Clarifying the identity of the Cleistes rosea complex (Orchidaceae) based on integrative taxonomy

Cleistes rosea and 10 other taxa comprise a complex widely distributed in various open vegetation types from eastern South America to Panama. The taxonomic identity of most of these taxa is unclear and their relationships with C. rosea have never been satisfactorily discussed or tested. Here, we employ molecular phylogenetics, morphometrics and morphological, ecological and geographic distribution data to investigate the relationships and taxonomy of Cleistes rosea and related species. Our results confirm that Cleistes rosea and Cleistes castaneoides are sister species and although highly similar in general floral morphology, they differ in vegetative characters, distribution, habitat, floral scent, and lip characters. Both species show local variations regarding floral characters, particularly the flower colour, and such differences were used here to delimit infraspecific taxa within the C. rosea/C. castaneoides clade. While C. rosea var. guianensis var. nov. is restricted to French Guiana, and C. rosea var. buenaventurae stat. & comb. nov. is restricted to low altitudes on the western side of the Cordillera Occidental in the Chocó region of Colombia, C. castaneoides var. castaneoides and C. castaneoides var. augusta comb. nov. are sympatric throughout their distribution. Cleistes abdita, formerly known as C. rosea f. pallida, is similar to C. castaneoides but distinguished by the smaller flowers and restricted distribution to Amazonian savannas. Cleistes rosea is distributed throughout the north of the Andes and northern South America. In turn, C. castaneoides occurs mainly in South-eastern and Central-western Brazil. Still, the distribution of these three species overlaps in the Guiana Shield, where the differentiation between them is less evident. Our results provide evidence of the plasticity of the floral morphology of Cleistes, especially the lip, and highlight that hitherto neglected characters, such as vegetative morphology, habitat, and floral scent, are particularly informative for the characterization of species and taxonomy of the genus.


Introduction
Cleistes Rich.ex Lindl. is a primarily South American genus of Orchidaceae with few species extending up to Central America (i.e., Costa Rica).Brazil is the country with the greatest species diversity of Cleistes (40 species), followed by Colombia (15 spp.) and Venezuela (11 spp.) (Govaerts et al., 2022).The name is derived from the Greek word kleistos (i.e., closed), probably in reference to the flowers that seem not to fully open.However, in many species of this genus, the flowers are completely open but short-lived and appear to be closed most of the time.There are 95 names published under Cleistes, of which 58 are currently accepted according to the World Checklist of Selected Plant Families (Govaerts et al., 2022).However, the identity of many of them is still not understood, and the genus is in need of a thorough revision (Pansarin pers. observ.).
Correspondence to: João A. N. Batista.E-mail: janb@ icb.ufmg.br the regions rps16, trnL-F, rbcL and psaB of the chloroplast DNA.The Maximum likelihood (ML) tree of each marker analysed separately along with bootstrap support (BS) values are provided in Supplemental Appendix 2. The number of characters and substitution models from each DNA region included in the combined data matrix is given in Supplemental Appendix 3.
Maximum parsimony analysis (MP) was run with PAUP Ã version 4.0b5 (Swofford, 2001) software.A heuristic search was conducted using 1000 replicates of random taxon entry additions, the MULTREES option, and the tree bisection-reconnection (TBR) swapping algorithm, holding 10 trees per replicate and saving all the shortest trees.Support for clades was assessed using 3000 bootstrap replicates (Felsenstein, 1985).For bootstrap support levels, we considered bootstrap percentages of 50-70% as weak, 71-85% as moderate and >85% as strong (Kress et al., 2002).
Maximum likelihood (ML) analysis was run using the MEGA X tool (Kumar et al., 2018) with bootstrapping for 3000 replicates.The analysis was based on the Kimura 2-parameter model (Kimura, 1980).Tests with other models resulted in trees with the same topology.Initial trees for the heuristic search were obtained automatically by applying Neighbour-joining and BioNJ algorithms to a matrix of pairwise distances estimated using the Maximum Composite Likelihood (MCL) approach and then selecting the topology with the superior log likelihood value.
Bayesian inference (BI) was conducted using the MrBayes program, version 3.1 (Ronquist & Huelsenbeck, 2003).The optimal model of sequence evolution for each partition was selected using the software jModelTest (Posada, 2008) and Bayesian Information Criterion (BIC).The evolution models selected for each of the regions are given in Supplemental Appendix 3. Four Markov chains were run simultaneously for three million generations, with parameters sampled every 100 generations.The consensus tree was calculated after the removal of the first 3000 trees, which were considered to be burn-in.Because Posterior probability (PP) values in Bayesian analysis are not equivalent to BS but are generally much higher (Erixon et al., 2003), we used criteria similar to a standard statistical test, considering groups with PP >0.95 as strongly supported, groups with PP ranging from 0.90-0.95as moderately supported, and groups with PP <0.90 as weakly supported.

Taxon sampling and selection of characters for morphometric analyses
We performed morphometric analyses to test for discrete clusters that could support or refute current taxon delimitation within the species complex.We analysed 192 individuals belonging to eight taxa related to the Cleistes rosea complex (Cleistes abdita, C. ayangannensis, C. castaneoides var.castaneoides, C. castaneoides var.augusta, C. rosea var.rosea, C. rosea var.buenaventurae, C. rosea var.guianensis and C. tamboana) (Table 1).All individuals were examined from exsiccates, mostly from images.We included only complete individuals, and selected them in order to sample the ecological, geographic, morphological and taxonomic diversity of the taxa.The type materials of the taxa were included in the analyses whenever they were available and had complete specimens.The full list of individuals analysed is given in Supplemental Appendix 4. Based on the morphological variation observed in the material examined, we selected and measured 10 continuous and six categorical morphological characters, of which 10 are vegetative and six floral (Table 2).For some characters not normally present on herbarium labels, such as the colour of fresh leaves and the scent of flowers, we followed the species standard characterized by examination of live plants, photographs or the labels with the available data.We standardized the flower measured as the most basal flower on the inflorescence at anthesis, but when it was not present the next available flower was measured.Missing data for quantitative variables (0.68%) were filled in with the average of the data for each taxa.Only the character 'flower fragrance' (Ffl) for C. tamboana remained with missing data because this information was missing for the taxon.

