Revisiting the taxonomy of the Neotropical Haemodoraceae (Commelinales)

Abstract Based on extensive herbarium, field, botanical illustration, and molecular phylogenetic research, five genera and eight species are recognised for the Neotropical Haemodoraceae. New taxa include Cubanicula Hopper et al., Xiphidium pontederiiflorum M.Pell. et al. and Schiekia timida M.Pell. et al. Two new combinations are made, Cubanicula xanthorrhizos (C.Wright ex Griseb.) Hopper et al. and Schiekia silvestris (Maas & Stoel) Hopper et al. We also correct the author citation for Xiphidium, provide the necessary typifications for several names and present an updated identification key, comments, and photo plates for all species. Finally, we provide high-quality illustrations for most of the recognised species and their diagnostic characters.

Etymology. Named for Cuba, in which the genus is narrowly endemic. The diminutive 'icula' is an allusion to the fact that this genus is second only to Pyrrorhiza in Haemodoraceae in its restricted geographical range.
Taxonomic history. The types of Xiphidium xanthorrizon were collected by the American botanist Charles H. Wright (1811Wright ( -1885, who, between 1856-1867, 'travelled all over Cuba with the exception of the highest mountains and tripled the number of the phanerogamous plant species known from this territory' (Borhidi 1991: 16). New taxa collected by Wright were described by Göttingen's Professor August H.R. Grisebach (1814Grisebach ( -1879, primarily in his Plantae Wrightianae e Cuba Orientali, published in two parts from 1860-1862. However, X. xanthorrizon was not published until 1866, in Griesebach's Catalogus Plantarum Cebensium, in which he attributed the new species' name to Wright. Ascertaining Wright's itinerary during his three periods on Cuban expeditions has been problematic: ' […] his travels were confined chiefly to the two ends of the island, leaving the great central portion largely unexplored. It is unfortunate that the labels on his plants, at least in most of the collections where they are to be found, bear only the inscription "Cuba" or "in Cuba orientali".' (Underwood 1905: 291). Moreover, many of Wright's collections made in western Cuba were irrevocably damaged in transport to the USA: 'It appears from Wright's correspondence that a considerable portion of his collection was lost, mainly that collected in the rich tobacco region of the western end of the island (Pinar del Rio). How extensive this loss may have been, probably cannot now be estimated, but it was certainly considerable.' (Underwood 1905: 291). The author also quotes some sentences found in Dr Gray's Letters (2: 555) that explain the cause of the loss of these specimens: 'April 8 th [1867] It grieves my heart and will grieve yours badly when I tell you that your boxes were put under a cargo of wet sugar, which drained into them and have [sic] ruined the collection. […] As to specimens to dispose of, say only one-half or one-third of the whole mass is left fit for it… [Ever your disconsolate A. GRAY.]' (Gray 1867 apud Underwood 1905: 291, 292).
These problems aside, Underwood (1905) managed to assemble a sketch of Wright's many Cuban itineraries through 200 letters written to Asa Gray and other sources that mentioned dates and place names. Perhaps because of a shipment earlier than the calamity referred to above by Asa Gray, Wright's collections of Xiphidium xanthorrizon persist. Wright probably collected X. xanthorrizon when he was stationed at Retiro -'a finca near Taco Taco where Don Jose Blain lived' (Underwood 1905: 297), either in June-September 1863 or, more likely, in January-May 1864. This can be deduced from labels on the types that provide the dates 1860-1864 and a statement in a letter written in Havana on 28 July 1864: 'plants boxed ready to embark' (Underwood 1905: 298).
The type location and Wright's collection number of X. xanthorrizon is cited by Maas and Maas-van de Kamer (1993: 31) as 'Cuba. Pinar del Rio: San Cristobal, Wright 3259'. The only reference to San Cristobal cited by Underwood (1905: 298) is for a letter written at Retiro on the 15 June 1866 -'went again to San Cristobal on the 10 th '. Since San Cristobal is only 10 km ENE of Retiro on the main road to Havana, it is clearly a place that Wright would have gone through whenever visiting Retiro in the years 1863, 1864, and 1866. For example, on 19 May 1864, Wright wrote: "Made an excursion of ten days eastward and southward to La Concordia, San Leon, etc." (Underwood 1905: 298).
