation. The average of these three measurements was assembled in a data matrix that included specimens in rows and variables in columns. Data were log-transformed and scaled to unit variance before analyses in an attempt to ensure normality and to reduce the effect of measurement units. Principal component analysis (PCA) based on correlation matrices of quantitative morphological variables was performed using the package stats on R 2.15.0 (R Development Core Team, 2012) to assess patterns of variation among the 87 qualitatively defined OTUs. Only those axes with eigenvalues > 1.0 were extracted using the eigenvalue-one rule or the Kaiser criterion (Quinn & Keough, 2002). As PCA is insensitive to collinearity and its success actually depends on the presence of correlated variables (Quinn & Keough, 2002; Gotelli & Ellison, 2004), we excluded none of them. In general, there were few specimens of each OTU with both flowers and fruits, and so analyses of flower and fruit variables were carried out separately. In addition, measurements were taken from mature organs to reduce size bias, which led to additional instances of missing values. Overall, we chose to maximize the number of OTUs and individuals in each analysis when missing data were an issue, thus occasionally excluding some characters. Results of the PCAs were plotted on twodimensional scatterplots for visualization. We also ran a stepwise discriminant function analysis (DFA) on SPSS (SPSS Inc.) to test the boundaries of the OTUs and identify the most taxonomically useful variables. To avoid spurious interpretation, we only carried out the DFAs when the sample size permitted the estimation of the group centroid. Likewise, to identify the climate space and assess patterns of niche variation for species, we analysed 19 bioclimatic variables and elevation (Supporting Information Table S5) using PCA, as this provides a quick assessment of the relative positions of each species in climate space (Zhu et al., 2012; and references therein). We obtained the 19 bioclimatic variables and elevation from the WorldClim database (Hijmans et al., 2005), which contains a set of global climate layers with a spatial resolution of c. 1 km × 1 km generated through interpolation of climate data (monthly total precipitation and monthly mean, minimum and maximum temperature) obtained from climate station records from 1950–2000. We used ArcGIS (ESRI Inc., Redlands, CA, USA) to extract the bioclimatic and elevation data for each specimen. We tested for multicollinearity among variables, but did not find any significant strong correlation (Pearson’s r ≥ 0.90) at the 0.05 level, suggesting that there is some variation in each variable that can potentially influence niche variation. Therefore, we did not remove any bioclimatic variables to avoid excluding those with potential biological relevance, acknowledging that overfitting may lead to poor performance of the analysis. However, given the property of PCA mentioned above, correlated variables are not expected to affect © 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161 DIVERSIFICATION OF NORONHIA 147 Figure 2. Maximum likelihood tree for Noronhia based on the combined plastid and internal transcribed spacer (ITS) dataset. Numbers above and below the branches are posterior probabilities (PP) and bootstrap support (BS), respectively. Bold letters refer to supported clades. Dark grey represents African species and light grey indicates species occurring in Madagascar and the surrounding islands (Comoros and Mascarenes). © 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161 148 C. HONG-WA and G. BESNARD greatly the interpretability of our data. Moreover, over-parameterization has been found to be less problematic than under-parameterization in model building (Warren & Siefert, 2011). PATTERNS OF DIVERSIFICATION We used different biogeographical zonations of Madagascar to represent the four hypotheses of diversification on the island: (1) the bioclimatic zones of Schatz (2000) for ECH, noting that the same four main types of biomes summarizing the 14 bioclimatic zones of Cornet (1974) were recognized by both Schatz (2000) and Yoder & Heckman (2006); (2) the zoogeographical zonations of Martin (1972) for RBH; (3) the centres of endemism of Wilmé et al. (2006) for WCH; and (4) the climate clusters of Pearson & Raxworthy (2009) for CCH (Fig. 1, adapted from the respective articles). Phylogenetic predictions could be derived for the four hypotheses of diversification (Vences et al., 2009). In particular, the ECH predicts an east–west partition between clades or sister species given the sharp bioclimatic distinction between these two regions or, by extension, a genetic break between clades or species from different bioclimatic zones. The RBH predicts genetic differentiation between clades or sister species occurring on either side of major rivers or, by extension, a genetic break between clades or species from areas separated by any biogeographical barrier, such as a mountain range. The WCH predicts a genetic differentiation among watersheds that served as zones of isolation. It also predicts that sister species will occupy contiguous watersheds. Likewise, the CCH predicts a genetic break between clades or species from different climate clusters. In the humid eastern escarpment and the central highlands, sister species are expected to occupy adjacent climate clusters along elevational gradients, whereas, in the dry western lowlands, sister species are expected to be separated along a north–south gradient. Overall, genetic differentiation is expected to increase with geographical distance or the presence of physical barriers to gene flow. We thus tested for correlation between pairwise genetic and pairwise geographical distances for each clade using a Mantel test performed in the R program and the package ade4 (Dray & Dufour, 2007). Pairwise genetic distances within a particular clade were estimated with the package ape (Paradis, Claude & Strimmer, 2004) on R by considering only a subset of the alignment that included the members of that clade exclusively. As the best-fit model of nucleotide substitution (GTR) is not implemented on the package ape for the estimation of genetic distance, we used the next best model available (K80) as an alternative. Similarly, pairwise geographical distances within each clade were obtained by calculating the Euclidean distance between the coordinates of the OTUs using the package stats on R. To assess the concordance between the observed phylogenetic splits and the phylogenetic and spatial patterns predicted by each of the four hypotheses, we first identified the physical barriers and biogeographical subdivisions in each hypothesis and other barriers that seemed to us to be effective. We then mapped these onto the phylogenetic tree, considering only the supported clades. Taking into account the ranges of species (Fig. 3), a match between a phylogenetic split and a biogeographical pattern is considered as support for the hypothesis it represents. For instance, a split between sister species or clades corresponding to an east–west partition supports the ECH if, and only if, the range of species A is confined to the east and that of species B to the west. As these hypotheses were formulated only for mainland Madagascar, excluding smaller islands such as Nosy Be or Sainte Marie, species or individuals occurring on these smaller islands were removed from this assessment of spatial patterns of diversification. This approach is rather qualitative, but, combined with the Mantel test, offers preliminary insights into the patterns of, and potential processes involved in, the diversification of Noronhia. RESULTS PHYLOGENETIC ANALYSES The combined plastid (trnL-F, trnT-L, trnS-G, trnKmatK) and nuclear (ITS) dataset contained 4500 bp, 691 bp of which were variable and 438 bp of which were potentially parsimony informative; the overall mean sequence divergence was 5.3%. The statistics of the different alignments are summarized in Table S3. The analyses of the combined dataset resulted in highly concordant Bayesian and likelihood phylogenetic