Eremiolirion, a new genus of southern African Tecophilaeaceae, and taxonomic notes on Cyanella alba

The generic affiliation of Cyanella amboensis Schinz has been uncertain since the species w as excluded from the genus Cyanella L. in 1991. The species has two leaves, a divaricately branching inflorescence, ebracteolate pedicels, and actinomorphic flowers with monomorphic anthers. It is endemic to the western parts of central and northern Namibia. Other species of Cyanella have several leaves, racemose inflorescences, bracteolate pedicels, zygomorphic flowers with dimor­ phic anthers, and are endemic or near-endemic to the winter rainfall region in southwestern South Africa and southern Namibia. These differences are consistent with the recognition of the species as a distinct genus within the family. Phylogenetic analysis of plastid DNA sequences indicates that C. amboensis Schinz is sister to the other species of Cyanella, a relationship that also supports its independent generic status. The monotvpic genus Eremiolirion is according­ ly erected to accommodate the species. Minor differences in flower colour and vegetative morphology in Cyanella alba L.f. are shown to correlate with the three disjunct groups of populations in which the species occurs, and these populations are recognized at the level of subspecies. Tecophilaeaceae are a small family of geophytic peren­ nials well circumscribed by their cormous rootstock, poricidal anthers and semi-inferior ovary (Simpson & Rudall 1998). The flowers of all species are attractive, brightly coloured and often fragrant (Manning etal. 2002). The fam­ ily comprises 24 species, currently distributed among eight small genera, three of which are monotypic. The genus Odontostomum Torr. (1 sp.) is endemic to California, Conanthera Ruiz & Pav. (5 spp.). Tecophilaea Bertero ex Colla (2 spp.) and Zephyra D.Don (1 sp.) are restricted to Chile, and the remaining four. Cyanastrum Oliv. (3 spp.), Cyanella L. (7 or 8 spp.), Kabuyea Brummitt (1 sp.) and Walleria J.Kirk (3 spp.), occur in sub-Saharan Africa (Simpson & Rudall 1998). An analysis of plastid sequence data for the rbcL gene and the trnL-F region for all genera and most species (18), places the three Chilean genera as sister to the Califomian-African clade and retrieves all genera as monophyletic (Forest & Manning 2005 in prep.). The African genera have been thoroughly mono­ graphed (Carter 1962; Scott 1991; Brummitt et al. 1998). Among them. Cyanastrum and Kabuyea are found only in the tropics, Walleria is widely dispersed through sub­ tropical Africa, and Cyanella is almost restricted to the winter rainfall region of South Africa in the extreme southwest of the continent. A single species of Cyanella, C. amboensis Schinz is found further north, in central Namibia. This species is morphologically anomalous among the rest of the species of Cyanella and was excluded from the genus by Scott (1991) pending a reap­ * Compton Herbarium, South African National Biodiversity Institute, Private Bag X7. 7735 Claremont. Cape Town. ** Leslie Hill Molecular Systematics Laboratory. South African National Biodiversity Institute. Private Bag X7. 7735 Claremont. Cape Town & Department of Botany. University of Cape Town. Private Bag. 7701 Rondebosch, Cape Town. *** National Herbarium of Namibia. National Botanical Research Institute. Private Bag 13184. Windhoek. MS. received: 2005-04-05. praisal of its relationships within the family. A label appended to the specimen Bean. VIok & Viviers 1824 (BOL) indicates that at the time she had considered that the species might be best transferred to the Chilean genus Conanthera, with which it shares a branched inflores­ cence and symmetrical anthers, but the similarity is not actually that close. Conanthera has numerous leaves and a zygomorphic, blue to purple perianth, in contrast to the bifoliate condition and actinomorphic. white perianth of C. amboensis. Currently, therefore, the species is of uncertain generic affinity. Molecular evidence places C. amboensis as sister to a monophyletic clade comprising the