Subsequent collections filled in knowledge of the geographical distribution of X. xanthorrizon, including an early collection from the 1860s by Jose Blain first recording the species from the northern portion of Isla de Juventud (= Isla de Pinos). The specimen (in the Field Museum) was annotated as Xiphidium floribundum Sw (= X. caeruleum), yet associated notes said (Millspaugh 1900: 426): ' […] In Cuba this species grows only in shady situations in glens, never on the open savannas; here, however, it seeks the open plains far from shade -Blain.' Moreover, an old handwritten slip attached to the Field Museum specimen, presumably written by Charles Wright, gave the species as X. xanthorrizon, and this is undoubtedly the identity of Blain's specimen. It is X. xanthorrizon, not X. caeruleum, that is common on open savannahs on Isla de Juventud, a view affirmed in subsequent maps and accounts of Cuban Haemodoraceae (Maas and Maas-van de Kamer 1993;Urquiola Cruz et al. 2000). The species' range has not been extended from the open pine woodlands on the white sands of Pinos del Rio Province and the Isla de Juventud, despite extensive modern collections across Cuba, such as the 20,000 sheets made by Borhidi (1991Borhidi ( ) and colleagues in 1969Borhidi ( -1970Borhidi ( and 1974Borhidi ( -1976 for phytogeographic and vegetation mapping purposes.  Until now, treatments of X. xanthorrizon after the original description have not challenged the generic placement of the species (León 1946;Simpson 1990Simpson , 1998bMaas and Maas-van de Kamer 1993;Urquiola Cruz et al. 2000). Indeed, Simpson (1990: 729) remarked, 'Xiphidium consists of X. coeruleum [sic] and X. xanthorrhizos [sic], which differ only in minor morphological features and are likely more closely related to one another than to any other genus. However, because no definitive synapomorphy is evident for Xiphidium, its monophyly cannot be affirmed.' Although he undertook a comprehensive examination of the morphology and anatomy of the genera of Haemodoraceae, Simpson (1990) did not include both species of Xiphidium in his study in order to test the genus' monophyly. Instead, he chose only to represent the genus by sampling X. caeruleum. An examination of seeds alone would have raised questions about the generic placement of X. xanthorrizon. Simpson (1993) discovered the unusual absence of septal nectaries in both Xiphidium species and interpreted this trait as an autapomorphy for the genus associated with buzz pollination by bees, which was known for X. caeruleum (Buchmann 1980), but the pollination ecology of X. xanthorrizon was not documented. Maas and Maasvan de Kamer (1993: 11) speculated that 'The differently coloured nectar guide on the three adaxial tepals of X. xanthorrizon suggest that an insect pollinator alights in a consistent orientation, forwardly directed to collect pollen from the shorter stamens, in the meantime being dusted by the largest stamen. ' Simpson (1993) affirmed an observation of Maas and Maas-van de Kamer (1993) that X. xanthorrizon has longitudinal anther dehiscence, whereas X. caeruleum anthers commence with nearly poricidal dehiscence, becoming longitudinal as flowers age or dry out (Buchmann 1980). Such a difference echoed a number of other traits overlooked by many authors that call into question the hypothesis that X. xanthorrizon and X. caeruleum are sister taxa.
Regarding generic relationships of Xiphidium, Simpson (1998a: 217) elaborated: 'Within this superior-ovaried group [of subfamily Haemodoroideae], Wachendorfia and Barberetta are united in having a similar pollen ultrastructure (Simpson 1983(Simpson , 1990 and Schiekia and Pyrrorhiza are united in having staminodes and similarities in ovule anatomy (M.G. Simpson, 1990, unpubl.). The exact relationships of Xiphidium to these genera is unclear.' Molecular phylogenetic analyses have yet to clarify the systematic position of Xiphidium in this clade (Hopper et al. 1999. Maas and Maas-van de Kamer (1993: fig. 5) were the first to illustrate and compare SEM micrographs of the seeds of X. xanthorrizon and X. caeruleum, which differ significantly. Indeed, seeds of X. xanthorrizon resemble those of Pyrrorhiza in being large (i.e., 2.5-3.5 mm long) and covered with 1-1.5 mm long coarse hairs (Fig. 4D, H), whereas X. caeruleum has cuboid, black seeds 0.5-1.0 mm in diameter and they are minutely tuberculate, lacking hairs (Fig. 4L), similar to seeds of Schiekia (i.e., S. orinocensis and S. timida). Maas and Maas-van de Kamer (1993: 10) suggested that 'the hairy seeds of Xiphidium xanthorrhizon and Pyrrorhiza neblinae, both savanna plants, might very well be dispersed by animals having seeds adhering to their body (i.e., exozoochoric dispersal).' Simpson (1990: 754) scored X. caeruleum as enantiostylous, but with 'actinomorphic and erect (not zygomorphic and horizontal) flowers without any bilaterally sym-metric nectar guides. ' Maas and Maas-van de Kamer (1993: 11) affirmed that X. xanthorrizon 'clearly displays' enantiostyly of the latter kind, differing significantly from the flowers of X. caeruleum. Despite these floral differences and significantly divergent seed morphologies between X. xanthorrizon and X. caeruleum, these authors retained the traditional circumscription of Xiphidium s.lat. With the recognition of a second species of Xiphidium s.str. in the present study, it became clear that the inclusion of X. xanthorrizon in Xiphidium s.lat. was untenable from the morphological perspective (Pellegrini 2019), added to strong molecular support (Hopper et al. in prep).

Distribution and habitat.