hypotheses within Noronhia (Fig. 2). The inclusion of additional taxa of Noronhia in this study supports previous findings that the genus is strongly supported as monophyletic (BS = 100%, PP = 1). The MFR taxa were clearly separated from African species (BS = 96%, PP = 1), supporting the single radiation of the genus on Madagascar, with subsequent colonization of surrounding islands. Relationships within this MFR clade were characterized by short internodes and large basal polytomies, and were resolved into only 16 clades with moderate to high support values (BS ≥ 70% and/or PP ≥ 0.85). These clades are identified here with capital letters (A–P; Fig. 2) and will be referred to accordingly. Within each clade, OTUs, usually represented by multiple individuals, separated clearly from each other (Fig. 2). For instance, the three OTUs (N. capu- © 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161 DIVERSIFICATION OF NORONHIA 149 Figure 3. Maps showing the distribution of species falling into clades and analysed in a phylogenetic context (Figs S2– S17). © 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161 150 C. HONG-WA and G. BESNARD ronii Bosser, N. gracilipes H.Perrier and N. sambiranensis H.Perrier) forming clade B were all reciprocally monophyletic. This was also the case in clades C, E, F, I, J, K, L, N and P. By contrast, ambiguous relationships were observed in clades A, D, G, H, M and O. Thus, in clade D, the three individuals representing N. linearifolia Boivin ex Dubard did not cluster together. Similarly, the two individuals of N. boivinii Dubard in clade H and the individuals of N. pervilleana (Knobl.) H.Perrier and N. sp32 in clade O failed to form monophyletic groups, but were either paraphyletic or formed parts of polytomies. Finally, individuals of N. brevituba H.Perrier and N. linocerioides H.Perrier in clade M were not reciprocally monophyletic, but formed a strongly supported mixed clade. The northern population of N. brevituba, represented by N. brevituba 2 and N. brevituba 3 in the tree (Fig. 2), is sufficiently distinct from the eastern population, represented by N. brevituba 1, and may deserve separate status. There were also instances in which different individuals of one OTU occurred in more than one clade or were parts of polytomies (e.g. N. grandifolia H.Perrier, the phylogenetic placement of which reflects some geographical patterning). However, within the large polytomy (Fig. 2), individuals of the same OTUs generally clustered together with high support values (e.g. N. crassiramosa H.Perrier, N. comorensis H.Perrier, N. decaryana H.Perrier and N. louvelii H.Perrier). The potentially new species of Noronhia included here showed overall good genetic differentiation from each other and from described species (Fig. 2). For instance, clade C was composed of only two putative species (N. sp11 and N. sp13), both of which were well separated. Clades L and P also included putative species (N. sp38, and N. sp2 and N. sp22, respectively) that were reciprocally monophyletic and distinct from the described species. However, such clear patterns were not always recovered. In some cases, individuals of these putative species did not cluster (N. sp28 and N. sp30) or, if they did, were part of the large polytomy (N. sp5, N. sp15, N. sp27). PATTERNS OF MORPHOLOGICAL AND NICHE VARIATION Patterns of both morphological and niche variation were interpreted in phylogenetic and geographical contexts. The phylogenetic context was established by focusing on clades with BS ≥ 70% and/or PP ≥ 0.90. In total, 16 such clades were identified (Fig. 2), and PCAs and DFAs of morphological and bioclimatic variables were carried out independently on these clades to estimate patterns of variation among closely related OTUs. For OTUs that were part of polytomies or that were not available for molecular study, analyses were conducted in a biogeographical context using the subdivisions proposed by Wilmé et al. (2006), as many species have narrow to moderate ranges confined to one or a few of these. Independent PCAs and DFAs of morphological and bioclimatic variables were performed within each of the 12 biogeographical units to assess patterns of variation among co-occurring OTUs. Results of the PCAs of morphological and bioclimatic variables on the 16 phylogenetic clades and 12 biogeographical zones are summarized in Tables 1 and 2 and Supporting Information Figs S2–S31. OTUs that clearly separate from others in morphospace and climatic space are represented by (+), whereas those that occupy contiguous or overlapping morphospace regions are indicated by (–) in Tables 1 and 2. Overall, the OTUs analysed here showed distinct patterns of variation. Individuals of most OTUs formed a cluster distinct from other such clusters in multivariate analyses of quantitative vegetative characters alone or a combination of quantitative vegetative and flower or fruit characters. In some cases, distinction among OTUs was observed only in separate analyses of flower or fruit characters (Tables 1 and 2, Figs S2, S10 and S19). In other cases, the presence of OTUs behaving as outliers as a result of extreme variations (e.g. very long leaf blades, very large flowers) obscured the patterns of variation of the remaining OTUs by constraining them in one space of the scatterplot. The exclusion of these outliers allowed us to focus on the patterns of variation of the remaining OTUs (e.g. Figs S9, S14 and S16). Of the 50 OTUs placed in clades in the phylogenetic analyses, it is noteworthy that all but six OTUs (N. boivinii, N. brevituba, N. linocerioides, N. luteola var. ankaranensis H.Perrier, N. mangorensis H.Perrier and N. sp9) formed discrete morphological clusters in at least one analysis based on quantitative vegetative, flower or fruit features only or some combination of these three datasets. Of the 37 OTUs falling into large polytomies or lacking molecular data, and thus analysed in a geographical context, only four (N. verrucosa, N. sp5, N. sp15 and N. sp40) showed overlapping patterns of variation. OTUs that included one or two specimens only were also mostly distinguishable. OTUs showing ambiguous morphologies, such as N. aff. candicans and N. candicans H.Perrier, also appeared distinct, although only vegetative characters were available (Fig. S30A). These two taxa were phylogenetically unrelated, but geographically sympatric. Noronhia aff. candicans was also distinct from other OTUs occurring in the same area (Fig. S30B). Likewise, N. sp21 is generally similar to N. buxifolia, but both were distinct in the morphospace (Fig. S31). Analyses of bioclimatic and elevation variables showed a high degree of overlap between the climate space of different OTUs. This is © 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161 DIVERSIFICATION OF NORONHIA 151 Table 1. Summary of the results of principal component analysis (PCA) and discriminant function analysis (DFA) on bioclimatic and quantitative morphological data of species belonging to clades A–P (Figs S2–S17). Signs indicate the presence (+) or absence (–) of distinction in the two-dimensional scatterplots of the PCAs. [1] indicates that only a single specimen was available in the analysis. Abbreviations: mi, material insufficient; NA, not applicable; nk, not known. Level of significance of Wilks’ λ: *0.01 < P ≤ 0.05, **0.001 < P ≤ 0.01, ***P ≤ 0.001 Morphology Bioclimate Vegetative Flower Fruit Clade Species PCA DFA Wilks’ λ PCA DFA Wilks’ λ PCA DFA Wilks’ λ PCA DFA Wilks’ λ A grandifolia introversa capuronii gracilipes sambiranensis sp11 sp13 candicans linearifolia ambrensis broomeana sp20 buxifolia myrtoides alleizettei boinensis tubulosa sp21 boivinii densiflora mangorensis ovalifolia tropophylla seyrigii divaricata sp34 minoriflora luteola ankaranensis emarginata oblanceolata peracuminata sp9 sp38 brevituba linocerioides verticillata sp25 obtusifolia edentata lanceolata cochleata humbertiana pervilleana sp32 incurvifolia insularis sp2 cordifolia sp22 – – – – – – – – – + + + – – – – – – – – – – – – – – [1] + + – + – – – – – – – – – – + – – – – – + – – – 