remaining species in the genus (six of the seven remaining species are included in this study). This topol­ ogy offers two options for resolving the generic position of C. amboensis'. the species can either be incorporated within an expanded circumscription of the genus Cyanella; or a new genus should be erected to accom­ modate it. The morphological differences between Cyanella s. str. and C. amboensis are substantial, includ­ ing leaf number, inflorescence structure, floral symme­ try, and arrangement of the androecium. Genera in Tecophilaeaceae as currently circumscribed are homoge­ neous assemblages of closely related species in which floral symmetry in particular plays a defining role. Including C. amboensis within Cyanella would not only enlarge the circumscription of the genus to the extent that it is no longer morphologically coherent but would ren­ der it uniquely heterogeneous in comparison to the other genera in the family. Cyanella s. str. is circumscribed on the basis of its tunicated corm, several to numerous leaves, zygomor­ phic flowers with dimorphic anthers, and bracteolate pedicels. All of the species are endemic to or centred in the winter rainfall region of southern Africa. Cyanella amboensis. in contrast, has just two leaves, the flowers are truly actinomorphic with monomorphic anthers, the pedicels are ebracteolate. and the species is distributed in summer rainfall Namibia. On the basis of these several 116 Bothalia 35,2 (2005) FIGURE 1.—Eremiolirion amboense, Mannheimer 2510 (NBG). A, whole plant; B, flower, front view; C, inner tepal; D, outer tepal; E, androecium; F, single anther, adaxial surface; G, half flower detail; H, tip of anther; I, detail of mouth of floral tube showing corona; J, capsule; K, seed. Scale bar: A-D, J, 10 mm; E-G, 2.5 mm; H, 3.75 mm; I, 1 mm; K, 2 mm. Artist: J.C. Manning. Bothalia 35,2 (2005) 117 morphological and ecological differences, the species is most appropriately accommodated in a separate genus closely allied to but distinct from Cyanella s. str., for which we propose the name Eremiolirion. 1. Eremiolirion J.C.Manning & F.Forest, gen. nov. Species unica, a speciebus Cy.anellae duobusfoliis, pedicellis sine bracteolis, floribus actinomorphis, antheris monomorphis distinguenda. TYPE.—Eremiolirion amboense (Schinz) J.C.Manning & C.A. Mannheimer (= Cyanella amboensis Schinz). Deciduous geophyte with deep-seated, subglobose to oblate tunicated cormii tunics decaying into firm-leathery, coarsely netted fibres extending into neck. Cataphyll pale membranous. Leaves 2, basal, narrowly lanceolate, canaliculate with prominent midrib abaxially, leathery. Inflorescence divaricately branching, paniculate cyme; bracts subtending branches and pedicels only; pedicels cemuous at tip, elongating slightly in fruit and straight­ ening. Flowers actinomorphic, nodding, campanulate, white flushed abaxially with pink or maroon; tepals 6 in two whorls, connate below into short tube with minute, fringed corona present at mouth of tube; tepals dimor­ phic, outer oblong, inner pandurate. Stamens 6, monomorphic, symmetrical, attached to perianth near mouth of tube; anthers basifixed, erect and connivent around style, narrowly lanceolate, dehiscing by oblong apical pores. Ovary half inferior, trilocular; ovules sever­ al per locule; style terete, erect, straight, slightly taper­ ing; stigma minute. Capsules ovoid to globose. Seeds ellipsoid-pyriform, blackish brown, testa surface rugose. Distribution: a single species in central and north­ western Namibia. Etymology: the name is a compound of the Greek eremios (desert or wilderness) and lirion (white lily). Eremiolirion amboense (Schinz) J.C.Manning & C.A. Mannheimer, comb. nov. Cyanella amboensis Schinz in Bulletin de THerbier Boissier, sér. 2, 2: 943 (1902). Type: South West Africa [Namibia], Amboland [Ovamboland], Ondonga, [Ondongwa], Rautenen 344 (Z, holo.!). Plants (60-) 100-250 mm high. Corms deep-seated, 30 mm diam; tunics decaying into firm-leathery, coarse­ ly netted fibres extending into neck, 10-60 mm long, pale whitish brown. Cataphyll