Lachnanthes caroliniana is known to occur from Nova Scotia (Canada) to Florida (USA), reaching Cuba (Fig. 9). It grows in marshy and acidic environments, swampy grasslands, and moist pine forests throughout its range, generally producing extensive clonal populations.
Phenology. Flowers and fruits from April to November. Conservation status. Lachnanthes caroliniana possesses a wide EOO (1,886,962 km 2 ) but a narrow AOO (ca. 616 km 2 ). Nonetheless, although generally abundant within its native range, L. caroliniana is listed as Endangered in four USA States (i.e., Connecticut, Maryland, New York, and Tennessee), as Threatened in Rhode Island and of Special Concern in Massachusetts (USDA-NRCS 2013) and as Threatened in Canada (COSEWIC 2009). Thus, following IUCN's (2001) recommendations, L. caroliniana should be considered as Vulnerable (VU).
Comments. Lachnanthes caroliniana is morphologically variable regarding stature and colouration, with much of this variation being related to environmental conditions. The roots and underground organs can range from yellowish-orange to dark red, the leaves, peduncles, bracts, and the outside of the tepals can range from light to dark green to bluish-green, and the tepals can be internal surface light green to yellowish-green to bright yellow. Aside from that, plants can range from 10 cm to over 100 cm tall.
Lachnanthes caroliniana is commonly considered a widespread weed in blueberry and cranberry crops (Meggitt and Aldrich 1959;Robertson 1976;Meyers et al. 2013), pastures (Ferrell et al. 2009) and to form extensive clonal populations followed by feral swine rooting disturbance (Boughton et al. 2016). Nonetheless, L. caroliniana    is an important nectar source for many insects (Hopper, pers. observ.) and a pollen source for bees and certain flies. It is viewed as an important "bridge species" supporting flower visitors in summer until fall (autumn) daisies begin to bloom (Boughton et al. 2016). Its seeds also constitute an important food source for sandhill cranes (Valentine and Noble 1970). Comments. Pyrrorhiza was initially considered as being closely related to Schiekia Meisn. (Maguire and Wurdack 1957), a view supported by the morphological phylogeny of Simpson (1990), but not supported by the anatomical studies of Aerne-Hains and Simpson (2017), the molecular phylogeny of Hopper et al. (in prep.) and the new morphological phylogeny for the family (Pellegrini 2019). As currently understood, Pyrrorhiza is sister to Cubanicula, with both being sister to Xiphidium s.str. (Hopper et al. in prep.). The supposed relation between Pyrrorhiza and Schiekia was thought to be supported by the zygomorphic perianth, dimorphic stamens, and the discontinuous subexterior exine wall (Simpson 1983(Simpson , 1990). However, the first two characters are clearly homoplastic in Haemodoroideae, while the third seems to be a convergence between Pyrrorhiza and Schiekia (Pellegrini 2019). Pyrrorhiza shares with Cubanicula and Xiphidium s.str. the sand-binding roots, campanulate and pollen rewarding flowers, mainly white perianth, tepals with an apical black mucron, anthers as long as to ca. ½ times shorter than the filaments and enlarged placental attachments subtending the ovules and fruits with thickened septal ridges (Pellegrini 2019). It shares exclusively with Cubanicula the peculiar lenticellate seeds with the testa's margin covered with coarse trichomes (Hickman 2019; Pellegrini 2019). Distribution and habitat. Pyrrorhiza neblinae is at present only known to occur at the Venezuelan side of the Cerro de la Neblina (Fig. 12), but most likely also reaches the Brazilian side. It grows in open, acidic, and swampy Heliamphora Benth. (Sarraceniaceae) and Bonnetia maguireorum Steyerm. (Bonnetiaceae) savannahs, with Euterpe Mart. (Arecaceae), along streams, between 1800-2100 m alt. Due to its cormose underground system producing cormlets, P. neblinae forms dense clonal clusters. Its pollination syndrome is unknown, but based on the vestigial pair of septal infralocular nectaries, it is most likely a pollen-rewarding, self-compatible species.