0.501*** – – + + + + + + + + + + + + – + – + – + – + – – + + [1] + + – + + + – + – – + – + + + – – – + – + – – + 0.367*** + + + + [1] + + + [1] + [1] + nk + [1] + + + + + [1] + + [1] – mi – + + + + + [1] + + + + mi + nk nk ± ± + nk + + + + + [1] + + + + [1] + mi nk 0.005* – – + + + + [1] + + + + + [1] nk + + – + – – [1] – mi – + [1] – – + + [1] + + – + + + – nk – – + + mi + + – + – + – + [1] – + [1] nk 0.569 B C D E F G H I J K L M N O P 0.053*** 0.284*** 0.249*** 0.000*** 0.453*** 0.608*** 0.071*** 0.421*** NA 0.222*** 0.321*** 0.396*** 0.136*** 0.013*** 0.000*** 0.056*** 0.003*** 0.044*** 0.269*** 0.275*** 0.174*** 0.011*** 0.646*** NA 0.124*** 0.000* 0.053*** 0.011*** 0.069*** 0.001*** NA NA NA NA 0.098* NA 0.416*** NA NA 0.000*** NA 0.072*** 0.000*** NA NA © 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161 0.052*** NA 0.064* NA 0.182* NA NA 0.697* NA 0.024** NA NA 0.012* NA NA 152 C. HONG-WA and G. BESNARD Table 2. Summary of the results of principal component analysis (PCA) and discriminant function analysis (DFA) on bioclimatic and quantitative morphological data of species found within the 12 centres of endemism (CE) and northern (RDN) and southern (RDS) retreat–dispersion watersheds of Wilmé et al. (2006), and the Comoros (COM) (Figs S18–F29). Signs indicate the presence (+) or absence (–) of distinction in the two-dimensional scatterplots of the PCAs. [1] indicates that only a single specimen was available in the analysis. Abbreviations: mi, material insufficient; NA, not applicable; nk, not known. Level of significance of Wilks’ λ: *0.01 < P ≤ 0.05, **0.001 < P ≤ 0.01, ***P ≤ 0.001. Species occurring in more than one watershed are shown in bold Morphology Bioclimate Vegetative Flower Fruit Zone Species PCA DFA Wilks’ λ PCA DFA Wilks’ λ PCA DFA Wilks’ λ PCA DFA Wilks’ λ CE1 aff. candicans crassinodis aff. crassinodis longipedicellata louvelii sp17 sp28 sp30 thouarsii crassiramosa decaryana jeremii louvelii verrucosa sp4 sp5 sp6 sp8 sp15 sp26 sp36 sp40 crassiramosa decaryana louvelii sp31 sp37 decaryana sp15 sp26 sp14 sp15 sp16 sp18 sp26 leandriana louvelii urceolata comorensis leandriana louvelii populifolia sp19 humblotiana jeremii sp19 sp36 sp39 – – – + – – – – – – – – – – – – – – – – – – – – + – – + + + – – – – – – + – + + – + – + – – + – NA + + + + + + + + – + – – – – + ± + – ± – – – + + + + + + + + + + + + + + + + + + + + + + + + + + NA nk + + mi + [1] nk nk nk mi nk + mi + nk nk ± nk nk ± + + [1] nk nk mi mi + + mi mi mi + + + nk + [1] + + [1] + + + [1] mi mi nk mi mi nk mi nk NA nk – – [1] – [1] mi + + [1] + [1] + [1] + [1] – + [1] – – [1] + [1] – + – – [1] – mi nk mi mi mi mi mi mi mi mi nk mi + [1] + [1] + [1] mi mi mi mi mi + [1] + [1] + + [1] + [1] – – [1] + NA CE2 CE3 CE4 CE5 CE8 CE9 + COM CE10 CE11 [1] 0.132*** NA [1] [1] NA [1] NA [1] NA [1] NA [1] [1] [1] NA [1] NA [1] [1] [1] 0.005*** [1] NA [1] [1] [1] NA [1] NA [1] [1] NA [1] NA [1] [1] [1] NA [1] NA [1] NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA © 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161 DIVERSIFICATION OF NORONHIA 153 Table 2. Continued Morphology Bioclimate Vegetative Flower Fruit Zone Species PCA DFA Wilks’ λ PCA DFA Wilks’ λ PCA DFA Wilks’ λ PCA DFA Wilks’ λ CE12 thouarsii crassinodis aff. crassinodis humblotiana longipedicellata sp27 sp30 decaryana humblotiana longipedicellata louvelii planifolia urceolata sp4 sp7 sp8 sp19 sp36 decaryana leandriana sp1 – – – – + – – – – – – – – – – – – – + + + 0.006*** + + – – – + – + + + + + – + – + + – – – + 0.001*** mi + [1] mi + [1] + + nk + [1] + mi mi mi + [1] nk + [1] nk nk mi mi + [1] + NA + mi + [1] mi – [1] + [1] + [1] + [1] mi mi + [1] + mi + [1] + [1] mi mi + – – + NA RDN RDS [1] NA [1] [1] [1] [1] NA [1] [1] NA [1] [1] [1] [1] not surprising as Noronhia spp. often co-occur in several forested areas. However, OTUs occupying the same or similar climatic spaces could usually be differentiated in morphospace (e.g. Figs S3, S4 and S13). Results of the DFAs are also summarized in Tables 1 and 2, with a particular focus on the value of Wilks’ lambda, which was assessed at the significance level of 0.05. Wilks’ lambda is a test statistic whose value ranges from zero to unity. A value close to zero indicates that groups are well separated, and a value close to unity suggests poor discrimination between them. The results presented here concern an entire clade or biogeographical zone, without excluding overly distinct OTUs as in the PCAs. The group centroid could not be calculated for OTUs with a small sample size, resulting in ‘NA’ in our summary Tables 1 and 2. In general, OTUs distinct in the PCAs were also well discriminated in the DFAs (e.g. clades B and F in Table 1, Figs S3 and S7). In some cases, OTUs were also well discriminated despite overlapping patterns of variation in the PCAs. For example, members of clades O and P showed little separation in the twodimensional scatterplot of the PCAs (Figs S16 and S17), but were found to be significantly distinct in the DFAs. Indeed, OTUs may overlap in some regions of the morphospace, but still have different centroids. Lastly, there were a few cases in which OTUs indistinct in the PCAs were also poorly discriminated in NA NA NA NA NA the DFAs (e.g. clades A and I, Figs S2 and S10), but these occurred only for a subset of the variables, i.e. either vegetative, flower or fruit. Similar results were also found for the analyses of bioclimatic variables (Tables 1 and 2). Overall, OTUs could be distinguished on any one set of variables or their combination. PATTERNS OF DIVERSIFICATION There was a lack of correlation between genetic and geographical distances among OTUs within each clade (Table 3, Fig. 3), suggesting that geographical distance or the presence of physical barriers to gene flow alone was not a sufficient predictor of genetic distance. Indeed, genetic distance was fairly high within clades J, M, N and P (4.5%, 5.0%, 4.9% and 4.8%, respectively) regardless of geographical distance between members of the clades (16, 32, 217 and 314 km, respectively). Similarly, genetic distance was relatively low within clades D and O (1.8% and 1.2%, respectively), but their members were 43 and 383 km apart, respectively. There was also an overall lack of evidence in support of the ECH, as there was no evidence of strong phylogenetic and spatial fit with the predictions of the RBH, WCH and CCH (Supporting Information Fig. S1). Indeed, the ECH predicted four bioclimatic subdivisions or at least an east–west par- © 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161 154 C. HONG-WA and G. BESNARD Table 3. Results of Mantel tests between cladewise geographical and genetic distances among species within each clade. Signs represent support for (+), contradictions with (–) or uncertainty over (+/–) the predictions of the four hypotheses. CCH, current climate hypothesis; ECH, ecogeographical constraint hypothesis; RBH, riverine barrier hypothesis; WCH, watershed contraction hypothesis; r = Pearson’s correlation coefficient Genetic distance (%) Clade Mean geographical distance (km) Mean Maximum r P ECH RBH WCH CCH B C D E F G H I J K L M N O P 55 62 43 110 80 169 307 554 16 139 53 32 217 383 314 2 1.5 1.2 1.9 0.3 1.5 1.7 1.7 3.1 1.8 2.1 3 2.5 0.8 2.6 3.3 3 1.8 2.8 0.5 3 3 3.1 4.5 2.5 4.1 5 4.9 1.2 4.8 0.06 0.75 0.06 0.99 0.98 0.58 0.34 0.91 –0.70 0.93 –0.33 0.15 0.61 0.12 –0.23 0.29 0.15 0.30 0.17 0.35 0.05 0.03 0.13 0.85 0.32 0.80 0.21 0.06 0.17 0.75 + – – – – – +/– – – + +/– – + – – + – – + – – +/– – – + +/– – – – – + – – + – – +/– – – + +/– – – – – + – – – – – +/– – – + +/– – + – – tition, none of which was apparent in the phylogenetic tree, either among or within clades. Instead, OTUs from the same region or occupying the same bioclimatic zone tended to cluster together within clades, but the overall pattern of the phylogenetic tree was a mosaic. Likewise, the correspondence between the phylogenetic tree and major genetic breaks predicted by the RBH was rather weak. In many instances, OTUs occupying different sides of a river were more closely related than OTUs occurring on the same side. For example, N. capuronii and N. ambrensis, both occurring north of the Mahavavy river, belong