Multivariate analyses
All analyses were performed in R and Rstudio using version 4.1.0in both (R Core Team, 2021;RStudio Team, 2021).To explore the structure of the data and detect possible outliers we performed a Principal component analysis (PCA), based on a correlation matrix of the 10 numeric variables and the 192 individuals, using the 'prcomp' function of the 'vegan' package (Oksanen et al., 2020).We performed a Principal coordinate analysis (PCoA) including 16 variables and 192 individuals (referred to as 'full dataset' hereafter).This method, in a similar way to other ordination analyses, such as PCA, aims to reduce the complexity of a dataset and identify its main patterns, but allows using any multivariate distance measure.First, we calculated a dissimilarity matrix using the Gower (1971) distance, which allows mixed data to be analysed, using the 'daisy' function of the 'cluster' package (Maechler et al., 2021).We implemented PCoA with the 'ape' package (Paradis & Schliep, 2019) from the square root of the dissimilarity matrix to avoid negative eigenvalues (Gower & Legendre, 1986;Legendre & Legendre, 1998).Given that the PCoA uses a distance matrix, inferring the contribution of the original variables to the observed patterns on the ordination axes is difficult, so to have an approximation, we calculated Pearson's correlation of the original variables with the object scores of the principal axis (Legendre & Legendre, 1998;Ramette, 2007).For the variable 'flower fragrance' (Ffl) this correlation was calculated for 187 individuals due to unavailable data for C. tamboana.To analyse finer patterns in C. rosea and C. castaneoides we performed a PCoA with a reduced dataset (referred to as 'reduced dataset' hereafter), in which we excluded C. ayangannensis and C. tamboana, which were the most distinct taxa according to the analysis of the full dataset.With the exclusion of C. ayangannensis the character 'fresh leaf colour' (Cfl) was invariant, so this variable was not included in the analysis.The reduced dataset matrix, therefore, consisted of 15 variables and 170 individuals.

Morphological, habitat and distribution data
In addition to the morphometric analyses, we performed a general comparison of the morphology, habitat preferences and distribution of C. rosea and C. castaneoides to analyse characters other than those used in morphometrics that could be used for taxon characterization and delimitation within the species complex.The habit and perianth were examined and illustrated from liquid-preserved or dry herbarium material.Details of floral morphology were examined using a stereoscopic microscope and measured using a digital calliper.Data concerning flowering periods, floral scent, habitat and distribution were based on herbarium labels and field observations.Morphological terminology is based on Stearn (1992) and Simpson (2006).A total of 425 specimens were examined from the following herbaria: AMES, BHCB, CAY, CEN, CEPEC, COAH, COL, DIAM, ESA, F, GH, HB, HEPH, HJBG, HRB, HUA, HUEFS, HUFU, IAN, IBGE, INPA, K, MBM, MO, MPU, NY, P, R, RB, RENZ, S, SEL, SP, SPF, TRIN, U, UB, UEC, UFP and US; acronyms following Thiers (2022).A list of species and total number of herbarium collections examined is provided in Table 1 and a list of exsiccatae given in Supplemental Appendix 5.

Distribution map
Occurrence records were obtained from herbaria and the iNaturalist website (2022).All records included had their identification revised according to the taxonomic circumscriptions adopted in this work (see taxonomic treatment).Latitude and longitude for the specimens without coordinates were inferred using Google Earth, and the distribution map was elaborated with ArcMap 10.5 (ESRI, 2016).The number of records for each taxa is given in Table 1 and a list of the selected records used for the distribution map is given in Supplemental Appendix 6.

Phylogenetic analyses
The trees obtained from a ML analysis of each marker did not present significant strongly supported topological incongruences (Supplemental Appendix 2), so we used a combined data matrix with all markers for the full phylogenetic inferences.The phylogenetic inferences (MP, ML and BI) obtained from the analyses of the combined data matrix resulted in trees with congruent topologies (ML is shown, Fig. 1).In all analyses (ML, MP and BI), the clade C. tenuis (Rchb.f.ex Griseb.)Schltr./C.rosea/C.castaneoides was recovered as sister to the remaining Cleistes (Fig. 1).Within this clade, C. tenuis emerged as sister to the C. rosea/C.castaneoides clade with strong support (PP 1.00, BS/MP 75%, BS/ML 98%).Within the C. rosea/C.castaneoides clade, C. rosea emerged as sister to a strongly supported clade (PP 1.00, BS/MP 100%, BS/ML 100%) that included all varieties of C. castaneoides.Sequentially, the clade including C. libonii (Rchb.f.) Schltr.and C. metallina (Barb.Rodr.)Schltr.(PP 1.00, BS/MP 100%, BS/ML 100%) was recovered as sister to the clade including the remaining species of the genus, with strong support (PP 1.00, BS/MP 100%, BS/ML 99%

Morphometric analyses
No outliers were found in the PCA analysis (data not shown).The PCoA of the full dataset showed a similar pattern to the results for the reduced dataset, with 5 and 3 eigenvalues greater than 1.0 explaining 71.09 and 63.85% of the total variance, respectively (Table 3, Fig. 2A-C).Cleistes rosea and C. castaneoides samples are separated on the first axis of both analyses (35.09% and 41.30%, respectively) (Fig. 2A, C).Axis 1 of the full dataset analysis also shows the separation of C. tamboana from C. castaneoides and C. abdita, as well as the marginal overlap of C. ayangannensis with the latter two taxa (Fig. 2A).The variables most correlated with axis 1 of both analyses are leaf base (Ble), flower fragrance (Ffl), leaf vernation (Vle), dry leaf colour (Cdl) and leaf length (Lle) (Table 3).Leaf width (Wle) and leaf number (Nle) also had a high correlation with axis 1 of the reduced dataset analysis.On axis 2 of the full dataset analysis (14.01%, Fig. 2A) C. ayangannensis, C. tamboana and C. rosea are separated from each other and there is a complete separation of C. ayangannensis from C. castaneoides and C. abdita, with which it showed overlap on axis 1.The variables with the highest correlation with this axis are fresh leaf colour (Cfl) and length of dorsal sepal (Lse) and lip (Lli).Axis 3 of the full dataset analysis (10.68%, Fig. 2B) shows, similarly to axis 2 of the reduced dataset (12.57%,Fig. 2C), the separation of C. rosea var.rosea from C. rosea var.guianensis and C. rosea var.buenaventurae, and further distinguishes C. castaneoides var.augusta and C. abdita from C. castaneoides var.castaneoides.Sepal colour (Cse) was the variable most correlated with these axes in both analyses.The remaining axes distinguished taxa that were already separated on the axes described, except C. rosea var.buenaventurae from C. rosea var.guianensis and C. abdita from C. castaneoides var.augusta, which continue to overlap.

Comparison of C. rosea and C. castaneoides general morphology
The general floral morphology is quite similar between C. rosea and C. castaneoides (Figs 3E-I, 8D-G) since both species have flowers of approximately the same size, similar colouring and a simple lip (Figs 4F-H, 9B, 9E-F), and after drying, there are no evident distinguishing characteristics to separate each other.In both species, the lip morphology is quite variable and ranges from elliptic, broadly elliptic, oblong to obovate, whereas the apex varies from acute, subobtuse to acuminate (Figs 4F-H, 9B, 9E-F).
Regarding vegetative morphology, C. rosea is quite variable in leaf outline, which range from elliptic, oblong or lanceolate to narrowly lanceolate (Fig. 4A-D).Nonetheless, the leaves are invariably decurrent (Fig. 4E), concave (Fig. 3D), subpatent to patent and the same size as or larger than the floral bracts (Figs 4A-D).Additionally, C. rosea tends to turn straw-coloured after drying (Fig. 4A-D).On the other hand, in C. castaneoides, the leaves are truncate, not decurrent, conduplicate, convolute, clasping the stem, and shorter than or up to the same length as the floral bracts (Figs 8C, 9A, D, H).Lastly, C. castaneoides usually becomes darker in colour after drying (Fig. 9D, H).
There are also some differences in flower colour.In C. rosea, the purple to dark purple stripes at the apex of the lip are wider and more conspicuous, and the background is purple to purplish-pink only at the apex, where the stripes occur, while the rest of the lip is white (Fig. 3E-I).In C. castaneoides var.castaneoides, the lip is white at the base, becoming intense magenta or purplishpink at the apex, and the stripes in a darker and more intense magenta or purplish-pink, with the contrast between the stripes and the background being more discreet (Figs 8D-G, 9B).Another distinguishing feature is the colour of the anthers, which are completely white in C. rosea (Fig. 3E, G-I