extending to ground level, pale membranous. Leaves 2, basal, suberect, narrowly lanceolate, (10-) 15-25 x (8-) 10-20 mm, attenuate, canaliculate with prominent midrib abaxially, leathery. Inflorescence divaricately branching, paniculate cyme, (l-)3-7-branched, up to 30-flowered; bracts subtending branches and pedicels only, lowermost linear-attenuate, conduplicate, up to 80 x 2 mm, becoming progressively shorter, uppermost lanceolate, attenuate to aristate, 3 x 1 mm; pedicels cemuous at tip, 15-25 mm long, elongat­ ing slightly in fruit and straightening, ultimately 20-40 mm long. Flowers nodding, campanulate, white flushed abaxially with pink or maroon at base of outer tepals, fra­ grant; perianth tube ± 4 x 5 mm with fringed corona, 0.5-1.0 mm high, at mouth of tube and forming collar extending over ovary to surround base of style; outer tepals spreading from base, oblong, obtuse, margins rev­ olute, 15-20 x 5-7 mm; inner tepals at first suberect but spreading in upper half, pandurate, shortly clawed, claw ± 2 mm long, blade ovate, 13-18 x 7-10 mm, margins crisped, apex slightly cucullate. Stamens attached to perianth near mouth of tube; filaments terete, ± 0.25 mm long, anthers narrowly lanceolate, 9-10 mm long; yel­ low, dehiscing by oblong apical pore, 1.5 mm long. Ovary half inferior; ovules ± 6 per locule; style extend­ ing ± 1 mm beyond anthers, white, 10-12 mm long. Capsules ovoid to globose, 10-12 x 8-12 mm. Seeds ellipsoid-pyriform, 4.0^4.5 x 3.0-3.5 mm, blackish brown; testa surface rugose. Figure 1. Distribution and ecology: locally common through the higher-lying parts of west-central and northwestern Namibia, occurring along the better w atered, western edge of the escarpment from west of Mariental in the south to Kaokoland in the north (Figure 2). The species typically occurs in colonies, often numbering many individuals, in sandy loam or heavy clay s

The African genera have been thoroughly mono graphed (Carter 1962;Scott 1991;Brummitt et al. 1998). Among them. Cyanastrum and Kabuyea are found only in the tropics, Walleria is widely dispersed through sub tropical Africa, and Cyanella is almost restricted to the winter rainfall region of South Africa in the extreme southwest of the continent. A single species of Cyanella, C. amboensis Schinz is found further north, in central Namibia. This species is morphologically anomalous among the rest of the species of Cyanella and was excluded from the genus by Scott (1991) pending a reap praisal of its relationships within the family. A label appended to the specimen Bean. VIok & Viviers 1824 (BOL) indicates that at the time she had considered that the species might be best transferred to the Chilean genus Conanthera, with which it shares a branched inflores cence and symmetrical anthers, but the similarity is not actually that close. Conanthera has numerous leaves and a zygomorphic, blue to purple perianth, in contrast to the bifoliate condition and actinomorphic. white perianth of C. amboensis. Currently, therefore, the species is of uncertain generic affinity. Molecular evidence places C. amboensis as sister to a monophyletic clade comprising the remaining species in the genus (six of the seven remaining species are included in this study). This topol ogy offers two options for resolving the generic position of C. amboensis'. the species can either be incorporated within an expanded circumscription o f the genus Cyanella; or a new genus should be erected to accom modate it. The morphological differences between Cyanella s. str. and C. amboensis are substantial, includ ing leaf number, inflorescence structure, floral symme try, and arrangement of the androecium. Genera in Tecophilaeaceae as currently circumscribed are homoge neous assemblages of closely related species in which floral symmetry in particular plays a defining role. Including C. amboensis within Cyanella would not only enlarge the circumscription of the genus to the extent that it is no longer morphologically coherent but would ren der it uniquely heterogeneous in comparison to the other genera in the family.