to different clades within which they were related to N. sambiranensis and N. sp20, respectively, both found south of the Mahavavy river (Fig. S1). Similar patterns were also observed for the WCH, despite the expected correlation between the distribution of Noronhia and that of its lemur dispersers, on which this hypothesis was based. However, even though OTUs from the same watershed were not always closely related, and OTUs from different watersheds were not always particularly distinct (Fig. S1), geographical proximity seems to be of some importance. There was, for instance, no clade with OTUs from northern and southern Madagascar. Instead, related OTUs were found within clusters of geographically close watersheds, e.g. clade B (CE1, CE10 and CE12) or clade H (CE2, CE4 and CE5) (Fig. S1), suggesting a broadly biogeographically constrained differentiation. The CCH also did not obtain strong support from the data; sister OTUs did not always occur in adjacent clusters, but more often would occupy the same cluster, e.g. clade C, or one OTU would straddle two or more zones, e.g. in clade J (Fig. S1). DISCUSSION EVIDENCE FOR MULTIPLE DISTINCT SPECIES In this study, we considered three lines of evidence as operational criteria to recognize species: bioclimatic, molecular and morphological distinctions. Nineteen OTUs lacked molecular data, and so only two lines of evidence could be used for them. Moreover, the combined dataset of plastid (trnL-F, trnT-L, trnS-G, trnKmatK) and nuclear (ITS) DNA regions resulted in large polytomies, which meant that it was more difficult to assess whether the OTUs that fell into polytomies were distinct both morphologically and phylogenetically, a satisfactory criterion for the ‘integration by congruence’ approach (Padial et al., 2010). Partially unresolved topologies were obtained from the separate analyses of both plastid and nuclear datasets (data not shown, but see also Hong-Wa & Besnard, 2013). The analyses of the combined dataset also resulted in partially unresolved topologies, suggesting that the data combination only marginally improved the phylogenetic resolution. To explain this pattern, we favour rapid diversification as the most likely explanation, given the occurrence of short interior branches and unresolved basal polytomies (Kim et al., 1996; © 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161 DIVERSIFICATION OF NORONHIA Seelanan, Schnabel & Wendel, 1997; Wendel & Doyle, 1998; Malcomber, 2002), and the apparent imbalanced diversity between the MFR Noronhia and its sister lineage (Schluter, 2000), although hybridization cannot be excluded (see Hong-Wa & Besnard, 2013). Consequently, we used the ‘integration by cumulation’ approach (Padial et al., 2010), which has been recommended for lineages in which diversification was recent and/or rapid relative to that of their sister lineages (Sites & Marshall, 2003; Padial et al., 2010), and incorporated various lines of evidence (bioclimatic, molecular and morphological) to recognize species in Noronhia. The strength of the various operational criteria to evaluate the limits among the 87 OTUs varied, the bioclimatic data being the least discriminating. Indeed, phylogenetically related OTUs usually occupied the same broad climatic niche (Table 1, Figs S2–S17); analyses carried out in the geographical context obviously focused on OTUs from the same region, and so could hardly show bioclimatic differentiation (Table 2, Figs S18–S29). Molecular data showed clear differentiation among most OTUs despite unresolved relationships at deeper levels (Fig. 2). Likewise, quantitative morphological data clearly discriminated most phylogenetically or geographically related OTUs (Tables 1 and 2, Figs S2–S31). Despite differences in discriminating power, each operational criterion provided useful insights into species limits in Malagasy Noronhia, especially in cases of unresolved phylogenetic placement or obscure morphology. Indeed, given the contingency of each line of evidence in separately evolving lineages, concurrent support for the same species is not expected and the lack of support from a particular criterion does not justify the rejection of that species as long as it is likely to represent a distinct evolutionary trajectory (de Queiroz, 2007; Padial & De la Riva, 2010). However, their additive value allows species recognition and permits the inference of major forces underlying species differentiation. Overall, 84 of the 87 initial OTUs could definitely be recognized as species using the ‘integration by cumulation’ approach (Table 4), and will be treated as such in an upcoming taxonomic revision of Noronhia (C. Hong-Wa, unpubl. data). OTUs that were ambiguous include the three pairs N. boivinii and N. mangorensis, N. luteola var. ankaranensis and N. sp9, and N. brevituba and N. linocerioides (Table 1, Figs S9, S13 and S14). The morphometric analysis barely supported their initial recognition as OTUs. Nevertheless, there was some genetic distinction between members of each pair. In particular, accessions of N. mangorensis and N. luteola var. ankaranensis were phylogenetically cohesive and fairly distinct from those of N. boivinii and N. sp9, respectively, suggesting perhaps cryptic or convergent morphology. As the 155 Table 4. Number of species recognized based on different types of data and their combination Number of species Number of species recovered out of recovered with Lines of evidence 87 initial OTUs additional data Bioclimatic only + Molecular + Morphological + Molecular + morphological Molecular only + Morphological Morphological only 19 50 81 84 – 31 62 65 43 out of 68 with molecular data 84 81 – 41 – sequences for N. sp9 were obtained from a herbarium specimen, our ability to make inferences about the possible causes and implications of this distinction is, however, limited. Further molecular sampling will provide a better understanding of the relationship between these two OTUs. There was also a genetic distinction between the northern and eastern populations of N. brevituba, which may suggest distinct entities. Similarly, N. sp5 and N. sp15 were not clearly distinct in both climatic and morphological spaces (Table 2, Fig. S19), but were genetically different (Fig. 2) and geographically distant. Noronhia sp15 is littoral and N. sp5 is inland, occurring above 1000 m elevation; both grow in the eastern part of Madagascar. Thus, phylogenetic data allowed morphologically nearly similar OTUs to be distinguished. Obviously, the lack of molecular or morphological data can hamper species recognition or rejection. For instance, N. sp40 did not clearly differ from N. sp15 on the basis of vegetative characters alone despite distinctive morphology (Table 2, Fig. S19). In such cases, the inability to discriminate the different OTUs quantitatively is mostly attributable to insufficient morphological data and lack of molecular data. Hence, recognition is not rejected until further samples are available. MECHANISMS OF SPECIES DIVERSIFICATION Our results provided evidence for extensive morphological variation in Noronhia, coupled with poor signals of basal divergence from the genetic markers used here. This contrasting pattern suggests a rapid, recent and/or incomplete radiation (Kim et al., 1996; Malcomber, 2002; Dunbar-Co, Wieczorek & Morden, 2008). Whether this diversification is ecologically mediated remains to be determined, but the lack of differentiation in climatic space indicates that factors © 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161 156 C. HONG-WA and G. BESNARD other than climate have played a major role in driving this pattern. Likewise, the weak correlation between the distribution of Noronhia and that of its lemur dispersers, as assessed through the WCH, is rather disappointing, and can perhaps be accounted for by repeated but uncoupled contraction and expansion of their respective ranges. In any case, a correlation, although faint, still exists and suggests that some patterns of distribution may be explained by co-evolutionary processes. Specific studies on species pairs and consideration of other potential dispersers, such