Comparison of C. rosea and C. castaneoides habitats
Cleistes rosea is quite tolerant regarding habitat and grows under a wide range of conditions, from sea level up to 2300 m, under partial shade to full sun, and in sandy or clayey, dry or wet soils.Habitats include sandstone or granite outcrops, grassy hillsides, grassy moist slopes, damp Curatella americana L. savannas, Brocchinia Schult.& Schult.f./Vellozia Vand.islands in savannas, white sand savannas, montane savannas, creekbanks on riverine forests, scrub forests, forest clearings, forest edges, seep and bog vegetation, and Paramo peat bogs.The species also grows opportunistically, often in ravines and along roadsides (Fig. 3C), clay banks, pastures, disturbed forests, or other anthropic habitats.On the other hand, Cleistes castaneoides and C. abdita are paludal species, growing in sandy-clay or sandy soils, in hydromorphic, permanently wet soils and are invariably associated with wet grasslands (Fig. 8A, L).

Comparison of C. rosea and C. castaneoides distribution
As delimited here, C. rosea is distributed along the Andes from Peru, Ecuador, Colombia and Venezuela, extending to Panama, Trinidad, the Amazon region, the Guianas and Northern Brazil (Amap a, Amazonas, Par a, Roraima) (Fig. 10).Cleistes rosea has also been recorded for Costa Rica and Bolivia (Carnevali & Ram ırez-Morillo, 2003;Garay, 1978;Szlachetko et al., 2020;Williams, 1956).Although the occurrence is possible, we did not find any sample of the species in these countries and the only herbarium samples identified as C. rosea that we examined so far from Bolivia (White 2308 -AMES) and Costa Rica (Luer et al.Cleistes castaneoides occurs in Brazil, Bolivia, Peru, Colombia (in the savanna zone or areas of the Guayan es shield formations) and Venezuela (Fig. 10).In Brazil, it occurs in the Cerrado, Amazon Forest and Atlantic Forest biomes, distributed in the States of Amazonas, Bahia, Goi as, Maranhão, Mato Grosso, Minas Gerais, Par a, Paran a, Roraima, São Paulo and the Distrito Federal.Most of the records are from the Cerrado, in the States of Goi as, Minas Gerais and Distrito Federal.In turn, C. abdita occurs in the Venezuelan (Amazonas, Bolivar), Brazilian (Roraima), Colombian (Vaup es) and Guyanian Amazon (Fig. 10).

Discussion
Our work with this species complex of the genus Cleistes is an example that shows perspectives and   challenges for advancing the systematics of taxonomically neglected groups with a wide geographic range.For complex groups, scientific collaboration, analysis across the taxon's range, the use of different sources of information and the use of citizen science platforms can help unravel their taxonomy.This work, although largely using images from herbarium collections, involved the analysis of a large number of samples that allowed us to identify and propose potentially informative morphological characters for the group and genus.Our morphometric analysis, as for other plant groups (Camelo-J unior et al., 2022;Cruz-Lustre et al., 2020;H€ opke et al., 2019;Sylvester et al., 2021), shows that this tool is useful for identifying patterns of variation in groups with high morphological variation and challenging taxonomy.The results of morphometry, complemented with the results of our molecular phylogenetic analysis and the analysis of morphology, habitat and geographic distribution, allowed us to characterize each taxon of the complex and support the discussion and taxonomic proposals presented below.

Relationship and differentiation between C. rosea and C. castaneoides
Overall the results of our phylogenetic analyses with the combined dataset were similar to previous studies (Pansarin et al., 2008(Pansarin et al., , 2012)).Since the focus of this study is the relationship between C. rosea and C. castaneoides, the phylogenetic relationships of the remaining species in the genus are beyond our objectives and a discussion on this topic can be found in the references above.Our phylogenetic analyses show C. rosea and C. castaneoides are related taxa forming a highly supported clade, which, together with C. tenuis, emerges as sister to the remaining species of the genus (Fig. 1), corroborating previous molecular phylogenetic studies for Cleistes (Pansarin et al., 2008(Pansarin et al., , 2012)).In turn, Cleistes tenuis, sister of C. rosea/C.castaneoides, presents a very different morphology and an analysis of this species is beyond the focus of this work.Although our molecular data are not available for all taxa considered here and sampling of each taxon is low, the high similarity between the sequences of the two varieties of C. castaneoides and the long branch lengths in the C. rosea/C.castaneoides clade (Fig. 1) show a large genetic difference between the two species, greater than other species pairs in the genus, and further support their separation as distinct species.In addition, our morphometric analyses with a broad sampling of species and specimens of the complex show that C. rosea and C. castaneoides are distinct species (Fig. 2A-C) based primarily on leaf base (decurrent in C. rosea vs. truncate in C. castaneoides), flower fragrance (fragrant vs. odourless), leaf vernation (concave/flat vs. conduplicate), dry leaf colour (brownish vs. blackish) and leaf length (long, usually longer than the floral bracts vs. short, generally shorter than the floral bracts).Meneguzzo and van den Berg (2020) have considered C. castaneoides, C. rosea f. pallida ( C. abdita) and C. angeliana Campacci as synonyms of C. rosea.In fact, C. rosea and C. castaneoides have similar gross morphology, flowers in the same size range, simple lip, and similar colour pattern, but as we show above differ in vegetative morphology, details of floral morphology, flower scent, habitat and distribution.

