Synoptic taxonomy of Cortaderia Stapf (Danthonioideae, Poaceae)

Abstract Cortaderia (Poaceae; Danthonioideae) is a medium-sized genus of C3 tussock grasses, widespread in the temperate to tropic-alpine regions of South America. It is particularly important in the subalpine and alpine zones of the Andes. We revised the classification of the genus, and recognize 17 species grouped into five informal groups. We describe one new species, Cortaderia echinata H.P.Linder, from Peru. We provide a key to the groups and the species, complete nomenclature for each species including new lectotypes, and notes on the ecology, distribution and diagnostic morphological and anatomical characters.


Introduction
Cortaderia Stapf (Danthonioideae, Poaceae) is best known for the pampas-grass, Corta deria selloana (Schult. & Schult. f.) Asch. & Graebn., which is globally cultivated as a garden ornamental (Grounds 1979), but which is also an aggressive invader in many warm-temperate regions (Gadgil et al. 1990;Harradine 1991;Lambrinos 2001;Okada et al. 2007;Robinson 1984). However, the genus is also a major component of the temperate C3 grasslands of South America, from Tierra del Fuego in the south to Venezuela in the north, and from the Atlantic coastal mountains near Rio de Janeiro to the Equadorian Andes, and from sea level at the southern extreme to over 4500 m at the equator.
Th e genus was erected by Stapf (1897), initially to include only the large tussocks allied to Cortaderia selloana, which until then had been classifi ed under Gynerium Bonpl. In the fi rst decades of the 20 th century Pilger (1906) and Hackel (in Dusén 1909) transferred four species from Gynerium into Cortaderia. During the second half of the last century fi ve new species were described by Swallen (1948;1956). Recently, Linder et al. (2010) segregated fi ve New Zealand species into Austroderia N.P. Barker & H.P.Linder and a New Guinean species into Chimaerochloa H. P. Linder on the basis of morphology and molecular data, and included Lamprothyrsus Pilg., a genus of two species distributed in South America, in Cortaderia.
Th ere are several taxonomic problems in the genus. Species delimitations of the three species closely related to C. selloana present a major challenge, as already noted by Stapf (1897). Th ese have recently been revised by Testoni and Villamil (2014), based on fi eld and herbarium studies. Similar species delimitation problems are also evident in some of the Andean tussock species. Th ere are several very local species, apparently known only from the type localities -the validity of these species could be questioned. As well, there are several apparent segregates from currently accepted species; these may be due to "over-splitting" of taxa. Finally, in 2008 Paul Peterson and Rob Soreng collected a putatively new species of Cortaderia from Peru, which needs a formal description.
Th e reproductive biology of Cortaderia is complex, with apparently hermaphrodite, dioecious, gynodioecious and apomictic species. Connor (1965) showed that gynodioecy, rather than dioecy, is the common and widespread condition in Cortaderia. Th e variation in reproductive structure in the genus was explored in detail in Connor (1973). He showed several syndromes. Only C. sericantha (Steud.) Hitchc. was monomorphic, with the reproductive organs in the male and bisexual plants scarcely diff erentiated. Testoni and Villamil (2014) established that species of the Selloana group (Sect. Cortaderia sensu Conert 1961) are gynodioecious and apomictic (e.g. C. selloana can form extensive clonal populations, one of which gave rise to C. selloana subsp. jubata (Lemoine) Testoni & Villamil) or only apomictic species (e.g. C. speciosa (Nees & Meyen) Stapf), whereas the species in the rest of the genus are dimorphic and dioecious, except for C. hieronymi (Kuntze) N.P. Barker & H.P.Linder, which Connor and Dawson (1993) showed to be apomictic. Th e genetics of the gynodioecism was explored for C. selloana by Connor and Charlesworth (1989), who showed that it was probably controlled by a male-sterility gene, expressed in the female-only plants of the species. Th e reproductive system infl uences the patterns of variation in the species, and might account for the taxonomic complexity of some species.
Th e phylogeny of Cortaderia is as yet incompletely known. Phylogenetic analyses have been published by Barker et al. (2003) and Pirie et al. (2008), based on which the New Zealand and New Guinean species were separated into the genera Austroderia and Chimaeorochloa, whereas the South American Lamprothyrsus was included in the genus Cortaderia (Linder et al. 2010). Cortaderia in its current circumscription is monophyletic, and most closely related to the "danthonioid" clade of the Danthonioideae (Pirie et al. 2009).
Th is paper presents a critical review of the species limits in Cortaderia based on leaf anatomical features, investigation of fi eld populations and the study of herbarium specimens. We also clarify the nomenclature and typifi cation of all names in the genus, and provide a key to the species. A descriptive monograph of the whole subfamily is in preparation, and full descriptions will be published in that account, as well as the full lists of specimens examined.