as birds, may provide useful insights into their role in the diversification and distribution of members of this genus. Other large-scale mechanisms suggested by the four models of diversification considered here also appeared to be of little significance in the diversification of Noronhia, given the overall poor fit between the phylogenetic tree and the physical and ecological barriers to gene flow (Fig. S1), and the lack of strong correlation between geographical and genetic distances (Table 3). Indeed, geographical isolation alone or the presence of physical barriers is not sufficient to explain patterns of diversification in Noronhia, as most species appeared to have diverged without a strong isolating role of rivers, drought or sharp ecogeographical features. Models invoking parapatric divergence along environmental gradients, such as the CCH (Pearson & Raxworthy, 2009), also account for some of the observed patterns (Table 3), and would presumably be more appropriate than the other three models as the CCH closely resembles the phytogeographical subdivisions of Humbert (1955). In general, the apparent lack of differentiation between different biogeographical zones and the pronounced differentiation in the same zone may reflect the signature of past climatic fluctuations and forest dynamics (e.g. Aguilée et al., 2011; Buerki et al., 2013). Moreover, the four models considered here have been formulated for terrestrial faunal diversification in Madagascar, and highly vagile species did not support these models (Weyeneth, Goodman & Ruedi, 2011). Thus, they may not be suitable for plants dispersed by these animals. In fact, it appears that these animal-based models of diversification have limited application in plant lineages because of intrinsic differences between these two groups. The explanation of plant diversification in Madagascar requires more complex models that take into account the particularities of the life histories of plants and the porosity of their genome. Indeed, patterns of diversification consistent with some of these four models have been found in some Malagasy plant groups (e.g. Andrianoelina et al., 2006; Maurin et al., 2007; Shapcott et al., 2007; Davis & Rakotonasolo, 2008; all cited in Vences et al., 2009), but others can hardly be explained by these alone (e.g. Malcomber, 2002; Janssen et al., 2008; Koopman & Baum, 2008; Anthony et al., 2010). This is not surprising as speciation studies of island endemic plants have often shown the importance of adaptive shifts in response to novel environments and chromosome evolution as the main drivers of radiation (Kim et al., 1996; Stuessy & Ono, 1998; Dunbar-Co et al., 2008; but see also Stuessy et al., 2006). The patterns of variation in morphology and ecology in Noronhia are comparable with those of other prominent island plant radiations (e.g. Plantago L., woody Sonchus L. alliance, Hawaiian silversword alliance), the insights from which can be useful for this study, and other Malagasy plant groups that show considerable variation in morphology and habitat preference with low genetic diversity (e.g. Coffea, Cyathea, Gaertnera, Pandanus). Certainly, the likelihood that fine-scale ecological and evolutionary processes also contributed to the diversification of Noronhia and most components of the Malagasy flora, in addition to the coarse-scale processes invoked by the four models, is very high. A major evolutionary force in plant diversification, having been demonstrated in ancient lineages and in more recently formed species (Soltis & Soltis, 2009; Wood et al., 2009) and allowing sympatric speciation (Rieseberg & Willis, 2007), is polyploidy or wholegenome duplication. Although conclusive data are still unavailable for Noronhia, the presence of multiple nuclear gene copies (Hong-Wa, 2012) can hint at polyploidy in some members of this genus, thus accounting for some of its diversification. It can also explain the distribution patterns of its species, many of which have narrow geographical ranges (see Fig. 3), and can account for the overall lack of phylogenetic resolution, as recently diverged species may have porous boundaries (Kane et al., 2009; and references therein). Moreover, because of the presence of multiple gene copies, polyploids are expected to exhibit less inbreeding depression (Lande & Schemske, 1985), which is often documented in small and isolated populations. Thus, there may have been recurrent polyploidization events in Noronhia, such as those reported in narrow endemics of Olea (Besnard et al., 2008; Besnard & Baali-Cherif, 2009). The pattern of diversity and endemism and the unresolved relationships in Noronhia may also be explained by sympatric speciation with gene flow, as demonstrated in Howea Becc. (Savolainen et al., 2006; Babik et al., 2009) and Coprosma J.R.Forst. & G.Forst. (Papadopulos et al., 2011), both from Lord Howe Island. Many large plant genera in Madagascar (e.g. Dalbergia L.f., Diospyros L., Dombeya Cav., Grewia L. and Memecylon; C. Hong-Wa, pers. observ.) exhibit patterns of morphological and ecological variation similar to those in Noronhia. Despite the lack of molecular data for most of these groups, other exam- © 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161 DIVERSIFICATION OF NORONHIA ples, e.g. Gaertnera, Adenia, Cyathea, Coffea subgenus Coffea and Pandanus (Malcomber, 2002; Hearn, 2006; Janssen et al., 2008; Anthony et al., 2010; Buerki et al., 2012), show that these variations are usually coupled with low genetic diversity, suggesting that the patterns observed in Noronhia are ubiquitous, or at least widespread, among diversified Malagasy plant lineages. Hence, this study provides useful insights into understanding the diversification of these taxa, whilst also emphasizing the need for more operational models aligned to the biology of plants in general. Addressing these gaps and developing more studies on Malagasy plant phylogeography and diversification will help to fulfil the recommendation of Vences et al. (2009) for the establishment of Madagascar as a model region for studies of both plant and animal species diversification. In addition, improving our perception of species boundaries and our knowledge of species distribution will contribute greatly to our assessment of plant diversification in Madagascar (Stone, 2012). So far, efforts to survey the Malagasy flora have provided useful information relevant to conservation and diversity estimates (Kremen et al., 2008; Callmander et al., 2011; Gautier et al., 2012), but have not been efficiently applied to understanding evolutionary history, which thus remains an important avenue for future research. IMPLICATIONS FOR DIVERSITY ESTIMATES Reconciling taxonomy and evolutionary patterns can be challenging, especially when complex patterns of diversification are involved. Indeed, the presence of large polytomies made it difficult to assess the phylogenetic cohesiveness of some Noronhia spp., and convergent morphologies obscured the limits of others. These situations lead to a taxonomic complexity that is at least partly attributable to the mechanisms underlying the diversification of Noronhia. It also affects the estimation of biological diversity. Clearly, considering different types of data independently will produce different outcomes in these instances. Although taxon sampling is still an issue, the integrated analyses of bioclimatic, molecular and morphological data applied in this study provided useful insights into the complex nature of divergence in Noronhia and allowed a better assessment of species boundaries. Population genetic analyses and anatomical and chromosomal data would also provide additional information for better separation of the species of this genus and for unravelling cryptic diversity. This study indicates an almost two-fold increase in species richness in this group since the last taxonomic treatment 60 years ago (Perrier de la Bâthie, 1952). This increase results mainly from many new collections accumulated since then, but also involves some 157 generic rearrangements (Hong-Wa & Besnard, 2013) and rank changes (C. Hong-Wa, unpubl. data). Such