Identity, taxonomy and infraspecific variability of C. rosea
Cleistes rosea was one of the first species described in the genus from material collected by R.H. Schomburgk in savannas adjacent to Lake Capoey [Capooey] on the coast near the mouth of the Essequibo River in Guyana [Guayana] (Lindley, 1840).Both the lectotype (K-L) and the isolectotype (P) contain only inflorescences without vegetative parts.The sample in K also contains an illustration of the lip and column made by Lindley.Since C. rosea and C. castaneoides differ from each other basically in vegetative morphology and ecological preferences, the absence of these features in the type raises doubts about the true identity of these specimens.Nonetheless, in addition to the type material, we also examined 27 additional collections from different regions of Guyana (Cuyuni-Mazaruni, Demerara-Mahaica, East Berbice-Corentyne and Potaro-Siparuni) as well as some images of Guyana plants from iNaturalist.Among the material examined, leaf development is quite variable, particularly in width, and some specimens with narrow leaves resemble C. castaneoides (Fig. 4A-D).However, in all specimens examined, the plants do not turn dark after drying, the leaves are always evident, more or less detached from the stem, decurrent, flat or concave, the same size or larger than the bracts, the stripes at the apex of the lip are strongly highlighted against the purple background, and the anther is completely white.All these are characteristics of C. rosea.Furthermore, all the 24 specimens of C. rosea from Guyana included in the morphometric analyses grouped with C. rosea (Supplemental Appendix 7).Thus, as far as we can determine, all specimens of the C. rosea/C.castaneoides group available from Guyana are referable to C. rosea.Nevertheless, in the sample Maguire et al. 40623 (NY) from Mount Ayanganna, the third specimen from left to right, consisting only of the apex of the plant, shows a dark colour contrasting with the straw colour of the others, conduplicate leaf and bract larger than leaf, more closely resembling C. castaneoides.Thus, additional studies, preferably by examining live material in the field, are still needed to determine if C. castaneoides also occurs in Guyana.
Our morphometric (Fig. 2A-C) and morphological (Figs 3-7) analyses also showed a subdivision of C. rosea into two subgroups, based mainly on the colour of the flowers.In one of the subgroups the inner side of the sepals is purplish while in the other it is green.Our analyses also showed that this variable is associated with other characteristics, such as the number of leaves, number of flowers, details of flower morphology, differences in floral fragrance, geographic distribution and ecoregion.This result indicates a substructure of the species and the existence of different variants, probably in process of speciation.Based on these results and on a critical analysis of all names and taxa similar to C. rosea described for the complex, we propose and justify below the recognition of two taxa as varieties of C. rosea.
Cleistes buenaventurae Szlach.& Kolan. is one of several species related to C. rosea described by Szlachetko and colleagues from Colombian material.In the protologue, Szlachetko and Kolanowska (2013) pointed out the lip prominently longer than the other flower parts and the venation of the lamellae on the lip base as the main distinguishing characteristics of the species.The first characteristic appears to be some kind of artefact, as it is not observed in the isotypes of the species, where the lip is approximately the same size as the other floral parts (Fig. 6E).Other specimens examined from the same region also present the lip with approximately the same length as the other floral parts (Fig. 6G, I), and in all taxa of the C. rosea complex the sepals are generally slightly longer than the petals and lip.As for the venation of the base of the lip lamella, the veins are somewhat prominent (Fig. 6I), but this feature is present to a greater or lesser extent in other species of the genus with flowers of similar size and prominent lamellae, but are only evident on dry material.As in the remaining characters, C. buenaventurae is morphologically similar to C. rosea, and these characteristics alone do not seem to be enough to separate them as distinct species.
Nevertheless, the description of the colour of the flowers in the type material and the examination of photographs from iNaturalist (2022) and herbarium samples from the region of Buenaventura, on the Pacific coast of Colombia, revealed a regional variation in flower colour, with the sepals light green or greenish-yellow to greenish-ochre, petals pale to dark mauve, and lip white slightly tinted with lilac to mauve at the apex (Fig. 5D-E).As this variation seems fixed in local populations, where all individuals have these same characteristics (C.E.Echeverri Garcia, personal communication, 28 May 2022; S. Camargo, personal communication, 2 June 2022), and is geographically restricted to low altitudes (from sea level to around 300 m) in the western side of the Cordillera Occidental in the department of Valle del Cauca, we followed the same criteria adopted for the other infraspecific taxa accepted in this work and recognized this local variation as a variety of C. rosea, which is treated here as C. rosea var.buenaventurae (Szlach.& Kolan.)J.A.N.Bat.& C.Castro stat.& comb.nov., following recommendation 24B.2 of the International Code of Nomenclature for algae, fungi and plants (Turland et al., 2018).
In addition to the difference in flower colour, the specimens we examined of C. rosea var.buenaventurae were generally more leafy ((2-)3-4(-5) larger leaves per plant) (Fig. 6E) than the typical variety, which usually has 1-2(-3) larger leaves per plant and which are more widely spaced on the stem (Fig. 4A-D).However, leafy plants also occasionally occur in C. rosea var.rosea.In the case of C. rosea var.buenaventurae, this feature may be related to the climate and vegetation of Buenaventura, which is basically covered by lowland forests and has high rainfall, with an annual average of 10,000 mm.Also, C. rosea var.buenaventurae generally has more flowers, (1-)2-3(-6) than C. rosea var.rosea (1-2(-3) flowers), but the overlap is great.According to S. Camargo (personal communication, 2 June 2022), who observed the living plants in situ, the variety has a pleasant and fragrant smell similar to that of vanilla.In the specimen Archer 2142, from Choc o, Colombia, treated here as C. rosea var.buenaventurae (see below), the scent is described as a sweet odour of black pepper.
Cleistes rosea f. xanthina (Ortiz-Valdivieso, 2012) was proposed based on a photograph depicting a flower with yellowish-green sepals, white petals and a white lip with pale violet stripes.The origin of the photo was not indicated, but it was provided by someone from Cali, the capital of the department of Valle del Cauca.Considering flower colour and geographic region, it is possible that it is based on C. rosea var.buenaventurae.Nevertheless, this name is a nomen nudum since it was not formally proposed and lacks a Latin diagnosis and type material.2B-C).In these analyses we coded the variable Sepal colour (Cse) of these specimens as purple, based on the herbarium labels where the flowers were described only as blue or pink.When this variable is coded as missing data, considering the uncertainty regarding the flower colour, these specimens are positioned between C. rosea var.rosea and C. rosea var.buenaventurae, or closer to the latter (Supplemental Appendix 7).
Some samples of C. rosea from Panama (Powell 113 (¼ 3400) -AMES, K, MO) recorded a form with white flowers (i.e., floribus albis) in the same sample with individuals with the typical reddish-purple flowers.One of these specimens located in herbarium K (K001461700) is particularly interesting.In a correspondence between C.W. Powell and R.A. Rolfe attached to the collection, it features a detailed description of the flower colour and two coloured drawings of the flowers.The colouring is reminiscent of other varieties with greenish sepals and purplish apex of the petals and lip, such as C. rosea var.buenaventurae.However, we were unable to examine any recent images of this variation.On iNaturalist, there are several records and images of plants from Panama, and all of them show the typical colour pattern of C. rosea var.rosea.It is possible that these variants represent C. rosea var.buenaventurae, since the Choc o ecoregion extends north into Panama along the Caribbean coast, or alternatively, a variation not yet characterized.Nonetheless, more grounded inferences on this issue will depend on the availability of more data on this variation.
Another variant of C. rosea is known from French Guiana.The general morphology, habitat and flowering period are similar to the typical variety.However, this taxon differs from the regular C. rosea in the colour of its flowers (green or light green or greenish-yellow sepals on both sides, white petals with a light green to cream-coloured base, and lip white with a light green to cream-coloured base, striped with purple on the apical half (Figs 5A-C, 6B) vs. sepals externally light brown to light greenish-brown, internally vinaceous, pinkishpurple, pinkish-brown, purplish-brown or magenta, petals purple, purplish-pink or magenta, and lip white at base becoming purple, purplish-pink or magenta on the apical half and striped with purple to dark purple (Fig. 3E-I)), lip suddenly depressed at 3/4 of its length (Figs 6B, 7B-C) (vs.usually not depressed, Fig. 4F-H) and in flower scent (strong cumin scent vs. rose-like or slightly sweet parsnip scent).The papillae on the lip crest also appear to be different (crested vs. lamellate), but a more detailed examination of this feature is needed to verify this possibility.
With the characters used in the morphometric analyses, C. rosea var.guianensis is almost inseparable from C. rosea var.buenaventurae (Fig. 2A-C).Both species have approximately the same general colour pattern and are usually more leafy than C. rosea var.rosea.However, they are separated by geographic distribution and details of floral morphology.Cleistes rosea var.guianensis is restricted to French Guiana (vs.restricted to the Choc o ecoregion for C. rosea var.buenaventurae) (Fig. 10), has completely white petals (Figs 5A-C, 6B) (vs.pale to dark mauve, Figs 5D-E), lip always suddenly depressed at 3/4 of its length (Figs 6B, 7B-C) (vs.not depressed, Figs 6G, I), anther white tinted mauve to pink along the margins (Figs 5B, 6C) (vs.completely white, Figs 5D-E), and flowers with a strong fragrance of cumin (vs.similar to that of vanilla).
In herbaria and literature C. rosea var.guianensis has been identified as C. rosea (Kolanowska et al., 2011;Szlachetko et al., 2016) or C. rosea f. pallida (Sambin & Aucourd, 2021;Sambin & Ravet, 2021), the basionym of C. abdita, due to association with the flower colour of this taxon, but which differs in a number of morphological and ecological characteristics, and distribution.In iNaturalist C. rosea var.guianensis is mainly identified as C. grandiflora (Aubl.)Schltr., described from French Guiana in 1775 and the type species of Cleistes (Cameron, 2003).Although the vegetative part of C. grandiflora resembles C. rosea, including the leaves with a decurrent base, the original description and illustration of the species (according to Aublet, 1775) depict a short lip, yellow dotted red, trilobed, with rounded lateral lobes and a median lobe with an emarginate and fimbriate apex.All these features contrast with the morphology and colour of the lip of C. rosea and related species.Apart from the protologue, a species of Cleistes with the above characteristics has never been found again in French Guiana and the identity of C. grandiflora is uncertain.Furthermore, in the same publication (Aublet, 1775), the illustration of Limodorum pendulum Aubl.(the basionym of Dichaea pendula (Aubl.)Cogn., the type species of Dichaea Lindl.)depicts a lip which is almost identical to that of C. grandiflora and which is very different from the morphology of the lip of Dichaea.In a systematic revision of Dichaea in Costa Rica, Pupulin (2007) also points out that both the description and the original illustration of L. pendulum are incongruous and that the lip is completely atypical of any of the known species of Dichaea.These discrepancies raise doubts about the accuracy of Aublet's descriptions and illustrations and reinforce uncertainty about the taxonomic identity of C. grandiflora.
Considering that C. rosea var.guianensis is common and widely dispersed in French Guiana (Fig. 10 and examined material of this taxon in the taxonomic treatment), occurs in the same locality as the type of C. grandiflora (commune of Kourou), and is the only large-flowered Cleistes species known from French Guiana, it is possible that C. grandiflora and C. rosea var.guianensis are the same taxon.One explanation for the discrepancies between the protologue of C. grandiflora and the morphological features of C. rosea var.guianensis may be that the specimen examined by Aublet had anomalous or damaged flowers, as suggested by the atypical flower with a trilobed lip illustrated by Cameron (2003).However, given the uncertainties on this issue and that an in-depth discussion on this subject is beyond the scope of this study, we considered it best for the moment to apply the name and taxonomic category proposed here for this taxon from French Guiana and keep C. grandiflora as a doubtful species.
In one collection from Guyana (Henkel 2152 -COL, GH, MO, NY, US), the perianth is described as light green.Based on the geographic origin and the colour of the column, described as white, we tentatively identified this specimen as C. rosea var.rosea.Nevertheless, it is also possible that it corresponds to C. rosea var.guianensis.In the morphometric analyses, the three examined individuals of this collection correspond to the C. rosea var.rosea samples grouped with C. rosea var.buenaventurae/C.rosea var.guianensis, since sepal colour (Cse) was the variable most correlated with axis 3 of the full dataset (Fig. 2B) and axis 2 of the reduced dataset (Fig. 2C).
Cleistes acuminata Schltr.was described from a single collection from Antioquia, Colombia, and the type lost during the bombing of Berlin in the Second World War.Besides the description, the only additional information available about the species is a drawing of the floral segments made by Schlechter and published by Mansfeld (1929).Ortiz-Valdivieso (2012) and Szlachetko et al. (2020) accepted the species, and the former published a photograph of a flower which looks very much like C. rosea.So far, the exact identity of C. acuminata remains uncertain.In the original description, Schlechter highlighted the similarity with C. rosea and pointed out the acuminate sepals, petals and lip as the main distinguishing features.However, our experience with the morphology of Cleistes has shown that the floral morphology can be quite variable, particularly in relation to the lip.For example, see the variations in the lip morphology of C. castaneoides depicted in Fig. 9B, E-F, I-L, where specimens with acuminate sepals and petals are found in both varieties (Fig. 9F, K).Additionally, among the examined flowers for C. rosea, the Renz 5720 (RENZ) specimen has the petals and lip somewhat acuminate (Fig. 4H), approaching Schlechter's drawing of C. acuminata.Since, in general, other characters of C. acuminata agree with C. rosea, and this variation alone does not seem sufficient to differentiate a species, we consider the two to be conspecific.
In the description of Pogonia venusta Schltr.( Cleistes venusta (Schltr.)Schltr.)Schlechter cited the wider petals as one of the distinguishing characteristics of the species.However, it is similar to C. rosea in all aspects, and there are no specific characteristics that allow the separation of the two.Thus, we follow Garay (1978) in treating them as conspecific.Cleistes venusta was described from a single collection from Cauca, Colombia, and the type was one of the specimens from the Berlin herbarium lost during the Second World War.Meneguzzo and van den Berg (2020) selected the collection A.H.G.Alston 8620 (P) as the neotype of C. venusta, which is also the type of Cleistes buenaventurae.However, both C. rosea var.rosea and C. rosea var.buenaventurae occur in Valle del Cauca, and the protologue of C. venusta does not provide enough information to allow us to know exactly which of these taxa the name C. venusta corresponds to.Based on the altitude (c.1500 m) provided in the protologue of C. venusta, it does not correspond to C. rosea var.buenaventurae, which is characteristic of lower altitudes.Therefore, we propose here the lectotypification of C. venusta using Schlechter s line drawing of a flower of the species, posthumously published by Mansfeld (1929).Although this illustration was not included in the protologue and there is no explicit reference to the material on which it was based, according to article 9.4 of the International Code of Nomenclature for algae, fungi and plants (Turland et al., 2018), it is part of the original material once it was most likely drawn from the holotype prior to, or at the time of, preparation of the description and is the element that best corresponds to Schlechter's original concept of the species.
Cleistes romeroana Szlach., Kolan.& Baranow, described from Colombia based on a single specimen, is similar to C. rosea in both vegetative and floral characters.The lip morphology is particularly similar to that of the C. rosea specimen illustrated here (Fig. 4F; Acosta et al. 44 -COL), and we could not find any distinctive feature or combination of characters that separate both species.The only flower in the holotype of C. romeroana (Raugh-Hink B16 -AMES) is somewhat smaller than the average C. rosea.Still, this difference might result from low sampling or that it represents a not fully developed flower.The unusually long length described for the ovary and pedicel (6 cm) most likely corresponds to a mistake (apparently the sum of the last stem extension and part of the ovary) since the ovary and pedicel in the only flower present in the type material, although broken, measures about 2.1 cm long, which is within the accepted variation of C. rosea.The comparison of C. romeroana with C. abdita in the protologue (Szlachetko et al., 2020) is unwarranted, as this last species is very distinct and characterized by slender plants, few and poorly developed leaves, smaller flowers, narrower lip and different habitat type (wet grasslands).
The type of C. romeroana (Raugh-Hink B16 -AMES) comes from the same region as C. rosea var.buenaventurae (western side of the Cordillera Occidental in Valle del Cauca) and shares some characteristics with this taxon, such as the leafy plant and the occurrence at low altitude (100 m).On the other hand, on the sample label the colour of the flower is described as pale lilac only, and it was not possible to determine exactly to which variety of C. rosea it belongs.In the morphometric analyses, we coded the variable Sepal colour (Cse) of the type specimen as purple and the specimen grouped with C. rosea var.rosea.When this variable is coded as missing data, the specimen is positioned between C. rosea var.rosea and C. rosea var.buenaventurae þ C. rosea var.guianensis (Supplemental Appendix 7), showing that sepal colour is the main variable determining the separation between C. rosea var.rosea and the other two varieties and that numerical variables alone are usually not sufficient to separate the varieties, except in the case of extreme values.Based on the morphometric analyses, vegetative morphology and geographic origin of the type specimen, we tentatively include C. romeroana under the synonymy of C. rosea var.buenaventurae.Additional collections from the same population of C. romeroana and a more detailed characterization of flower colour will be necessary to resolve this issue.
Cleistes lehmannii Kraenzl.ex Szlach.& Kolan, described from Colombia (Szlachetko et al., 2020;Szlachetko & Kolanowska, 2014) based on a single collection (Lehmann 752 -W 87212), is also very similar to C. rosea var.buenaventurae, and we were unable to find any consistent characters to support the separation between them.The type is from R ıo Dagua, next to Buenaventura in the Departament of Valle del Cauca, the plants are leafy, have 4-5 well developed leaves, up to five flowers, and the sepals are described as light olive brown and the lip and petals as light carmine.Based on these characters, we tentatively consider C. lehmannii conspecific with C. rosea var.buenaventurae.The few flowers present in the holotype are smaller than average in C. rosea var.buenaventurae, but they apparently correspond to the apical flowers, which are generally smaller, since the basal floral bracts are with fruits.
Cleistes aboucharii Szlach., Kolan.& Baranow, described from Colombia based on a single specimen (Szlachetko et al., 2020), is also very similar to C. rosea, and we were unable to find consistent characters to support the separation between them.All the characteristics mentioned as distinctive, such as the short stem, a single leaf and an obovate lip with an apiculate apex, occur to a greater or lesser extent in C. rosea and in the morphometric analyses the type specimen (Abouchar s.n.-COL 119652) grouped with C. rosea var.rosea (Supplemental Appendix 7).The size of the plants, for instance, varies greatly in C. rosea, and between the two extremes, we cite the specimen Sullivan 446 (MO) with a stem only 12 cm long and the specimen Cañas s.n.(HUA 76164) with a 95 cm long stem.The unusual 20 cm long leaf described in the protologue, and used as one of the distinctive characters of the species, is a mistake since in the holotype (Abouchar s.n.-COL000431381), the single most developed leaf measures 4 Â 1.1 cm.