Materials and methods
Th e morphological descriptions were compiled from the analysis of the available herbarium material at B, BA, BBB, CONC, CORD, K, LOJA, M, NY, P, RB, SI, SGO, QCA, US, W, Z and ZT (acronyms follow Th iers (continuously updated)). Where suffi cient material was available, spikelets were dissected, sketched and measured. Th e lemmas were mounted in glycerine, and drawn using a camera lucida. Anatomical investigation was based on leaf fragments of herbarium specimens (Table 1). Fragments ca. 1 cm long, from near the middle of blade, were fi rst softened and reconstituted in warm, soapy water for 20-40 min. Transverse sections were hand-cut at 20-40 μm, and epidermal scrapes were prepared of the abaxial surface. Sections and scrapes were diff erentially stained with a combination of Safranin Red and Alcian Blue (Tolivia and Tolivia 1987), dehydrated in an alcohol series, and mounted in Histomount. In addition, for Cortaderia selloana, C. speciosa and C. vaginata samples were fi xed in formalinacetic acid-alcohol, dehydrated in an ethyl alcohol-tertiary butyl alcohol series and embedded in Paramat. Th e study plant sections (20 μm) were stained with safraninfast green and mounted in Canada balsam. Th e anatomy of each species was described using the characters and character states proposed by Ellis (1976;. Th e assignation of holo-and lectotype status follows the analysis and recommendations of (McNeill 2014). Description. Gynodioecious, dioecious, hermaphrodite or apomictic perennials, ranging from rounded vegetable hedgehogs less than 0.5 m tall to erect 4 m tall tussocks; innovations intravaginal; spreading stolons rare. Leaf sheaths variable: persisting intact, or fragmenting transversely, or decaying into a tangled mass of fi bres, or occasionally persisting as burnt-off sheaths; glabrous or more rarely covered in a dense indumentum. Ligule of one or many rows of cilia, to 5 mm long. Leaf blades to 2 m long, tough, expanded, rolled or folded, occasionally pungent, usually persistent but occasionally disarticulating above the ligule, sometimes with an adaxial weft of hairs directly above the ligule; margins sometimes roughly scabrid and cutting. Infl orescences paniculate, sometimes compact but usually plumose, to 1 m long, many-spikeleted, pedicels and pulvini glabrous, scabrid or villous. Spikelets to 30 mm long, with 2-10 fl orets, disarticulating above the glumes, male spikelets usually less hairy than female spikelets and glabrous in the Selloana group; glumes glabrous, often papery or membranous, 4-30 mm long, usually 1-veined and rarely with no veins, upper and lower glumes similar. Lemmas ( Fig. 1) 3-7 nerved, mostly with the central three nerves continuing into a more or less twisted awn; the lateral nerves sometimes terminating in lateral bristles, the lemmas often continuing up the awns, consequently with the bristles apparently borne on the awn, in C. selloana the lemma continues to the tip of the awn and so obscures the awn; lemmas usually long-villous on the back, rarely glabrous. Palea membranous, linear, often longer than the lemma, keeled, sometimes variously villous on the back. Lodicules two. Anthers three, fertile or sterile, to 3.5 mm long. Ovary stalked, styles two. Caryopses 1.5-3.5 mm long, variable in shape, glabrous, embryo mark from ¼ to more than ½ length of caryopsis, hilum linear, from ¼ to ¾ caryopsis length. Leaf anatomy. Leaf in transverse section sclerophyllous, leaves varying from expanded to setaceous, margins not thickened but with a sclerenchyma cap. Adaxial furrows vary from deep and cleft-like to absent; abaxial ribs sometimes present. Vascular bundles diff erentiated into two, rarely three, orders; primary vascular bundles 6-30, symmetrically distributed in the two leaf sections; either ad-or abaxially or centrally positioned, circular or elliptical, sometimes with sclerosed phloem; outer bundle sheath cells always distinct from the chlorenchyma and sometimes lignifi ed, entire or interrupted by bundle sheath; adaxial sclerenchyma as narrow girders, as trapezoidal girders, as T-shaped girders or inversely anchor-shaped girders; abaxial sclerenchyma as small strands, as narrow girders, as wide girders, as trapezoidal girders, or as massive linked girders forming a continuous subepidermal layer; tertiary vascular bundles 1-several between the primary vascular bundles, adaxial sclerenchyma as small strands, as narrow girders, as trapezoid girders narrowing towards vascular bundles, as T-shaped girders or inversely anchor-shaped girders; abaxial sclerenchyma absent, as small strands, as narrow girders, as broad girders, as trapezoidal girders or as massive linked girders forming a continuous subepidermal layer. Mesophyll of small, angular isodiametric chlorenchyma cells with small air spaces; mesophyll islands of colourless cells usually absent, sometimes with colourless collenchyma cells connecting the adaxial and abaxial furrows and so partitioning the chlorenchyma. Abaxial subepidermal layer sometimes with collenchymatous or non-chlorophyllous cells in 1-several layers only along the margins, or fl anking the midrib, and sometimes with this layer extending over the whole width of the leaf. Abaxial epidermal zonation present or absent; microhairs or macrohairs absent; silica bodies absent, or tall and narrow, or round and single. Adaxial epidermis sometimes with papillae, pricklehairs, and microhairs.
Distribution and ecology. Widespread in South America, from Tierra del Fuego (Argentina) to Venezuela, from Brazil to Peru, from sea level to the Páramo.