changes are not unique to Noronhia and taxonomic studies of Malagasy plants often report a considerable increase in species number. For instance, a study of Memecylon revealed that most of the specimens collected in the last 25 years represented undescribed entities and will incur a 70% increase in the diversity of this genus (Stone, 2012). Similarly, at least a hundred species remain to be described in Diospyros (Ebenaceae), based on currently available herbarium material (G. E. Schatz, pers. comm.), and > 50 new Impatiens spp. have been described between 2003 and 2007, and the taxonomic work is still in progress (see the Catalogue of the Vascular Plants of Madagascar webpage: http://www.tropicos.org/Project/MADA). Species discovery, in animal and plant groups, has increased exponentially in Madagascar over the last decade (see WWF Report, 2011), and will probably continue for years to come as a result of better sampling, improved analytical techniques and more researchers (Scheffers et al., 2012). Unfortunately, gaps in geographical sampling still represent a potential bias, often leading to inflated diversity. This can be addressed by applying an integrative approach, which can be useful in deciphering variation, and thus avoiding taxonomic confusion, in highly diverse groups such as those mentioned above. Although not often used in the assessment of the Malagasy flora, such an approach has already permitted a refinement of species hypotheses in a broad range of animal taxa, such as lemurs, tenrecs and frogs (Olson, Goodman & Yoder, 2004; Yoder et al., 2005; Groeneveld et al., 2009; Vieites et al., 2009). Given the patterns of microendemism in Madagascar, the traditional approach to plant species discovery on this island can particularly benefit from such integrative and analytical assessments to provide robust hypotheses of species, which are a prerequisite for effective conservation. ACKNOWLEDGEMENTS We are grateful to P. F. Stevens, E. A. Kellogg, A. J. Miller, A. E. Zanne, the associate editor and two anonymous reviewers for constructive comments. The herbaria G, MO, P, TAN and TEF made their collections available for morphological study. T. Andriamihajarivo, C. Birkinshaw, R. Letsara, P. Lowry, F. Rakotonasolo, R. Randrianaivo, A. Randrianasolo, F. Ratovoson, S. Razafimandimbison, A. Razanatsima, G. Schatz and S. Trigui provided leaf material for molecular analyses. The US National Science Foundation (NSF DEB 1011208), Garden Club of America, Intra-European fellowship (PIEF-GA-2008-220813) and Harris World Ecology Center at UMSL provided financial support. The Kellogg Laboratory at UMSL, © 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161 158 C. HONG-WA and G. BESNARD EDB Laboratory of the ‘Laboratoire d’Excellence (LABEX)’ TULIP (ANR-10-LABX-41) and CEBA (ANR-10-LABX-0025), Missouri Botanical Garden in St. Louis and its Madagascar Program, Madagascar National Parks, Ministère des Eaux et Forêts de Madagascar, Rio Tinto Madagascar (QMM) and Idea Wild provided administrative, financial, logistic or material support. REFERENCES Aguilée R, Lambert A, Claessen D. 2011. Ecological speciation in dynamic landscapes. Journal of Evolutionary Biology 24: 2663–2677. Andriamaharoa H, Birkinshaw CR, Ludovic R. 2010. Day-time feeding ecology of Eulemur cinereiceps in the Agnalazaha forest, Mahabo-Mananivo, Madagascar. Madagascar Conservation and Development 5: 55–63. Andrianoelina O, Rakotondraoelina H, Ramamonjisoa L, Maley J, Danthu P, Bouvet J-M. 2006. Genetic diversity of Dalbergia monticola (Fabaceae) an endangered tree species in the fragmented oriental forest of Madagascar. Biodiversity and Conservation 15: 1109–1128. Angel F. 1942. Les lézards de Madagascar. Tananarive: Académie Malgache. Anthony F, Diniz LE, Combes M-C, Lashermes P. 2010. Adaptive radiation in Coffea subgenus Coffea L. (Rubiaceae) in Africa and Madagascar. Plant Systematics and Evolution 285: 51–64. Babik W, Butlin RK, Baker WJ, Papadopulos AST, Boulesteix M, Anstett M-C, Lexer C, Hutton I, Savolainen V. 2009. How sympatric is speciation in the Howea palms of Lord Howe Island? Molecular Ecology 18: 3629–3638. Barker FK, Lutzoni FM. 2002. The utility of the incongruence length difference test. Systematic Biology 51: 625–637. Barrett CF, Freudenstein JV. 2011. An integrative approach to delimiting species in a rare but widespread mycoheterotrophic orchid. Molecular Ecology 20: 2771– 2786. Besnard G, Baali-Cherif D. 2009. Coexistence of diploids and triploids in a Saharan relict olive: evidence from nuclear microsatellite and flow cytometry analyses. Comptes Rendus – Biologies 332: 1115–1120. Besnard G, Christin PA, Malé PJG, Coissac E, Ralimanana H, Vorontsova M. 2013. Phylogenomics and taxonomy of Lecomtelleae (Poaceae), an isolated Panicoid lineage from Madagascar. Annals of Botany 112: 1057–1066. In press. Besnard G, García-Verdugo C, Rubio de Casas R, Treier UA, Galland N, Vargas P. 2008. Polyploidy in the olive complex (Olea europaea L.): evidence from flow cytometry and nuclear microsatellite analyses. Annals of Botany 101: 25–30. Birkinshaw CR. 1999. The importance of the black lemur (Eulemur macaco) for seed dispersal in Lokobe forest, Nosy Be. In: Rakotosaminana B, Rasamimanana H, Ganzhorn JU, Goodman SM, eds. New directions in lemur studies. New York: Kluwer Academic Publishers, 189–199. Birkinshaw CR. 2001. Fruit characteristics of species dispersed by the black lemur (Eulemur macaco) in the Lokobe forest, Madagascar. Biotropica 33: 478–486. Buerki S, Callmander MW, Devey DS, Chappell L, Gallaher T, Munzinger J, Haevermans T, Forest F. 2012. Straightening out the screw-pines: a first step in understanding phylogenetic relationships within Pandanaceae. Taxon 61: 1010–1020. Buerki S, Devey DS, Callmander MW, Phillipson PB, Forest F. 2013. Spatio-temporal history of the endemic genera of Madagascar. Botanical Journal of the Linnean Society 171: 304–329. Callmander MW, Phillipson PB, Andriambololonera S, Rabarimanarivo M, Rakotonirina N, Raharimampionona J, Chatelain C, Gautier L, Lowry II PP. 2011. The endemic and non-endemic vascular flora of Madagascar updated. Plant Ecology and Evolution 144: 121–125. Cornet A. 1974. Notice explicative no. 55. Essai de cartographie bioclimatique à Madagascar. Paris: ORSTOM. Darlu P, Lecointre G. 2002. When does the incongruence length difference test fail? Molecular Biology and Evolution 19: 432–437. Davis AP, Rakotonasolo F. 2008. A taxonomic revision of the Baracoffea alliance: nine remarkable Coffea species from western Madagascar. Botanical Journal of the Linnean Society 171: 355–390. Davis AP, Tosh J, Ruch N, Fay MF. 2011. Growing coffee: Psilanthus (Rubiaceae) subsumed on the basis of molecular and morphological data; implications for the size, morphology, distribution and evolutionary history of Coffea. Botanical Journal of the Linnean Society 167: 357–377. Dayrat B. 2005. Towards integrative taxonomy. Biological Journal of the Linnean Society 85: 407–415. Dewar RE, Richards AF. 2007. Evolution in the hypervariable environment of Madagascar. Proceedings of the National Academy of Sciences of the United States of America 104: 13 723–13 727. Donati G, Lunardini A, Kappeler PM. 1999. Cathemeral activity of red-fronted brown lemurs (Eulemur fulvus rufus) in the Kirindy forest/CFPF. In: Rakotosaminana B, Rasamimanana H, Ganzhorn JU, Goodman SM, eds. New directions in lemur studies. New York: Kluwer Academic Publishers, 119–137. Dray S, Dufour AB. 2007. The ade4 package: implementing the duality diagram for ecologists. Journal of Statistical Software 22: 1–20. Duminil J, Di Michele M. 2009. Plant species delimitation: a comparison of morphological and molecular markers. Plant Biosystems 143: 528–542. Dunbar-Co S, Wieczorek A, Morden CW. 2008. Molecular phylogeny and adaptive radiation of the endemic Hawaiian Plantago species (Plantaginaceae). American Journal of Botany 95: 1177–1188. Farris JS, Källersjö M, Kluge AG, Bult C. 1995. Constructing a significance test for incongruence. Systematic Biology 44: 570–572. © 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161 DIVERSIFICATION OF NORONHIA Gautier