Additional species related to C. rosea
Another species morphologically related to C. rosea is C. tamboana, described from Ecuador (Dodson, 1997).In common with C. rosea, C. tamboana has leafy plants and flowers with a simple lip but differs in the smaller and generally more numerous flowers ((1-)2-5(-8) flowers vs. 1-2(-3) in C. rosea), among other characteristics.The latter is a striking feature, as C. rosea usually has one flower, occasionally two, and very rarely three.Among other differences C. tamboana has on average more leaves, including the floral bracts ((3-)5-7(-9)), the leaves are smaller (4.5-8.5 Â 1.5-3.2cm, length/width ratio 2.4-3.6),approximately all the same size, regularly arranged on the stem and tinted red.Cleistes rosea is very variable in vegetative morphology.However, generally, the leaves are more concentrated in the upper half of the stem, are less numerous, including the bracts ((2-)4-6(-8) leaves), larger (4.2-13(-16)Â 0.8-4 cm), proportionately narrower (length/width ratio 2.6-8.3), and completely green.The colour and general morphology of the flower are similar between the two species, but in C. tamboana, the flowers are slightly smaller (e.g., lip 3.4-4.3Â 1.5 cm vs. (4.0-)4.5-7.0(-8.7)Â 2.4-3.0 cm in C. rosea), the lip has a light green basal half, and the mauve stripes at the apex are discreet and not very prominent against the magenta background (Fig. 5F).In the protologue (Dodson, 1997) the column is described as winged, which would be another outstanding distinguishing feature in relation to C. rosea, but we did not have suitable material available for analysis to confirm this feature.Cleistes tamboana is known so far only from a few collections from the province of Esmeraldas, Ecuador, in the Choc o region, where it grows on roadsides among secondary tropical forests, at altitudes between 400 and 800 m.It should probably also be expected for neighbouring regions of Colombia.
One last species that deserves comment is C. ayangannensis, described from Guyana (Szlachetko & Baranow, 2012).Although it has been confused with C. rosea very few times, it shares with C. tamboana some characteristics, such as leafy plants, small flowers and simple lip.Nonetheless, it differs in geographic distribution, habitat and flower colour.In C. ayangannensis, the stem is red, the leaves are maroon, tinted vinaceous or green with red margins, plants usually drying black (vs.straw-coloured in C. tamboana), the flowers are smaller (e.g., lip 1.9-3.0cm long vs. 3.4-4.3cm long in C. tamboana), petals are salmon-coloured, the lip is bright pink without stripes at the apex (Fig. 5G), and the capsules are vinaceous to red.Cleistes ayangannensis appears to be restricted to the Guayana highlands in Guyana, Venezuela and Brazil, where it grows at altitudes between 1200 to 2600 m in swampy open places, steep slopes, wooded streamlets, stream banks, at the edge of boggy meadows, bromeliad-dominated savanna, and scrub forests on sandstone and peat.
Cleistes ayangannensis specimens with narrower leaves resemble and can be confused with C. parviflora Lindl.
[¼Cleistes moritzii (Rchb.f.) Garay & Dunst.],which has flowers of roughly the same size and more or less the same colour, but the two species differ in leaf morphology and colour (leaves wide, maroon, tinted vinaceous or green with red margins in C. ayangannensis vs. leaves narrow, evenly green in C. parviflora) and lip morphology (entire in C. ayangannensis vs. with a small lobe or apical projection in C. parviflora).