Pyrrorhiza neblinae
Phenology. It was found in bloom and fruit from November to February. Conservation status. As aforementioned, Pyrrorhiza neblinae is only known from a single Amazonian mountain. It possesses very narrow EOO (20 km 2 ) and AOO (ca.   Comments. Pyrrorhiza neblinae is still poorly known, with only a handful of collections. Nonetheless, it is known that P. neblinae is restricted to swampy and rocky montane savannah (i.e., tepuis). The peculiar cormose underground system of P. neblinae is only comparable to those of Barberetta Harv., Wachendorfia Burm. (both Haemodoroideae) and Tribonanthes Endl. (Conostylidoideae) (Simpson 1998b). Nonetheless, the corms in Barberetta and Wachendorfia are further connected by long, stolon-like flagelliform-shoots, which are unique in the family (Pellegrini 2019). The seeds covered with coarse trichomes might function in adherence to animal fur or feathers as an aid to dispersal (Maas and Maas-van de Kamer 1993). Alternatively, the seeds covered with coarse trichomes might also be an adaptation to hydric stress. These projections might help the seed to quickly absorb and store water, which could come in handy in such an inconstant environment such as the Amazonian tepuis (i.e., Pyrrorhiza), white sand savannahs (i.e., Cubanicula), and the seasonally-dry fynbos from South Africa (i.e., Wachendorfia) (Pellegrini, pers. observ.). Seeds with coarse trichomes are recovered as a synapomorphy for the clade composed by Barberetta, Cubanicula, Pyrrorhiza, Schiekia, Wachendorfia, and Xiphidium. Nonetheless, coarse trichomes in the seed testa are independently lost several times, such as in Barberetta (smooth), Schiekia (reticulate in S. orinocensis and S. timida), Wachendorfia (smooth in W. thyrsiflora Burm.), and Xiphidium (tuberculate) (Pellegrini 2019 Comments. Schiekia is indisputably closely related to Wachendorfia (Hopper et al. 1999Hickman 2019;Pellegrini 2019;Hopper et al., in prep.), which is shown by its taxonomic history and due to several morphological characters. Schiekia and Wachendorfia share some unique floral traits, such as the perianth apertures (produced by the connation of five tepals, giving the flowers a peculiar bilabiate appearance and producing two basal pouches ;Simpson 1990) and the infralocular septal nectaries with commissure slits which channel the nectar to the perianth apertures (Simpson 1993;Pellegrini 2019). These features serve as strong morphological synapomorphies that support the clade composed by Schiekia + (Wachendorfia + Barberetta), with a posterior loss of the perianth apertures in Barberetta (Pellegrini 2019). The nectary apparatus in Barberetta is also remarkably similar to that of Wachendorfia and Schiekia and only lacks the ducts that would carry the secreted nectar to the perianth apertures (Simpson 1993). Furthermore, Schiekia and Wachendorfia share the presence of tapering trichomes, while Barberetta and Wachendorfia share the unifacially-plicate leaves, which are unique in the family and the order as a whole (Simpson 1990;Pellegrini 2019). The staminode-like structures are synapomorphic to Schiekia (Pellegrini 2019) and cannot be considered actual staminodes, in fact, representing a de novo structure (Simpson 1990;Pellegrini 2019). These staminode-like structures seem to represent some kind of corona (i.e., a perianth projection), comparable to the ones observed in many Amaryllidaceae and Passifloraceae. Their function is most likely associated with the genus' floral biology and could represent enlarged osmophores, which would aid in the attraction of pollinators, together with the nectar. Nonetheless, reproductive biology studies in Schiekia are entirely lacking and are necessary to understand the function of these staminode-like structures. Furthermore, ontogenetic studies are also necessary to understand the origin and to propose a more suitable and definite name to these structures.  (1816) mentions a collection made on Isla de Pararuma, Río Orinoco, but makes no reference to the collector, collection number, or herbarium. During a visit to P herbarium, we came across two specimens in which the labels matched the locality in the protologue and also had a label indicating it had been part of the Bonpland & Humboldt herbarium. The specimen P00669614 is clearly what the majority of the original illustration was based upon, while P00669615 was only used to illustrate the fruits. Thus, since the specimen P00669614 possesses well-preserved leaves and stems, floral buds, and mature flowers, it is here designated as the lectotype. When describing Schiekia flavescens, Maury (1889) mentions two collections, Gaillard 52 and Chaffanjon 185. During a visit to P, we were unable to locate the collection Chaffanjon 185 but managed to find Gaillard 52. The latter was cited by Maury as a mixed gathering, with two specimens of his S. flavescens and a central specimen of S. orinocensis. Thus, we designate the two lateral specimens (right and left) as comprising the lectotype for S. flavescens.

Schiekia orinocensis
Distribution and habitat. Schiekia orinocensis, in its current circumscription, is a far more geographically-restricted taxon than traditionally accepted. It is known to occur in Colombia, Guyana, Venezuela, and Brazil (States of Amazonas, Pará, and Roraima) (Fig. 14), in tepuis and other montane formations in the Guyana Shield, in seasonally-flooded environments.
Phenology. It was found in flower and fruit from June to October, during the dry season.
Conservation status. Schiekia orinocensis possesses a wide EOO (1,193,173 km 2 ) but a relatively narrow AOO (ca. 224 km 2 ). This narrow AOO might be related to the relatively reduced number of collections, especially when compared to S. timida. The relatively small number of specimens might be due to the difficulty of reaching and collecting in tepuis and other mountainous formations in the Amazon Region. Nonetheless, field observations by one of us (EJH) indicate that S. orinocensis forms consid-  erably smaller and more restricted subpopulations than S. timida, which might indicate it is ecologically more specific in its requirements. Thus, following IUCN's (2001) recommendations, S. orinocensis should be considered as Vulnerable [VU, A2ab+C2a(i)].