Systematics
We arranged the species into fi ve informal groups, which are coherent morphologically and anatomically.

Key to the species (anatomical characters in brackets)
1 Lemma body continued up the awn, for at least the same length as the expanded portion of the lemma; plants forming massive tussocks to 4 m tall, infl orescences plumose (leaves with abaxial groves (Fig. 2a-c) Old leaf sheaths intact, or shattering transversally, rarely some lacerated (sometimes in C. boliviensis); (leaves, except in C. echinata, with a multilayered wide collenchyma below the adaxial epidermis and no sclerenchyma girder connecting the vascular bundle to the epidermis, Fig. 3(1) In this group three morphologically and anatomically similar species, with both gynodioecious and apomictic breeding systems, are included. Th ey are easily distinguishable from other species in the genus: they form big tussocks 1.5 to 3 m in diameter and up to 4 m in height, and the leaf edges are strongly cutting. Th e panicles are large and plumose, very showy, and much larger than in most of the other species. Th e spikelets have 1-veined glumes, the lemmas are long-acuminate, with or without evident awns, unlobed, 3-veined and with long hairs only in female plants (hermaphrodites glabrous). Th e leaf, in transversal section ( Fig. 2A-C), is ribbed, with moderately deep square ribs on adaxial and ribbed on abaxial surface. Colourless chlorenchya cells occur between the vascular bundles on the abaxial epidermis. Aerenchyma is sometimes present (Fig. 2B)  A. dioeca Spreng. is cited in the protologue, and the diagnoses are identical. Furthermore, both description cite a Sellow collection, without number, from Montevideo. It is most likely that the type is Sellow 570 from Montevideo, which is in B, and is designated here as lectotype. Curiously, Conert (1961) proposed Sellow 396 from Brasilia as holotype of A. dioica, although Sprengel explicitly mentions that the type is from Monte Video. Gynerium argenteum Nees is also based on the same collection as Arundo dioica Spreng., plus some additional material. All three names are based on the same type.
Common names. pampas grass, cortadera, cola de zorro, carrizo de las pampas. Th e origins of the popular name "pampas grass" are somewhat obscure, and do not refl ect the ecology of the species (Stapf 1897). Taxonomy. Cortaderia selloana ssp. selloana can be diagnosed by the glumes about as tall as the basal lemma, and lemma without a distinct awn. Th e plants are generally larger than those of C. araucana and C. speciosa, and the panicles are larger (0.5 to 1 m long), more lax, and coloured white, pink or yellowish. Th e similar size of basal lemmas and glumes (6-15 mm) further separates it from C. araucana (glumes 9-17 mm long, ca. ½ length of basal lemmas), whereas the larger glumes separate it from C. speciosa (glumes 6-8 mm, ca. ¾ length of basal lemmas). Th e large size may also Structures referred to in the descriptions are labelled as follows: 1, multi-layered abaxial sub-epidermal collenchyma layer; 2, aerenchyma; 3, chlorenchyma; 4, primary vascular bundle; 5, midrib; 6, colourless cells; 7, empty cells; 8, bulliform cells. lead to confusion with C. nitida, but it is easily separated by the larger and laxer panicles, 3-veined, awnless lemmas that are glabrous on hermaphrodite plants; and female plants with tiny staminodes. For the distinction from ssp. jubata see below.
Cortaderia selloana ssp. selloana was originally described as dioecious, but Astegiano et al. (1995) showed that it is gynodioecious, and Testoni and Villamil (2014) recorded several populations with only pistillate individuals (so presumably apomictic) in central and northern Argentina. Th is subspecies presents the greatest morphological variability and geographical range in the genus. Th e morphological characterization is also complicated by interbreeding between natural populations and cultivated plants.  Etymology. jubata (Lat.): Having mane, crest, in allusion to the panicle. Common names. pink pampas grass, jubata grass, cortadera Taxonomy. Th is subspecies is generally similar to ssp. selloana, and includes all the morphologically homogenous apomictic populations of the Yungas region. It can be separated from ssp. selloana by the infl orescences which extend far beyond the foliage, and the pink, 75-90 cm long, very lax, pyramidal and nodding panicles. In Ecuador it is sympatric with C. nitida, from which it can be separated by its larger size and its spectacular pink panicles. Th ey can also easily be distinguished by the leaves: in subsp. jubata they are fl at and folded V-shaped, while in C. nitida leaves are inrolled from both margins. Etymology. -ana, indicating connection. From the Araucania region of Chile. Common names. cortadera Taxonomy. In the Selloana group, C. araucana is readily diagnosed by the basal lemmas longer than 12.5 mm and much longer than the glumes. Th e spikelets are 20-35 mm long and the lemma of the basal fl oret 14-25 (30) mm long (including awn of 5-11 mm long). Th e species is found in the southern (austral) Andean region.