Taxonomy and infraspecific variability of C. castaneoides
Similarly to C. rosea, C. castaneoides presents two distinct colour patterns: one with sepals, petals and lip internally intense magenta or purplish-pink at the apex, usually with stripes at the apex of the lip in a darker and more intense magenta or purplish-pink (Figs 8D-G,   9B).This pattern corresponds exactly to the species type, as evidenced in the original colour illustration published by Hoehne (1940: plate 136).A second pattern presents sepals light green to greenish-yellow on both sides, petals white, sometimes light green at the base, with the apex internally tinted mauve to pale pink, and lip white or cream-coloured tinted mauve, pink or pale pink at the apex, with the stripes usually in a darker mauve or more intense pink (Figs 8H-K, 9I).This second pattern exactly matches the one described as Pogonia rosea var.augusta Hoehne.
Except for colour, there are no conspicuous morphological differences between these two morphs, as highlighted in the morphometric analyses (Fig. 2A-C) and both occur across the entire geographic range of the species (Fig. 10).However, each pattern always forms homogeneous populations.Even in the few cases where the two patterns were found in the same location on the same date, as in the collections by Pansarin & Mickeliunas 873,888,946 (UEC), each pattern forms homogeneous, spatially separated subsets.Although the two patterns may be sympatric and flower simultaneously, we have never observed, and there is no record of any intermediate forms between the two.Together, these data indicate some degree of differentiation within the species and, in our opinion, justify a taxonomic separation between the two patterns.Meneguzzo and van den Berg (2020) treated Pogonia rosea var.augusta as a form of C. rosea.Based on our results and observations above, and Stace's concepts, we treat here the pattern with the lightest colours as a variety [Cleistes castaneoides var.augusta], which is defined as 'a population of one or several biotypes, forming more or less distinct local facies of a species' (Stace, 1989).