Comments. Schiekia has consistently been treated as a monospecific genus until the present study, given that S. flavescens has been considered a synonym of S. orinocensis since very early days. Nonetheless, previous studies, such as Maguire and Wurdack (1957) and Maas and Maas-van de Kamer (1993), have treated the polymorphism observed in herbarium specimens by recognising different subspecies. Both previous attempts to divide S. orinocensis were almost entirely based on vegetative morphology (Maguire and Wurdack 1957;Maas and Maas-van de Kamer 1993), with the second one also relying on the proportion between the leaves and the inflorescences (Maas and Maas-van de Kamer 1993). The observed variation in plant stature and leaf length and width, which was used by previous authors to recognise subspecies (Maguire and Wurdack 1957; Maas and Maasvan de Kamer 1993), seems to be environmental and, thus, is here disregarded as taxonomically relevant. Our present treatment is based on extensive field and herbarium studies. It suggests that three species can be recognised based on ecological preferences, rhizome morphology, leaf morphology, tepal arrangement and colouration, the width of the filiform staminode-like projections, capsules morphology and colouration, and seed ornamentation. Schiekia orinocensis s.str. is morphologically similar to S. timida due to its rhizome morphology, leaf arrangement and consistency, inflorescence architecture, upright to patent flowers, inflated medial filament, and tuberculate seeds. Schiekia orinocensis s.str. can be differentiated by its leaves with inconspicuous veins (vs. conspicuously veined in S. timida), chasmogamous and bilabiate flowers (vs. cleistogamous and narrowly tubular), pedicels gibbose at the apex (vs. not gibbous), tepals with apex reflexed and apricot to cream (vs. straight and light to medium green), upper tepals with three dark orange to orange-brown nectar guides (vs. lacking nectar guides), staminode-like projections fusiform and almost as long as its subtending tepal (vs. filiform and 1/3 the length of its subtending tepals) and capsules broader than long (vs. slightly longer than broad or as broad as long). Schiekia orinocensis s.str. and S. silvestris share the chasmogamous flowers and upper tepals with nectar guides, thick and fusiform staminode-like projections and capsules slightly longer than broad or as broad as long. Nonetheless, they can be easily differentiated based on vegetative morphology, flower orientation, inflation of the medial filament, capsule colouration, and seed ornamentation (see below).  (Fig. 18). Found growing in the seasonallyflooded forest understorey, near rivers. Phenology. It was found in flower and fruit from January to November, but peaking during the dry season.

Schiekia silvestris
Conservation status. Schiekia silvestris possesses a wide EOO (1,634,289 km 2 ) but a relatively narrow AOO (ca. 392 km 2 ). This narrow AOO might, once again, be related to the difficulty for collection in the Amazon Region. Nonetheless, the number of known collections is relatively large, which leads us to believe this species might be  Comments. Schiekia silvestris is by far the easiest species to differentiate from the three accepted by us in the present study. It is the only species to exclusively inhabit understorey and other mesic habitats and has a growth form similar to that of Xiphidium caeruleum, with its long and trailing rhizomes and leaves evenly distributed along the stem. Aside from that, the leaves are considerably more delicate and broader, and herbarium specimens of S. silvestris are commonly misidentified as X. caeruleum in Brazilian herbaria. Furthermore, the inflorescences of S. silvestris generally possess a corymb-like appearance, added to the diminutive and strongly bilabiate, pendulous, apricot to orange-yellow flowers, with tepals recurved in the upper half and non-inflated medial filament. The capsules of S. silvestris also tend to be much broader than those of S. orinocensis and S. timida, ranging from green when immature to chocolate brown when mature. Finally, it is the only species of Schiekia to present seeds with short and coarse trichomes scattered across the reticulate testa ( Fig. 17U-W). On the other hand, S. orinocensis and S. timida (Fig. 20T-V) present evenly reticulate testa. Diagnosis. Similar to Schiekia orinocensis (Kunth) Meisn. in rhizome morphology, leaf arrangement and consistency, inflorescence architecture, floral orientation, and filiform staminode-like projections, but differs due to its leaves with impressed veins, narrowly tubular and cleistogamous flowers, tepals with apex straight and light to medium green, upper tepals lacking nectar guides, medial filament inflated, staminodelike projection 1/3 the length of its subtending tepal and capsules slightly longer than broad or as broad as long.
Specimens seen ( Etymology. The epithet means "shy" and makes reference to the cleistogamous flowers, which open only a few millimetres. This is the first record of cleistogamy in Neotropical Haemodoraceae, which was previously recorded only for the Paleotropical genus Haemodorum. Distribution and habitat. Schiekia timida is currently known for Bolivia, Brazil (States of Amazonas, Pará, Roraima, Tocantins, Maranhão, Goiás, and Mato Grosso), Colombia, Guyana, and Venezuela (Fig. 21). Found growing in seasonally-flooded grasslands.