Cortaderia araucana
Cortaderia araucana includes extensive morphological variation, and both gynodioecious and apomictic populations. Th is variability led Acevedo Vargas (1959) to recognize three varieties, which are no longer maintained. In northern Patagonia C. araucana and C. selloana are sympatric, but the plants of C. araucana are somewhat smaller, with less lax panicles and fl owering in the austral spring (late November and early December), whereas C. selloana fl owers in the austral summer (January and February). Further, the spikelets are diff erent: the glumes are shorter than the basal fl oret, the lemma may terminate in an awn that arises between two lower lateral setae. Th e leaf anatomy of both species is similar.  (1834), from Copiapo (Chile) and Lake Titicaca (Peru), respectively -are synonyms of Cortaderia speciosa, but are invalid (nomina nuda) as no descriptions were published. Th eir identity can be determined, because the specimens in B! were annotated with the Meyen names. Gynerium speciosum was validated by Nees in 1943. Tropicos (Downloaded 14 December 2016) lists the species as described by Nees in 1841 (Nees ab Esenbeck 1841), but this is erroneous. Conert (1961) designated Philippi 1024 (B photo!) from Chile ("Atacama oppidum, 1824") as lectotype of Gynerium atacamense Phil. However, the type has been found in the herbarium SGO (Connor, 1983)  Common names. cortadera Taxonomy. In the Selloana group, C. speciosa can be diagnosed by the short basal lemmas, which are less than 13 mm long. Th e spikelets are 8-15 mm long and the basal lemma 7.0-12.5 mm long (including awn, 1-4 mm). It diff ers from other species in the group by its very compact, bright brown panicles with ascending, short and stiff branches. Th e species is readily distinguished by the small fl oret sizes. Th e leaf anatomy is also somewhat diff erent from the other species of the group (Fig. 2C): the midrib is rounded and somewhat lower; the outer sheath of the central vascular bundle without projections to the adaxial epidermis; and with a massive abaxial sub-epidermal collenchyma layer, only in the middle part of the leaf. Th e latter occurs in the Nitida group but along the leaf. It is known only by pistillate plants from desert regions (the Puna) of Argentina, Bolivia and Chile.

Cortaderia speciosa
Th is species is completely apomictic, and several morphological subgroups can be recognized. As these are all apomicts, it is presumed that they derive from the same ancestral sexual population. Th e material previously separated as C. rudiuscula has longer (9-12 mm) and more slender lemmas, than the material previously separated as C. speciosa (lemmas ca. 8 mm), but there is no clear separation between these two forms.

Lamprothyrsus group
Th is group is very distinct within Cortaderia. Morphologically, it diff ers by the long, fi liform awns, 14-35 mm long; glumes without veins; and by the sheaths which are always intact. Furthermore, the leaf anatomy diff ers by the primary vascular bundles with lignifi ed sheaths and girders, tertiary vascular bundle sheaths and girders collenchyma (Fig. 3C). Th e group includes only one species. Th e enormous variation with this species complex could be due to its apomictic reproduction (Connor and Dawson 1993). Taxonomy. Th is species contains substantial variation in the robustness of the plants. Conert (1961) partitioned this variation into three species (L. peruvianus, L. venturi and L. hieronymi) and Pilger (1906) recognized varieties in his L. hieronymi. Study of the herbarium material suggests that this is most likely all one taxon (Bernardello 1979), but an analysis of variation within natural populations in the fi eld would be useful to understand the range of variation possible. Cortaderia hieronymi diff ers from the other species in Cortaderia by the very long hair-like lemma awns and setae, the glumes without veins, and the small fl owers with relatively short and sparse lemma hair.