Cyanella s. str. is circumscribed on the basis of its tunicated corm, several to numerous leaves, zygomor phic flowers with dimorphic anthers, and bracteolate pedicels. All of the species are endemic to or centred in the winter rainfall region of southern Africa. Cyanella amboensis. in contrast, has just two leaves, the flowers are truly actinomorphic with monomorphic anthers, the pedicels are ebracteolate. and the species is distributed in summer rainfall Namibia. On the basis of these several morphological and ecological differences, the species is most appropriately accommodated in a separate genus closely allied to but distinct from Cyanella s. str., for which we propose the name Eremiolirion.
Distribution: a single species in central and north western Namibia.
Etymology: the name is a compound of the Greek eremios (desert or wilderness) and lirion (white lily).
Distribution and ecology: locally common through the higher-lying parts of west-central and northwestern Namibia, occurring along the better w atered, western edge of the escarpment from west of Mariental in the south to Kaokoland in the north (Figure 2). The species typically occurs in colonies, often numbering many individuals, in sandy loam or heavy clay soils, especially in stony or grav elly situations. The flowers close up at night around 21:00, re-opening in the morning around 09:00. They are fragrant during the day, with a jasmine-like fragrance at first but later smelling of stale urine, and are visited by bees and the occasional moth (Ward, Ward & Ward 10518). The dor mant corms sprout only in response to good summer rains, flowering mainly in February and March, sometimes as early as mid-January and rarely into early April, depending on the timing of the rains. In drier years when rainfall is below the average annual of ± 175 mm. the species flow ers poorly or not at all. The extent of flowering is also dependant on the amount of rain received. In addition, if the rains are patchy, migratory herds of springbok and zebra move into the areas very quickly and consume both the leaves and the inflorescences eagerly. The corms com prise part of the traditional diet of the San around Etosha (Giess, Volk & Bleissner 6039).
History: the species was first collected by Rev. Martti Rautanen of the Finnish Missionary Society, who arrived in Cape Town in late December 1868 w ith his colleagues on their way to establish mission stations in Ovamboland in northern South West Africa, now Namibia (Gunn & Codd 1981). Rautanen's interest in natural history was stimula ted by the visit of the German botanist. Dr Hans Schinz, who spent seven months of his botanical tour of South West Africa [Namibia] from 1884-1887 in the vicinity of the Olukondo Mission Station where Rautanen was based. Schinz was greatly assisted by Rautanen and returned the favour by naming several plant species for him. Since then the species has been collected from numerous localities to the south and is well known enough among locals to have received the charming sobriquet, desert snowdrop.  5 + 1 )........................................................................

Cyanella alba L.f.
The genus Cyanella comprises seven species of deci duous geophytes concentrated in the winter rainfall region at the southwestern tip of southern Africa (Scott 1991). Among them, Cyanella alba is easily the most distinctive, distinguished from all others by its linear-filiform leaves and contracted inflorescence axis. In this species the indi- vidual flowers are borne on elongated pedicels up to 200 mm long, and because of the highly condensed inflores cence axis, appear to arise singly from among the leaves.
The leaves in the remaining species are lanceolate to nar rowly lanceolate and the inflorescence is distinctly race mose, with individual flowers borne on short pedicels. There is significant variation in flower colour within C. alba, with the tepals varying from white to yellow, or rarely pale pink. Scott (1991) referred to minor morphological differences between the colour forms and suggested that further investigation might provide a clearer picture of the 'distribution...and infraspecific variation' of the species. Examination of the species in the field and in the herbari um indicates that the geographical distribution of the species is strongly correlated with differences in flower colour and some other morphological traits.