Phenology. It was found in flower and fruit from November to June, rarely during July and August, but peaking during the rainy season.
Vernacular name and use. According to specimen labels, S. timida is called "ahtu" in the language spoken by the native Brazilian Krahô tribe. It seems to be used in some religious ceremonies, mixed in a drink with some confirmed psychoactive plants.
Comments. Schiekia timida is morphologically similar to S. orinocensis due to its rhizome morphology, leaf arrangement and consistency, inflorescence architecture, floral orientation, and inflated medial filament. Nonetheless, it differs due to its conspicuously veined leaves, narrowly tubular and cleistogamous flowers, pedicels not apically gibbous, tepals with apex straight and light to medium green, upper tepals lacking nectar guides, staminode-like projections filiform and 1/3 the length of its subtending tepals and capsules slightly longer than broad or as broad as long. Until the present work, both species were treated under a broad concept of S. orinocensis subsp. orinocensis, as proposed by Maas and Maas-van de Kamer (1993). However, as noticed during fieldwork, S. timida seems to be a cleistogamous species, with flowers never opening more than a few millimetres. Durandia macrophylla Boeckeler (= Xiphidium caeruleum Aubl.).

Xiphidium
Type species. Xiphidium caeruleum Aubl. Nomenclatural history. It has been widely accepted that the original place of publication of the generic name Xiphidium is "Histoire des Plantes de la Guiane Françoise" by Aublet (1775). Nonetheless, Aublet never clearly states to be proposing a new genus. This seems to follow his publication's formatting, where none of the new taxa present any explicit statement indicating that they are newly proposed. At the end of the Latin diagnosis and French comments, Aublet (1775: 35) mentions that his new species differs from the one described by Loefling (1758) due to its "fine stems and leaves furnished with hairs, blue flowers, and oval and acute petals". This statement makes it clear that Aublet had access to Loefling's publication (1758) and knew of the description of his new genus Xiphidium. Finally, Dorr and Wiersema (2010) give the final support to our interpretation when they explain that in several instances, Loefling (1758) cited a genus published earlier by Linnaeus or P. Browne, followed by a full stop, (an) alternative generic name(s) and a description. The authors also point out that, on some occasions, this formatting has been misinterpreted as the proposal of species' names (i.e., binary combinations), which they are not. That was the case of Xiphidium Loefl., which was misinterpreted as representing a new species, Ixia xiphidium Loefl. (e.g., Maas and Maas-van de Kamer 1993), instead of the publication of a new genus. Thus, the genus Xiphidium was originally described by Loefling (1758), without the inclusion of any species. The proposal of Xiphidium by Loefling (1758) is based on the author not agreeing on the inclusion of all elements/species by Linnaeus in his Ixia L.
The first species name to be validly published in Xiphidium was only proposed almost 20 years later, by Aublet (1775), as X. caeruleum Aubl. The publication of the generic name Xiphidium by Loefling (1758) makes it clear that the author recognised a sole species for that genus. The practice of not providing a specific epithet when describing monospecific new genera was common practice at the time. A similar situation, with the description of the type genus of Haemodoraceae -Haemodorum (Smith 1798) -, supports this interpretation. When first described, Haemodorum was considered monospecific and, therefore, was not given a specific epithet, according to the standard practice of J.E. Smith (1798). Only seven years later, another author (Vahl 1805) provided an epithet for Smith's plant, as H. corymbosum Vahl. Thus, as the first species formally published and associated with Xiphidium, X. caeruleum automatically typifies this generic name.