Cortaderia hieronymi
Only apomictic populations are known, but a few fertile staminate specimens with long hairs on the lemmas were found (Bernardello 1979). It is not known if they can form viable caryopses, and if the species is dioecious or gynodioecious.
In the central and northern Argentina to Ecuador C. hieronymi is sympatric with the two subspecies of C. selloana, but it is easily separated by its smaller panicles, spikelets with glumes without veins, and 5-veined, 3-awned lemmas. In Peru and Ecuador it is sympatric with C. bifi da, with which it is often confused: in both species the old leaf sheaths are lacerated and the spikelets have long awns, but the spikelets of C. hieronymi are bigger, and the lemmas with longer and robust central awns.

Egmontiana group
Th is group includes three quite distinctive species. Cortaderia vaginata and C. modesta have an unusual (for Cortaderia) leaf anatomy lacking ribs, and with deeply split phloem poles (Fig. 3B), and large bulliform cells (Fig. 2E-G), which are rare in the other groups.  (1829) described Ampelodesmos australis, and explicitly included Arundo pilosa D'Urville as a synonym, noting that this species is better placed in Ampelodesmos.

Cortaderia egmontiana
Taxonomy. Th e species can be readily diagnosed by the combination of compact infl orescences, almost glabrous leaves, and either no, or poorly developed, awns and setae on the lemmas. Th e habit and dense infl orescences are as in C. sericantha, but C. egmontiana diff ers by the absence of setae, and by the almost completely glabrous leaves. Th e lemma and spikelet morphology (reduced or absent awns and setae) suggests an affi nity to the eastern Brazilian species C. vaginata and C. modesta. From these two species C. egmontiana can be separated by the compact infl orescences and the tendency of the leaf blades to disarticulate from the sheaths. It is the only Cortaderia species in southern South American temperate zone. Th e leaf anatomy (Figs 2E, 3A) does not show any distinctive peculiarities.
Th ere is remarkable intraspecifi c variation in the spikelet and fl oret sizes, and Conert (1961) separated the forms with smaller spikelets as C. minima. Moore (1983) suggested that the two taxa were latitudinally separated, with the southern populations constituting C. pilosa, and the northern C. minima. On the available material, there is indeed a break in the glume length variation. However, this fi ts no ecological or geographical pattern, and both small and large-glume forms occur in both the Falkland / Malvinas islands and Tierra del Fuego. Further north, indeed, only the small-glume form is found. Th is suggests that this size variation has no biological signifi cance, accordingly it is ignored here. Etymology. modesta (Latin) = moderate, presumably referring to the culms of average height. Nomenclatural comments. Th e locality information given by Connor and Edgar (1974) is incorrect. Note that Glaziou made several collections of the same species from the same area.

Cortaderia modesta
Common names. cabeça de negro, capim-de-anta. Taxonomy. Some specimens show a poorly developed axillary infl orescence developed at the penultimate node of the fl owering culm. Th e almost awnless lemmas, with the paleas as long as the lemmas, and the very dense callus hairs compared to the short lemma back hairs, are almost unique in the genus. Its closest relative might be C. vaginata from Santa Catarina, further south along the Brazilian Atlantic coast. It is readily distinguished from C. vaginata by the persistent leaf sheaths and the awnless lemmas. According to herbarium labels the plant forms massive tussocks with persistent red, burnt sheaths. Etymology. vagina (Latin) = sheath. Possibly referring to the conspicuous leafsheaths, a feature that is common to most of the genus.

Cortaderia vaginata
Common names. Penacho, Capim-Penacho. Taxonomy. According to Swallen (1956) this species resembles C. parvifl ora, but diff ers by the glabrous lemmas and long-villous calli. It is unusual among the species assigned to Cortaderia by the glabrous lemma, and the almost glabrous pedicels and infl orescence axes. Th is species may be a local endemic, and might be quite rare. It is probably most closely related to C. modesta, which also has a reduced awn, but diff ers by the sheaths which are lacerated, lax panicles (without axillary panicles), and the glabrous lemma. Geographically it can be immediately identifi ed as the only Cortaderia species from Santa Catarina in southern Brazil.
Th e leaf anatomy is identical to that of C. modesta, except that all sections appear to have large empty cells in the middle of the leaf (Fig. 2D), between the vascular bundles. Th ese were seen on some sections of C. modesta, but rarely.