Cyanella alba is restricted to mountain renosterveld, a sclerophyllous, evergreen shrubland dominated by the asteraceous shrub Elytropappus rhinocerotis (L.f.) Less., occupying fine-grained clay soils of the Bokkeveld Series that receive moderate amounts of winter rainfall, from 250 to 450 mm per year (Low & Rebelo 1996). The species has a discontinuous distribution along the moun tain ranges of the west coast of Western Cape, from Nieuwoudtville in the north to Karoo Poort, east of Ceres in the south, a distance of some 200 km. Within this arc, the species occurs in three disjunct groups of populations ( Figure 3). The northern populations, distributed along the Bokkeveld Escarpment, are characterized by their white or rarely pale pink flowers flushed with darker pink on the reverse of the tepals. The anthers are uni formly yellow and the upper five are either coherent or free from one another. The plants are often well grown, producing more than three and up to nine flowers on long pedicels, (80-) 100-200 mm in length. Each pedicel bears a well-developed bracteole arising at a point between one third and three-quarters along its length.
A second group of populations occurs just south of the Bokkeveld, in the northern Cederberg Mountains around Clanwilliam and the Biedouw Valley. They are separated from the Bokkeveld populations by the arid expanse of the Doom River Valley, which supports a succulent karoo vegetation in which C. alba does not grow. The Cederberg populations resemble the Bokkeveld popula tions in stature but are distinguished from them by their flower colour. In these plants the tepals (sometimes only the inner whorl but usually both whorls) are pale yellow rather than white, and the upper five anthers are invari ably coherent and marked on their outer (upper) face with a prominent dark maroon or blackish blotch near the base. Like the Bokkeveld plants, the pedicels are bracteolate.
A third group of populations occurs well to the south at Karoopoort, in small patches of renosterveld that fringe the arid Tanqua River Basin. These plants are typically smaller in stature with fewer, up to three and often just one flower per plant, borne on pedicels that are mostly less than 100 mm long, rarely up to 150 mm. The flowers are white or pale pink, flushed darker pink on the reverse, and the upper five anthers are coherent and uniformly yel low. Superficially the plants resemble small forms of the Bokkeveld genotype, and it could be argued that they rep resent nothing more than depauperate forms from mar ginal habitats at the edge of the distribution. However, closer examination reveals that the bracteoles on the pedicels have been completely suppressed and are thus lacking. In this respect the plants are unique in the genus.
The three disjunct groups of populations of Cyanella alba are thus morphologically distinct from one another. These morphological distinctions point to long-standing separation, sufficient to allow genetic drift, suggesting that the disjunctions are interpreted to be the result of isolation due to the intervening areas of aridity rather than the alternative hypothesis of recent fragmentation through farming activities also offered by Scott (1991). Since each of the groups of populations is geographical ly isolated from the other, and is morphologically distin guishable. it is appropriate to recognize them at the level of subspecies. The homogeneous nature of each of the groups of populations points to genetic continuity within each group, while the differences between them indicate significant levels of genetic discontinuity. This is highly significant in the context of the high levels of microgeo graphic speciation that characterize much of the Cape flora (Goldblatt & Manning 2000 Distribution: Northern Cape, Bokkeveld Escarpment from just north of Nieuwoudtville southwards to Menzieskraal ( Figure 3). The collection Stokoe SAM55591 from the Nardouwsberg north of Clanwilliam, while certainly this form, is doubtfully correctly localized. The Nardouwsberg is a sandstone range that provides no suitable habitats for the species and although it is well explored, the species has never been collected there again. There are other instances of mislabelling by Stokoe (Goldblatt & Manning 2000: 507). Plants 80-150 mm high. Leaves filiform, 0.5-1.5 mm wide. Pedicels ebracteolate. Flowers 1-3, white to pale pink with darker pink on reverse. Anthers uniformly yellow.
Distribution: Western Cape, east of Ceres at Karoo poort ( Figure 3). Curiously, although Scott (1991: 46) did not cite any specimens corresponding to subsp. minor, she did provide dots indicating its distribution. The solitary, disjunct dot to the west of the main range of the subspecies is an error.