Comments. Xiphidium has traditionally been considered an ill-circumscribed genus, lacking any obvious synapomorphy (Simpson 1990(Simpson , 1993(Simpson , 1998b. However, with the transfer of X. xanthorrhizon to Cubanicula, Xiphidium s.str. can be easily defined by its introrsely rimose, but functionally poricidal anthers (an adaptation to buzz-pollination; Buchmann 1980), the complete loss of septal nectaries (also an adaptation to buzz-pollination), capsules bright-coloured, indehiscent, lacking thickened septal ridges and somewhat fleshy at maturity (a possible adaptation to endozoochory) and cuboid seeds (Hickman 2019;Pellegrini 2019). All these characters are unique in the family and observed on the two species of Xiphidium accepted by us in the present study. The anther morphology of Xiphidium and its floral biology are reminiscent of some species of Dichorisandra J.C. Mikan (Commelinaceae, Commelinales) that also possess introrsely rimose but functionally poricidal anthers (Pellegrini and Faden 2017). However, studies on the reproductive biology of Xiphidium are non-existent, save that by Buchmann (1980). Further studies focusing on effective pollination and seed dispersal are necessary. The genus is well-documented as medicine for snakebite (Odonne et al. 2013) and has antimalarial and leishmanicidal properties (Valadeau et al. 2009). Xiphidium caeruleum also shows the most significant genetic divergence levels for any species of Haemodoraceae amongst populations across its wide Neotropical range (Hopper et al., in prep.). A further detailed taxonomic study is recommended, combining extensive fieldwork, molecular data, and traditional taxonomy. Nomenclatural notes. The taxonomic circumscription of X. caeruleum is greatly impaired by the lack of knowledge of the current whereabouts of the type material of several of its associated synonyms. Types for the names X. caeruleum and X. fockeanum were successfully located and designated by Maas and Maas-van de Kamer (1993), while types for the names X. rubrum, Eccremis scabra, and Durandia macrophylla were located by us and had lectotypes designated when necessary. Nonetheless, we have been unable to locate a type specimen, or illustration for X. giganteum, which prevents us from knowing if this name matches any of the X. caeruleum morphs recognised by us. Maas and Maas-van de Kamer (1993) erroneously designated plate 66 from Lindley (1846) as the lectotype of X. giganteum. The indicated plate actually depicts Swainsona greyana Lindl. (Fabaceae) and obviously cannot be the type for X. giganteum. In fact, the original publication (Lindley 1846) provides no illustration for X. giganteum. Lindley (1846) mentions that a live specimen was brought from Caraccas and flowered in Syon [Park], London, UK. After searching for specimens that matched these data at K herbarium, we were unable to locate any. We have also searched for a possible unpublished illustration that might serve as the type for X. giganteum, but were also unsuccessful. Thus, we are currently unable to designate a lectotype for X. giganteum, since this name completely lacks any original material (Art. 9.4., Turland et al. 2018). Since the original description is not enough to undoubtedly apply this name, we also feel it is premature to designate a neotype until natural populations from Caraccas have been studied. Finally, we also choose to tentatively retain it under the synonymy of X. caeruleum until further information becomes available.

Xiphidium caeruleum
As explained by Dorr and Wiersema (2010), Ixia xiphidium Loefl. represents a misinterpretation by Maas and Maas-van de Kamer (1993) of Loefling's (1758) publication. The author never intended to publish a new species but published a new genus, rejecting the application of Ixia L. for American plants. Thus, Ixia xiphidium Loefl. was never published and should not be included in databases.
When describing X. rubrum, Don (1832) mentions his new species is based on a Ruiz & Pavón collection, but without indicating a collection number or herbarium information. We came across a specimen matching the protologue with a label in Pavón's handwriting during a visit to BM, saying, "Ornithogalum rubrum sp. n., Fl. Per.". This specimen is here selected as the lectotype. Kuntze (1898) described Eccremis scabra, based on a collection from Río Juntas, Bolivia. The author mentions a specimen at B, but we were unable to locate it, and it might have been lost during WWII. Luckily, we were able to locate a duplicate at NY, which is designated here as the lectotype.
Distribution and habitat. Xiphidium caeruleum is widely distributed in the Neotropics, ranging from Mexico, reaching the Antilles, to northern South America (Fig. 25). It can be found growing in permanently or seasonally-wet environments, more rarely in dry and rocky environments.
Phenology. It was found in bloom and fruit throughout the year. Conservation status. As currently circumscribed, Xiphidium caeruleum is widely distributed, with equally wide EOO (14,922,959 km 2 ) and AOO (ca. 3,056 km 2 ). Thus, following IUCN's (2001) recommendations, X. caeruleum should be considered as Least Concern (LC).
Comments. Xiphidium caeruleum is a widely-distributed species and still a variable taxon even in our present circumscription. Despite our best efforts, we have been  unable to correlate any of the observed morphological variability to any of the previously proposed names in Xiphidium. After careful study of protologues, we concluded that X. loeflingii Mutis, X. caeruleum var. albidum (Lam.) Backer, X. floribundum var. albiflorum Hook., X. album Willd., X. albidum Lam.. and X. floribundum Sw. actually represent homotypic synonyms and are unambiguously conspecific with the type of X. caeruleum. Alternatively, Durandia macrophylla Boeckeler, Eccremis scabra Kuntze, X. fockeanum Miq. and X. rubrum D.Don represent heterotypic synonyms. Xiphidium giganteum Lindl. is tentatively kept here as a heterotypic synonym of X. caeruleum until further information on its type specimen is acquired.
All diagnostic characters provided by the original authors in their respective protologues can be easily observed in the typical morph of X. caeruleum. Some peculiar specimens of X. caeruleum are recorded for French Guiana (in which the specimens seem to present peculiarly large, red, crustose, and trigonous fruits), Costa Rica (where some specimens possess flowers with three inconspicuous green nectar guides at the base of the upper tepals) and Mexico (where specimens present inner tepals much longer than the outer tepals and perianth generally with apricot to pinkish hue). Furthermore, it is also known for berries of X. caeruleum to range from yellowish-orange to orange with reddish-orange spots, to completely red. We were unable to find any obvious correlation between the different colours of berries, geographical distribution, and the observed genetic diversity. Nonetheless, due to limited access to such morphs and also due to herbarium specimens in Xiphidium being generally poorly preserved, we consider it premature to recognise or propose any taxonomic status for these morphs. Thus, we propose that studies focusing on population genetics and reproductive biology, associated with a morphometric study and intense field studies, are necessary to properly deal with the issue. Diagnosis. Similar to Xiphidium caeruleum Aubl. in habit and inflorescence morphology, differing due to its leaves marginally ciliate at apex, apricot to light orange flower buds, larger and zygomorphic flowers, inner lobes obovate with obtuse to round apex, upper tepals connate in the basal third or halfway through with three orange-yellow to orange nectar guides, dark red to vinaceous mature capsules and dark reddish-brown to reddish-black seeds.