Nitida group
Th e Nitida group can be characterized by the leaf sheaths which generally remain intact, and the leaves which, in transverse section, show a massive abaxial sub-epidermal collenchyma layer. Cortaderia nitida, C. pungens and C. boliviensis are very similar, whereas C. sericantha is quite distinct by the villous, folded leaves with no adaxial ribs. Th e distinction of C. pungens is not clear, and needs fi eldwork. Th e new C. echinata is also included in here although anatomically it fi ts into the next group. Leaf anatomically, C. nitida and C. boliviensis are very similar, with papillate adaxial surfaces, deep adaxial grooves, and a well developed abaxial collenchyma layer. Th e leaf anatomy of C. pungens is not known. Etymology. niteo (Latin) = shine. It may refer to the persistently intact, more or less white, leaf sheaths. Common names. "Sigse de Páramo". Taxonomy. Cortaderia nitida is a distinctive grass. It is the tallest and most robust species of this group. Th e lamina margins are inrolled. Th e basal sheaths gradually become shorter with age, but do not become lacerated, the leaf blades are scabrid in the upper half but not the lower, and the infl orescence branches which are scaberulous while the pulvini often have a few long hairs (the latter seems to be unique in the genus). Th e callus usually has very long spreading hairs (more than 2 mm, almost equivalent to the lemma hairs), and the setae are less than 2 mm long. Th e other tall Cortaderia, C. bifi da, has central awns that are longer than 8 mm, and very well developed setae. Th e lemma shape is similar to C. columbiana, but the infl orescence branches are scaberulous in C. nitida, and villous in C. columbiana.

Cortaderia nitida
Th is species also approaches the Selloana group by it large size, big plumose infl orescences, and especially by the lemma shape. It is easy to confuse the lemmas of the two groups, but in Nitida group the lemmas are 5-7 veined, hairy in both sexes, while in Selloana group the lemmas are 3-veined, hairy in female plants and glabrous in hermaphrodite plants. Th e plastid sequence data also places this species as sister to the Selloana group, but this is not corroborated by the ITS-based phylogeny. Laegaard (1997) mentions a distinct form of smaller and more delicate plants from the province of Azuay in Ecuador, and with three-nerved glumes, but we have not seen any material of it.
Th e leaf anatomy (Fig. 3D) is similar to that of C. boliviensis, and approaches that of C. sericantha. A well-developed layer of collenchyma is found below the abaxial epidermis, and overall there is little evidence of lignifi cation. It diff ers from C. sericantha by the well-developed adaxial grooves and the not quite so massive collenchyma, and by the presence of adaxial papillae. Etymology. -ense (Latin), denoting origin. From Bolivia. Taxonomy. Th is species is very similar to C. nitida, with which it shares the (usually) non-lacerated, entire leaf sheaths and the shape of the lemmas, as well as largely similar leaf anatomy. However, neither chloroplast nor nuclear genome indicates such a relationship for C. boliviensis (Pirie et al. 2009). It diff ers by the horizontally shattering sheaths. More inconsistent diff erences are in the indumentum of the fl oret, with the callus indumentum of C. boliviensis being shorter than in C. nitida. Lyle (1996) diagnosed C. boliviensis against C. bifi da, under which it was originally described as a variety by Henrard in 1921. Mostly it is very diff erent from C. bifi da: the latter has much longer lemma setae and the basal sheaths are lacerated and not shattered. Th e type collection, however, is easily confused with C. bifi da due to the long awns and setae, and somewhat fragmented leaf sheaths. Th e leaf anatomy is also quite diff erent.

Cortaderia boliviensis
Th e leaf anatomy (Fig. 3E) is like that of C. nitida, with adaxial grooves and a well-developed abaxial collenchyma layer. Th ere are diff erences in detail, and wider sampling may well indicate that this is within-species variation. Etymology. serios (Greek) = silken + Anthos (Greek) = fl ower. Presumably this refers to the silky-haired leaves, a diagnostic trait for this species. Taxonomy. Th is species is very distinctive in Cortaderia by its very villous leaves, which are rolled rather than fl at, and quite pungent; the compact infl orescences with short infl orescence branches; the glumes with three veins and which are much longer than the packet of fl orets; and the tuft of hair at the base of the spikelets. Th e infl orescences are similar to those of C. egmontiana, but the villous leaves immediate distinguish our species from C. egmontiana. Th e intact leaf sheaths, pungent leaf tips, and compact growth form related this species to C. pungens and C. echinata. Th e remarkably large glumes, much overtopping the packet of fl orets, are shared with C. echinata.
Th e leaf anatomy (Fig. 3F) could be unique in the genus. Th e abaxial half of the leaf, in cross-section, consists of colourless collenchyma. Th e vascular bundles are very slender, and the girders taper towards the adaxial epidermis. Adaxially the leaves are only very slightly grooved. Etymology. pungens (Latin): piercing, terminating in a sharp point. Th is describes the leaf tips. Taxonomy. Th is species is often placed with C. hapalotricha, from which it diff ers by (a) shorter growth-form (less than 1 m tall); (b) the intact leaf bases; (c) the rolled, pungent leaves; and (d) deeply lobed lemmas. Th e two species have much in common (leaf anatomy, spikelet and infl orescence structure). It is possible that they are ecotypes of each other, and the problem needs critical fi eld work. We keep them separate on the very diff erent growth-form. Th e intact leaf bases and pungent leaves suggest a relationship to C. sericantha and C. echinata, but the species is readily separated from these two by the much shorter glumes.