L, Chatelain C, Callmander MW, Phillipson PB. 2012. Richness, similarity and specificity of Madagascar flora compared with sub-Saharan Africa. Plant Ecology and Evolution 145: 55–64. Goodman SM, Benstead JP. 2003. The natural history of Madagascar. Chicago, IL: The University of Chicago Press. Gotelli NJ, Ellison AM. 2004. A primer of ecological statistics. Sunderland, MA: Sinauer Associates. Groeneveld LF, Weisrock DW, Rasoloarison RM, Yoder AD, Kappeler PM. 2009. Species delimitation in lemurs: multiple genetic loci reveal low levels of species diversity in the genus Cheirogaleus. BMC Evolutionary Biology 9: 30. Hearn DJ. 2006. Adenia (Passifloraceae) and its adaptive radiation: phylogeny and growth form diversification. Systematic Botany 31: 805–821. Hijmans RJ, Cameron SE, Parra JL, Jones PG, Jarvis A. 2005. Very high resolution interpolated climate surfaces for global land areas. International Journal of Climatology 25: 1965–1978. Holmgren PK, Holmgren NH, Barnett LC. 1990. Index herbariorum. Part 1: the herbaria of the world. The Bronx: International Association for Plant Taxonomy. Hong-Wa C. 2012. Diversification and coexistence in the Madagascar olive (Noronhia, Oleaceae). PhD Dissertation, University of Missouri, St. Louis, MO. Hong-Wa C, Arroyo TPF. 2012. Climate-induced range contraction in the Malagasy endemic plant genera Mediusella and Xerochlamys (Sarcolaenaceae). Plant Ecology and Evolution 145: 302–312. Hong-Wa C, Besnard G. 2013. Intricate patterns of phylogenetic relationships in the olive family as inferred from multi-locus plastid and nuclear DNA sequence analyses: a close-up on Chionanthus and Noronhia (Oleaceae). Molecular Phylogenetics and Evolution 67: 367–378. Humbert H. 1955. Les territoires phytogéographiques de Madagascar. Année Biologique 31: 439–448. Janssen T, Bystriakova N, Rakotondrainibe F, Coomes D, Labat J-N, Schneider H. 2008. Neoendemism in Madagascar scaly tree ferns results from recent, coincident diversification bursts. Evolution 62: 1876–1889. Kane NC, King MG, Barker MS, Raduski A, Karrenberg S, Yatabe Y, Knapp SJ, Rieseberg LH. 2009. Comparative genomic and population genetic analyses indicate highly porous genomes and high levels of gene flow between divergent Helianthus species. Evolution 63: 2061–2075. Kim S-G, Crawford DJ, Franscico-Ortega J, SantosGuerra A. 1996. A common origin for woody Sonchus and five related genera in the Macaronesian islands: molecular evidence for extensive radiation. Proceedings of the National Academy of Sciences of the United States of America 93: 7743–7748. Koopman M, Baum D. 2008. Phylogeny and biogeography of tribe Hibisceae (Malvaceae) on Madagascar. Systematic Botany 33: 364–374. Kremen C, Cameron A, Moilanen A, Phillips SJ, Thomas CD, Beentje H, Dransfield J, Fisher BL, Glaw F, Good TC, Harper GJ, Hijmans RJ, Lees DC, Louis JED, Nussbaun RA, Raxworthy CJ, Razafimpahanana A, 159 Schatz GE, Vences M, Vieites DR, Wright PC, Zjhra ML. 2008. Aligning conservation priorities across taxa in Madagascar with high-resolution planning tools. Science 320: 222–226. Kuntner M, Agnarsson I. 2011. Biogeography and diversification of hermit spiders on Indian Ocean islands (Nephilidae: Nephilengys). Molecular Phylogenetics and Evolution 59: 477–488. Lande R, Schemske DW. 1985. The evolution of selffertilization and inbreeding depression in plants. I. Genetic models. Evolution 39: 24–40. Malcomber S. 2002. Phylogeny of Gaertnera Lam. (Rubiaceae) based on multiple DNA markers: evidence of a rapid radiation in a widespread, morphologically diverse genus. Evolution 56: 42–57. Martin RD. 1972. Adaptive radiation and behavior of the Malagasy lemurs. Philosophical Transactions of the Royal Society 264: 295–352. Martinez BT. 2010. Forest restoration in Masoala National Park, Madagascar: the contribution of red-ruffed lemur (Varecia rubra) and the livelihoods of subsistence farmers at Ambatoladama. PhD Dissertation, University of Minnesota, St. Paul, MN. Maurin O, Davis AP, Chester M, Mvungi EF, Jaufeerally Y, Fay MF. 2007. Towards a phylogeny for Coffea (Rubiaceae): identifying well-supported lineages based on nuclear and plastid DNA sequences. Annals of Botany 100: 1565– 1583. Meudt HM, Lockhart PJ, Bryant D. 2009. Species delimitation and phylogeny of a New Zealand plant species radiation. BMC Evolutionary Biology 9: 111. Miller MA, Pfeiffer W, Schwartz T. 2010. Creating the CIPRES Science Gateway for inference of large phylogenetic trees. Proceedings of the Gateway Computing Environments Workshop (GCE), 14 November 2010, New Orleans, LA, 1–8. IEEE. Nylander JAA, Wilgenbusch JC, Warren DL, Swofford DL. 2008. AWTY (are we there yet?): a system for graphical exploration of MCMC convergence in Bayesian phylogenetics. Bioinformatics Applications Note 24: 581–583. Olson L, Goodman SM, Yoder AD. 2004. Illumination of cryptic species boundaries in long-tailed shrew-tenrecs (Mammalia: Tenrecidae; Microgale), with new insights into geographic variation and distributional constraints. Biological Journal of the Linnean Society 83: 1–22. Padial JM, De la Riva I. 2010. A response to recent proposals for integrative taxonomy. Biological Journal of the Linnean Society 101: 747–756. Padial JM, Miralles A, De la Riva I, Vences M. 2010. The integrative future of taxonomy. Frontiers in Zoology 7: 16. Papadopulos AST, Baker WJ, Crayn D, Butlin RK, Kynast RG, Hutton I, Savolainen V. 2011. Speciation with gene flow on Lord Howe Island. Proceedings of the National Academy of Sciences of the United States of America 108: 13 188–13 193. Paradis E, Claude J, Strimmer K. 2004. APE: analyses of phylogenetics and evolution in R language. Bioinformatics 20: 289–290. © 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161 160 C. HONG-WA and G. BESNARD Pearson RG, Raxworthy CJ. 2009. The evolution of local endemism in Madagascar: watersheds versus climatic gradient hypotheses evaluated by null biogeographic models. Evolution 63: 959–967. Perrier de la Bâthie H. 1949. Révision des Oléacées de Madagascar et des Comores. Mémoire de l’Institut Scientifique de Madagascar, Série B 2: 275–310. Perrier de la Bâthie H. 1951. Notes biologiques sur les Oléacées de Madagascar et des Comores. Mémoires de l’Institut Scientifique de Madagascar, Série B 3: 175–185. Perrier de la Bâthie H. 1952. Oléacées. In: Humbert H, ed. Flore de Madagascar et des Comores. 166e famille. Paris: Muséum National d’Histoire Naturelle, 1–89. de Queiroz K. 2007. Species concepts and species delimitation. Systematic Biology 56: 879–886. Quéméré E, Crouau-Roy B, Rabarivola C, Louis EE, Chikhi L. 2010. Landscape genetics of an endangered lemur (Propithecus tattersalli) within its entire fragmented range. Molecular Ecology 19: 1606–1621. Quinn GP, Keough MJ. 2002. Experimental design and data analysis for biologists. Cambridge: Cambridge University Press. R Development Core Team. 2012. R: a language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. ISBN 3-900051-07-0. Available at: http://www.R-project.org Radespiel U. 2007. Ecological diversity and seasonal adaptations of mouse lemurs (Microcebus spp.). In: Gould L, Sauther ML, eds. Lemurs: ecology and adaptation (developments in primatology: progress and prospects). New York: Springer, 211–234. Rakotoarisoa JE, Raheriarisena M, Goodman SM. 2010. Phylogeny and species boundaries of the endemic species complex, Eliurus antsingy and E. carletoni (Rodentia: Muroidea: Nesomyidae), in Madagascar using mitochondrial and nuclear DNA sequence data. Molecular Phylogenetics and Evolution 57: 11–22. Raxworthy CJ, Nussbaum RA. 1995. Systematics, speciation and biogeography of the dwarf chameleons (Brookesia; Reptilia, Squamata, Chamaelontidae) of northern Madagascar. Journal of Zoology 235: 525–558. Reddy S, Driskell A, Rabosky DL, Hackett SJ, Schlenberg TS. 2012. Diversification and the adaptive radiation of the vangas of Madagascar. Proceedings of the Royal Society of London B 279: 2062–2071. Rieseberg LH, Willis JH. 2007. Plant speciation. Science 317: 910–914. Rieseberg LH, Wood TE, Baack EJ. 2006. The nature of plant species. Nature 440: 524–527. Ronquist F, Huelsenbeck JP. 2003. MrBayes 3: Bayesian phylogenetic inference under mixed models. Bioinformatics Applications Note 19: 1572–1574. Russell AL, Goodman SM, Cox MP. 2008. Coalescent analyses support multiple mainland-to-island dispersals in the evolution of Malagasy Triaenops bats (Chiroptera: Hipposideridae). Journal of Biogeography 35: 995–1003. Savolainen V, Anstett M-C, Lexer C, Hutton I, Clarkson JJ, Norup MV, Powell MP, Springate D, Salamin N, Baker WJ. 2006. Sympatric speciation in palms on an oceanic island. Nature 441: 210–213. Schatz GE. 2000. Endemism in the Malagasy tree flora. In: Lourenço WR, Goodman SM, eds. Diversity and endemism in Madagascar. Paris, ORSTOM: Mémoires de la Société de Biogéographie, Société de Biogéographie, 1–9. Schatz GE, Birkinshaw C, Lowry PP II, Randriatafika F, Ratovoson F. 2000. The endemic plant families of Madagascar project: integrating taxonomy and conservation. In: Lourenço WR, Goodman SM, eds. Diversity and endemism in Madagascar. Paris, ORSTOM: Mémoires de la Société de Biogéographie, Société de Biogéographie, 11– 24. Scheffers BR, Joppa LN, Pimm SL, Laurance WF. 2012. What we know and don’t know about Earth’s missing biodiversity. Trends in Ecology and Evolution 27: 501–510. Schluter D. 2000. The ecology of adaptive radiation. Oxford: Oxford University Press. Seelanan T, Schnabel A, Wendel JF. 1997. Congruence and consensus in the cotton tribe (Malvaceae). Systematic Biology 22: 259–290. Shapcott A, Rakotoarinivo M, Smith RJ, Lysakova G, Fay MF, Dransfield J. 2007. Can we bring Madagascar’s critically endangered palms back from the brink? Genetics, ecology and conservation of the critically endangered palm Beccariophoenix madagascariensis. Botanical Journal of the Linnean Society 154: 589–608. Simmen B, Sauther ML, Soma T, Rasamimanana H, Sussman RW, Jolly A, Tarnaud L, Hladik A. 2006. Plant species fed on by Lemur catta in gallery forests of the southern domain of Madagascar. In: Jolly A, Sussman RW, Koyama N, Rasamimanana H, eds. Ringtailed lemur biology: Lemur catta in Madagascar (developments in primatology: progress and prospects). New York: Springer, 55–68. Sites JW, Marshall JC. 2003. Delimiting species: a Renaissance issue in systematic biology. Trends in Ecology and Evolution 18: 462–470. Soltis PS, Soltis DE. 2009. The role of hybridization in plant speciation. Annual Review of Plant Biology 60: 561–580. Stamatakis A. 2006. RAxML-VI-HPC: maximum likelihoodbased phylogenetic analyses with thousands of taxa and mixed models. Bioinformatics Applications Note 22: 2688– 2690. Stone RD. 2012. Endemism, species richness and morphological trends in Madagascan Memecylon (Melastomataceae). Plant Ecology and Evolution 145: 145–151. Stuessy TF, Jakubowsky G, Gómez RS, Pfosser M, Schlüter PM, Fer T, Sun B-Y KH. 2006. Anagenetic evolution in island plants. Journal of Biogeography 33: 1259–1265. Stuessy TF, Ono M. 1998. Evolution and speciation of island plants. Cambridge: Cambridge University Press. Swofford DL. 2002. PAUP*. Phylogenetic Analysis Using Parsimony (* and other methods). Version 4. Sunderland, MA: Sinauer Associates. Takhtajan A. 1986. Floristic regions of the world. Berkeley, CA: University of California Press. © 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161 DIVERSIFICATION OF NORONHIA Thorén SIK. 2011. Comparative feeding ecology of two sympatric mouse lemurs (Microcebus spp.) in northwestern Madagascar. PhD Dissertation, University of Veterinary Medicine, Hannover. Vences M. 2005. Madagascar as a model region for the study of tempo and pattern in adaptive radiations. In: Huber BA, Lampe KH, eds. African biodiversity: molecules, organisms, ecosystems. Proceedings of the 5th International Symposium on Tropical Biology, Museum Koenig. Bonn: Springer, 69–84. Vences M, Wollenberg KC, Vieites DR, Lees DC. 2009. Madagascar as a model region of species diversification. Trends in Ecology and Evolution 24: 456–465. Vieites DR, Wollenberg KC, Andreone F, Köhler J, Glaw F, Vences M. 2009. Vast underestimation of Madagascar’s biodiversity evidenced by an integrative amphibian inventory. Proceedings of the National Academy of Sciences of the United States of America 106: 8267–8272. Warren BH, Strasberg D, Bruggemann J, Prys-Jones RP, Thébaud C. 2010. Why does the biota of Madagascar region have such a strong Asiatic flavor? Cladistics 26: 526–538. Warren DL, Siefert SN. 2011. Ecological niche modeling in Maxent: the importance of model complexity and the performance of model selection criteria. Ecological Applications 21: 335–342. Wendel JF, Doyle JJ. 1998. Phylogenetic incongruence: window into genome history and molecular evolution. In: Soltis DE, Soltis PS, Doyle JJ, eds. Molecular systematics of plants II: DNA sequencing. Norwell, MA: Kluwer Academic Publishers, 265–296. Weyeneth N, Goodman SM, Ruedi M. 2011. Do diversification models of Madagascar’s biota explain the population structure of the endemic bat Myotis goudoti (Chiroptera: Vespertilionidae)? Journal of Biogeography 38: 44–54. Wilmé L, Goodman SM, Ganzhorn JU. 2006. Biogeographic evolution of Madagascar’s microendemic biota. Science 312: 1063–1065. 161 de Wit MJ. 2003. Madagascar: heads it’s a continent, tails it’s an island. Annual Review of Earth and Planetary Science 31: 213–248. Wood TE, Takebayashi N, Barker MS, Mayrose I, Greenspoon PB, Rieseberg L. 2009. The frequency of polyploid speciation in vascular plants. Proceedings of the National Academy of Sciences of the United States of America 106: 13 875–13 879. WWF Report. 2011. Treasure Island: new biodiversity on Madagascar (1999–2010). Available at: http://assets.wwf .org.uk/downloads/madagascarspeciesreport.pdf Yoder AD, Heckman KL. 2006. Mouse lemur phylogeography revises a model of ecogeographic constraint in Madagascar. In: Fleagle J, Lehman SM, eds. Primate biogeography: progress and prospects. New York: Kluwer Press, 255–268. Yoder AD, Irwin JA, Payseur BA. 2001. Failure of the ILD to determine data combinability for slow loris phylogeny. Systematic Biology 50: 408–424. Yoder AD, Nowak MD. 2006. Has vicariance or dispersal been the predominant biogeographic force in Madagascar? Only time will tell. Annual Review of Ecology Evolution and Systematics 37: 405–431. Yoder AD, Olson LE, Hanley C, Heckman KL, Rasoloarison R, Russell AL, Ranivo J, Soarimalala V, Karanth KP, Raselimanana AP, Goodman SM. 2005. A multidimensional approach for detecting species patterns in Malagasy vertebrates. Proceedings of the National Academy of Sciences of the United States of America 102: 6587–6594. Zapata F. 2010. Phylogenetics and diversification of Escallonia (Escalloniaceae). PhD Dissertation, University of Missouri, St. Louis, MO. Zhu G, Bu W, Gao Y, Liu G. 2012. Potential geographic distribution of brown marmorated stink bug invasion (Halyomorpha halys). PLoS ONE 7: e31246. SUPPORTING INFORMATION Additional Supporting Information may be found in the online version of this article at the publisher’s web-site: Figure S1. Biogeographical zones recognized under the ecogeographical constraint (ECH), riverine barrier (RBH), watershed contraction (WCH) and current climate (CCH) hypotheses mapped onto the phylogenetic tree. Figures S2–S31. Results of principal component analyses of bioclimatic and morphological data carried out in phylogenetic (Figs S2–S17) and geographical (Figs S18–S31) contexts. BIO, FL, FR, VEG, VEG + FL and VEG + FR are bioclimatic, flower, fruit, vegetative, vegetative and flower, and vegetative and fruit variables, respectively. Number (2) after an abbreviation refers to a second analysis with the same set of variables, but excluding the species indicated in brackets. Table S1. List of qualitative characters used to separate specimens into groups. Table S2. List of operational taxonomic units (OTUs) included in this study with corresponding GenBank accession numbers for the different loci. Table S3. Statistics of the alignment. Table S4. List of quantitative characters measured for morphometric analyses. Table S5. Elevation and bioclimatic variables used in the principal component analyses and discriminant function analyses. © 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161