Distribution and habitat. Xiphidium pontederiiflorum is known to occur in Colombia, Ecuador, and Panama (Fig. 27), in the understorey in rainforests, generally near rivers, along streams, and other water bodies.
Phenology. Blooms and fruits from March to August. Conservation status. Xiphidium pontederiiflorum possesses a relatively narrow EOO (849,856 km 2 ) and AOO (ca. 132 km 2 ). Thus, following IUCN's (2001) recommendations, X. pontederiiflorum should be considered as Endangered [EN, A2ac+C2a(i)]. Comments. Xiphidium pontederiiflorum is morphologically similar to X. caeruleum in overall habit and inflorescence morphology. However, X. pontederiiflorum can be differentiated by its leaves marginally ciliate at apex (vs. glabrous in X. caeruleum), apricot to light orange flower buds (vs. white to cream, rarely apricot in Mexican populations), larger and zygomorphic flowers (vs. smaller and actinomorphic flowers), inner lobes obovate with obtuse to round apex (vs. elliptic with acute apex), upper tepals connate in the basal third or halfway through with three orange-yellow to orange nectar guides (vs. only basally connate and lacking nectar guides, rarely with green nectar guides in some Costa Rican populations), capsules dark red to vinaceous when mature (vs. orange to medium red) and dark reddish-brown to reddish-black seeds (vs. black). Added to that, X. pontederiiflorum is generally a more robust plant, growing erect up to 2 m tall, while X. caeruleum reaches up to 1 m tall, and its stems tend to lean due to the plant's weight, especially when in bloom or fruit.
Xiphidium pontederiiflorum was first collected in 1923 in Panama by the pioneering Neotropical botanist P.C. Standley (1884Standley ( -1963 from the United States (Williams 1963). Reference to it was included under X. caeruleum in Standley's (1928) Flora of the Panama Canal Zone.

Conclusion
The Neotropical species of Haemodoraceae represent morphological outliers in the family that have remained poorly studied for far too long, despite previous comprehensive studies dealing with macro-and micromorphology and the systematics of the Haemodoraceae (Simpson 1985(Simpson , 1987(Simpson , 1990(Simpson , 1993(Simpson , 1998a(Simpson , 1998bHopper et al. 2006Hopper et al. , 2009Smith et al. 2011;Aerne-Hains and Simpson 2017). Furthermore, most of its species dwell deep in the Amazon Forest, and key and enigmatic taxa, like Pyrrorhiza neblinae, are restricted to almost impossible to reach tepuis. This paper is the result of the author's combined efforts, as part of a global collaboration, hoping that these new data will update our current knowledge on Haemodoraceae and encourage further studies on the family, as well as in Commelinales.
All Neotropical Haemodoraceae are placed in subfamily Haemodoroideae and, except for Lachnanthes, are also placed in a well-supported clade by both molecular (Hopper et al. 1999 and morphological data (Simpson 1990;Pellegrini 2019). Ongoing studies seem to indicate the need to revisit the family's classification and formally recognise this clade, as well as others (Hopper et al., in prep.;Pellegrini 2019). A similar scenario is observed for the other families of Commelinales, where several systematic-based classification updates are still needed for several groups (Pellegrini 2019). Pontederiaceae is currently the most systematically up-todate family in the order, thanks to recent contributions (Pellegrini 2017;Pellegrini and Horn 2017;Pellegrini et al. 2018). Nonetheless, the remaining four families (i.e., Commelinaceae, Haemodoraceae, Hanguanaceae, and Philydraceae) are still in need of much updating.
Finally, the present study takes the first vital step towards standardising the morphological terminology used in Haemodoraceae. As part of the first authors' systematics studies in Commelinales (Pellegrini 2019), it became clear that much of the difficulty in finding morphological synapomorphies for the order, as well as its backbone and families, is related to the disparate terminology used in each of the five families. Thus, it is crucial for the descriptive terminology used for Commelinales to be standardised to enable the inclusion of morphology in phylogenetic studies. This standardisation also dramatically decreases the degree of homoplasy in the morphological dataset and increases its congruence with the molecular data (Pellegrini 2019). A publication focusing on the standardisation of the morphological terminology for Commelinales is in the works and should be published in the near future.