Cortaderia pungens
Th e leaf anatomy was not studied. Diagnosis. Similar to C. pungens by the small compact habit and pungent leaves, but diff ering by the shattering leaves and the longer spikelets.
Leaf anatomy. Leaf in transverse section expanded, sclerophyllous; margins gently tapering, sclerenchyma caps well-developed; adaxial furrows located between all vascular bundles, the same over primary and tertiary vascular bundles, about half depth of leaf, forming narrow clefts, ribs fl at-topped; abaxial ribs and furrows present. Vascular bundles closer to abaxial surface, 3 primary vascular bundles in half a leaf section, with 1-2 tertiary vascular bundles between the primary vascular bundles. primary vascular bundles elliptical; phloem without lignifi ed cells; metaxylem vessels narrower than outer bundle sheath cells; outer bundle sheath clearly distinct from chlorenchyma, cells larger and colourless, with adaxial and abaxial interruptions; inner bundle sheath walls thickened anticlinally, cells smaller than outer bundle sheath cells; adaxial sclerenchyma as inversely anchor-shaped girders; abaxial sclerenchyma as trapezoidal girders. tertiary vascular bundles outer bundle sheath cells distinct from and larger than chlorenchyma cells, walls thickened anticlinally or all round; with abaxial interruption only; adaxial bundle sheath extension present with cells smaller than outer bundle sheath cells; adaxial sclerenchyma inversely anchor-shaped girders; abaxial sclerenchyma as trapezoidal girders; phloem without lignifi ed cells or with only the inner bundle sheath lgnifi ed. Mesophyll of small, angular isodiametric chlorenchyma cells with small air spaces. Abaxial epidermal cells all larger than adaxial ones; outer wall twice as thick as inner wall; walls equal to mesophyll walls. Subepidermal layer of sclerifi ed fi bres only in marginal regions of leaves, absent from the middle of the leaf (directly next to leaf margins), 2-3 cells thick; with large clear parenchymatous cells below abaxial furrow present, connected via collenchyma cells to the adaxial furrow to the epidermis and so partitioning the chlorenchyma. Bulliform cells absent; abaxial epidermal zonation present (Fig. 3G).
Etymology. echinus (Latin) = hedge-hog or sea-urchin. Th e plant is spiny like a hedgehog.
Habitat. Rock ledges (bedrock slabs); moisture regime: in soil pockets on rock. Forming cushions on almost fl at rock slabs, in pockets of soil.
Conservation status. Known only from the type collection. Phenology. Flowering month March or April. Taxonomy. Th e small compact hedgehog form with pungent leaves is similar to C. pungens, from which it diff ers by the shattering leaves and the longer spikelets (glumes 15-25 mm long). Th e shattering leaf-sheaths link the species to C. boliviensis, but it diff ers by the very diff erent growth form. Th e compact infl orescences are reminiscent of C. egmontiana, but the pungent leaves provide a simple diagnostic diff erence.
Th e leaf anatomy is reminiscent of that of C. bifi da, but the outer bundle sheath is not lignifi ed, and form an extension adaxially on the vascular bundles, connecting them to the lignifi ed anchor-shaped girders.

Bifida group
Th e leaf sheaths of this group are highly lacerated and form a tangled mat around the base of the plant. Anatomically there is nothing unusual about these species. Th e distinction between C. columbiana and C. roraimensis needs critical investigation. Etymology. bis (Latin) = twice + fi du, divide, this presumably refers to the lemma setae. Taxonomy. Th is species can be diagnosed by the combination of the lacerated sheath bases, the long awns and especially the long setae. Th e shape of the lemmas with lobes and setae are shared with C. hapalotricha, and the curly fi brous leaf remains are similar to C. roraimensis. It is separated from C. roraimensis by the hairy lemmas and by the much longer awns and setae. It diff ers from C. hapalotricha by the glabrous adaxial surface above the ligule and the scaberulous infl orescence branches. From the other tall species, C. nitida, it can be separated by the longer awns (more than 8 mm long). Th e long awns and setae result in the infl orescences looking similar to those of C. peruviana, but the bases of the plants are quite diff erent. Consequently, it can be diffi cult to determine collections which consist only of infl orescences.