Circumscription and characterization of C. abdita
Cleistes abdita was initially described as Cleistes rosea f. pallida (Carnevali & Ramirez, 1990) and later raised to specific level (Romero-Gonz alez & G omez, 2009).In both cases, it was compared with C. rosea, from which it is quite distinct.It is, in fact, similar to C. castaneoides, both in vegetative morphology and habitat and, for some time, we thought the two could be conspecific.Both C. abdita and C. castaneoides grow in flooded grasslands and have reduced leaves, appressed to the stem, with a truncate base, which separates them from C. rosea.Accordingly, in our morphometric analyses C. abdita is always grouped with C. castaneoides and separated from C. rosea.However, C. abdita has smaller flowers (lip 3.3-5.3Â 1.0-1.4cm vs. 3.9-7.6Â 2.8-3.5 cm in C. castaneoides).The narrower lip is particularly distinctive, and the presence of up to three flowers per inflorescence (vs.1(-2) in C. castaneoides), and slender plants that do not become darker when dry (vs.usually not slender and becoming darker when dry) also provide additional qualitative and quantitative differences.Romero-Gonz alez and G omez (2009) also cited the denticulate margin of the petals and apex of the lip as a distinctive characteristic of C. abdita that, although absent in C. rosea, can also occur in C. castaneoides (Fig. 9B).The relationship between C. abdita and C. castaneoides seems to be similar to the one between C. tamboana and C. rosea, meaning they resemble each of these species in general morphology but differ in the smaller size of the flowers, among some other characters.
Some collections of C. abdita, such as Barbosa & Costa 1423 (INPA), contain both small flowers (sepals 3.5 cm long), as in the protologue of C. abdita, and larger flowers (sepals 5.0 cm long), which overlaps the smallest C. castaneoides, showing that flower size is variable within the same population and making the separation between the two species difficult in these cases.Although all the characteristics mentioned above are variable and there is an overlap in the size of the flowers, the extremes are quite different, and we have never observed plants with the characteristics of C. abdita occurring in populations of the typical C. castaneoides in South-eastern and Central-Western Brazil.Therefore, for now, we choose to retain both as distinct species.No samples of C. abdita suitable for molecular studies were available, and the phylogenetic positioning of C. abdita will depend on adequate samples for this analysis.Based on this species' morphology and type of habitat, we presume that it is related to the C. castaneoides clade.