Cortaderia bifi da
Th e leaf anatomy (Fig. 3H), in transverse section, shows shallow abaxial groves and deep adaxial clefts. Adaxially there appear to be no papillae (diff erent from the C. hapalotricha anatomy). Abaxially below the epidermis are large colourless cells. Th e outer bundle sheath of the primary vascular bundles are completely lignifi ed. Th us broadly similar to the C. halalotricha anatomy, but diff ering in a number of traits. Etymology. planus (Latin) = fl at + folium (Latin) = leaf. Leaf-blades fl at. Taxonomy. Cortaderia planifolia has many similarities to C. pungens, but is separated by the fl at or folded, but not rolled, leaves; somewhat taller tussocks (05-1 m, compared to 0.2-0.5m); adaxial leaf surface above the ligule glabrous; glumes 8-15 mm long, compared to 12-16 mm; lemmas 4-8 mm, compared to 3-4 mm long; lemma awn less than 8 mm long, compared to more than 9 mm in C. pungens. Th ese numerous small diff erences suggest that these are two species.

Cortaderia planifolia
It has also been grouped with C. hapalotricha, from which it diff ers by the smaller size, the fl at leaves glabrous above the ligule, the shorter lemma awn and setae.
Leaf anatomy not investigated. Etymology. hapalos (Greek) = soft + thrix (Greek) = hair. It presumably refers to the densely pubescent rhachilla. Nomenclatural comments. Th e type specimen of Cortaderia scabrifl ora is intermediate between C. hapalotricha, C. pungens and C. planifolia. It has the lemma structure of C. pungens, the folded leaves typical of C. planifolia, the pungent leaves typical of both, but the size of C. hapalotricha. Overall, it approaches C. hapalotricha.

15.
Taxonomy. Connor and Edgar (1974) note "Th e golden brown panicles with very hairy branches are obvious characteristics of this species.", but these characters are variable in the species. Cortaderia hapalotricha is morphologically very close to C. columbiana, especially by the inner leaf surfaces directly above the ligule being densely and fi nely woolly. Genetically, the two species are strongly supported as sister species. Cortaderia hapalotricha can be separated from C. columbiana by the longer glumes, which are much longer than the spikelets, by the denser infl orescences, and by the lemmas which have well developed setae. It is also similar to C. bifi da, but the lemmas are longer and the setae shorter. Most convincing might be anatomical diff erences, these need to be corroborated with more sections. Th e leaf anatomy and spikelet structure indicate a very close relationship with C. pungens, and the two might just be ecological variants of each other. However, the growth form is quite diff erent, and we keep them separate on this basis.
Leaf anatomically (Fig. 3I) this species is very similar to C. columbiana, with well developed adaxial ribs, and girders linking the vascular bundles to both surfaces, as well as well developed adaxial epidermal papillae. Th e only diff erence may be the absent or poorly developed abaxial subepidermal sclerenchyma layer. Etymology. -ana, indicating connection. From Republic of Colombia. Taxonomy. Connor & Edgar (1974) imply a similarity to C. hapalotricha, but note that the panicle is longer, more laxly fl owered, and dull brown, and that this separates the two species. Cortaderia columbiana is superfi cially similar to C. hapalotricha, and also has short felty hair on upper leaf surface above the ligule, but is diff erent by the shorter setae. Leaf anatomically they can be separated by the presence of a continuous lignifi ed sub-epidermal layer on the abaxial side. It is also very similar to C. roraimensis by the lemma shape, in particular with the very short setae. However, the plant bases diff er: in C. roraimensis the leaf bases are lacerated and curly, a feature less well developed in C. columbiana. Possibly the best way to separate the two species might be by the much more villous leaf margins, and often the villous adaxial leaf surface of C. columbiana. Geographically, the two species are also adjacent.

Cortaderia columbiana
Th e leaf anatomy is like that of C. hapalotricha, but diff ers by a continuous sclerenchyma layer below the abaxial epidermis. Cortaderia roraimensis (N.E.Br.)  Etymology. -ensis (Latin), denoting place of origin. From Mt Roraima, Guyana. Taxonomy. Th is is the only Cortaderia species from the tepuis. It is very similar to C. columbiana. It shares with C. columbiana and C. bifi da a base of dense clustered lacerated sheaths. From the similar C. columbiana it is separated by the almost absent indumentum on the leaf margin directly above the simple ligule. From C. bifi da it is distinct by the lobed lemma, where the lobes are not extended into slender setae.

17.
Th e leaf anatomy (Fig. 3J) follows the same basic plan as that of C. hapalotricha.