Taxonomic treatment
Cleistes rosea Lindl.var.rosea, Gen. Sp.Orchid.Pl.: 410.1840.Pogonia rosea (Lindl.)Rchb.f., Xenia Orchid.2: 89.1865 Pogonia rosea (Lindl.)Hemsl., Biol. Centr.-Amer., Bot. 3: 304. 1884 Notes.The illustration of C. rosea in the treatment of Cleistes in Genera Orchidacearum (Cameron, 2003) is particularly interesting as it shows a flower (Milliken 367 -K) with a tripartite lip.We only examined images of this material, including duplicates in MO and NY, and were unable to confirm this detail.In any case, if correct, it must correspond to an anomaly or exceptional variation since there is no other record of this characteristic in C. rosea.
Etymology.The name of the variety refers to French Guiana, where it is restricted.
Diagnosis.Similar to Cleistes rosea var.rosea in the general morphology, habitat and flowering period, but distinct in the colour of the flowers, with green or pale green or pale yellow sepals, white to cream-coloured petals, white to cream-coloured lip, striped with purple on the apical half (vs.brown to greenish-brown sepals, pale lilac petals, lip white, striped with lilac on the apical half) and a lip suddenly depressed at 3/4 of its length (vs.usually not depressed).
Distribution area and habitats.So far, only known from French Guiana, where it is locally common.The plant grows from 5-660 m altitude, in full sun or slightly shaded environments, most often on granite outcrops, on the edge of low vegetation or islets of humicolous mounds; in ravines, embankments and roadsides, on lateritic, clayey or sandy soils; in sandy savannahs, on the edge of the forest, in areas not subject to flooding and, more rarely, in savannah on hydromorphic soil, on the edge of the forest.
Illustrations.Kolanowska et al. (2011, photo 16), Sambin and Ravet (2021, photos 49-50), Sambin and Aucourd (2021, photo 132) Notes.This variety occurs throughout almost the entire geographic distribution of the species, in northern South America (Colombia, Peru and Venezuela) and Brazil, in the States of Amazonas, Bahia, Goi as, Maranhão, Mato Grosso, Minas Gerais, Par a, Roraima and the Distrito Federal.The general morphology, habitat and flowering period are similar to the typical variety.The lip is often wider than in the typical variety (Fig. 9J, L), but variations occur in both taxa.The sepals are light green, green or greenish-yellow on both sides (vs.externally light brownish-green, light brownish-pink or brownishpink tinted magenta, internally light brown, brownishpink, magenta or purplish-pink; in C. castaneoides var.castaneoides), the petals are white with a light green base, apex sometimes internally tinted mauve to pale pink (vs.mauve, magenta or purplish-pink, sometimes with pink base), and the lip is white or cream-coloured, apex tinted mauve, pink or pale pink, with the stripes dark mauve or intense pink (vs.white, upper half magenta to purplish-pink, with the stripes dark and more intense magenta or purplish-pink), and the crest is predominantly pale yellow to light greenish-yellow in colour, sometimes becoming tinted with mauve, pink or white at the apex (vs.predominantly white with a golden to orange-yellow centre, rarely with the outermost fimbriae tinted with magenta) (Figs 8H-K, 9I).
Occasionally the petals and lip are almost entirely white, with only very faint pale pink stripes near the apex margin (Fig. 8K).
An additional difference appears to be the lip crest.In both taxa, the crests are lamellar and longitudinal.In C. castaneoides var.castaneoides, near the terminal portion, along almost their entire length, the lamellae have papillate projections that elongate into fimbriae (Fig. 9G).In turn, in C. castaneoides var.augusta, the lamellae are crenate with very small or absent projections that only form papillate projections near the terminal portion of the crest (Fig. 9M).However, as with other floral characteristics, it is possible that this is also variable, so further analyses with a larger sample are needed to verify these preliminary observations.Considering the differences in flower colour and lip crest morphology, it is possible that the varieties are visited and pollinated by different pollinators, which could explain the apparent reproductive isolation between them.So far, there is no information about the pollination of the species and studies in this sense should shed light on this question.
Notes.As delimited here, C. abdita occurs in the Venezuelan (Amazonas, Bolivar), Brazilian (Roraima) and Colombian (Vaup es) Amazon, and in Guyana (Fig. 10), in wet savannas, generally below 500 m.Plants are slender (stem 1.5-4.0mm wide), with 1(-2) reduced, narrow leaves (2.0-7.2Â 0.5-1.6 cm), base truncate, embracing the stem, 1(-3) flowers with the floral bracts 3.2-10.5cm long, 1.0-1.9times longer than the leaves, and lip 3.3-5.3Â 1.0-1.4cm.Sepals are green or yellowish-green, petals and lip completely white or cream-coloured, sometimes with a pale pink or lilac apex, callus yellow, and column white.Similar to C. castaneoides, there is no record of odour for this species' flowers.Most of the specimens cited by Szlachetko et al. (2020) in the treatment of C. abdita for the Orchid Flora of Colombia correspond to C. rosea, and the description of C. abdita is consequently a mixture of the two species.
There is a discrepancy in the colour of the flowers between the protologue of C. rosea f. pallida (Carnevali & Ramirez, 1990) and C. abdita (Romero-Gonz alez & G omez, 2009).While in the first, the sepals are described as greenish-cream, the petals as white, and the lip white with a pink apex.In the second description, the sepals are described as purplish-brown, the petals as intense purple, and the lip basally pink becoming darker at the apex.Based on the herbarium labels, the colour of the flowers in the protologue of the former is based on the specimen Carnevali et al. 2661, while in the latter it is based on the specimen Carnevali et al. 2662, both from the same site and collected on the same day.Cleistes abdita presents a variation in flower colour similar to C. castaneoides and C. rosea, with 4b.Sepals internally light green, green, greenish-yellow or greenish-brown; petals white with a light green base, sometimes tinted mauve at the apex; lip white slightly tinted with lilac to mauve at the apex, stripes accentuated in darker mauve or purple.. Sepals externally light brownish-green, light brownish-pink or brownish-pink tinted magenta, internally light brown, brownish-pink, magenta or purplishpink; petals mauve, magenta or purplish-pink, sometimes with pink base; lip white, upper half magenta to purplish-pink, stripes dark and more intense magenta or purplish-pink, crest predominantly white with a golden to orange-yellow centre, rarely with the outermost papillae tinted with magenta, terminally fimbriate; anther magenta to purplish-pink..........................C. castaneoides var.castaneoides 7b.Sepals light green, green or greenish-yellow on both sides; petals white with a light green base, apex sometimes internally tinted mauve to pale pink; lip white or cream-coloured, apex tinted mauve, pink or pale pink, stripes dark mauve or intense pink, crest predominantly pale yellow to light greenish-yellow in colour, sometimes becoming tinted with mauve, pink or white at the apex, terminally papillose; anther white...................................C. castaneoides var.augusta
) and pale mauve to intense purplish-pink in C. castaneoides (Figs 8G, 9C).Additionally, several of the records of C. rosea indicate that the flowers are fragrant, while the flowers of C. castaneoides are odourless.Records range from a rose-like scent (DeWolf & Aristeguieta 2130) to a slightly sweet parsnip odour (Maguire et al. 43840).Finally, the sepals of C. castaneoides are usually markedly reflexed in fully open flowers (Fig. 8D-G), while in C. rosea, they are patent with at most only a reflexed apex (Fig. 3E-I).
4144 -SEL) are referable to C. castaneoides and C. costaricensis Christenson, respectively.In a survey of Costa Rican orchids, Dressler (2003) records only C. costaricensis for the country.

Fig. 2 .
Fig. 2. Representation of the first PCoA axes of the C. rosea complex analysed in this study.(A, B) Complete dataset (192 individuals belonging to eight taxa and 16 variables).(C) Reduced dataset (170 individuals (C.ayangannensis and C. tamboana excluded) and 15 variables).
numbers above the average found in C. rosea.Both the Buenaventura and Choc o departments are located in the Pacific/Choc o region, characterized by an equatorial climate and high rainfall, with areas near Quibd o, Choc o, reaching up to 13,000 mm annually.Although we do not know the exact colour pattern of the flowers of these specimens from the department of Choc o, we include them here under C. rosea var.buenaventurae due to the abovementioned similarities.Still, a more detailed characterization of these populations is needed to establish their exact relationship with other taxa in the group.In the morphometric analyses, the specimens Archer 1751 (two individuals) and Archer 2142 (one individual) correspond to the three samples of C. rosea var.buenaventurae grouped with C. rosea var.rosea (Fig.

Table 1 .
Number of herbarium collections, individuals from herbarium samples or records (¼ herbarium collections þ iNaturalist records) of Cleistes rosea and related taxa used in the morphology, morphometry and distribution map respectively.

Table 2 .
Characters used as variables in the morphometric analysis of the Cleistes rosea complex.Excluding basal sheaths.QT ¼ quantitative, QL ¼ qualitative.

Table 3 .
Correlation between variables and axes with eigenvalue >1.0 from PCoA analysis of data from 16 variables and eight taxa of the Cleistes rosea complex.High correlations (>0.6) are indicated in bold.Except for the variable 'Ffl' where the correlation was calculated with 187 individuals due to unavailable data for C. tamboana.ÃÃ Invariant character in the set of 170 individuals, not included as a variable.