A revision of Lachenalia ( Hyacinthaceae ) in the Eastern Cape , South Africa

A taxonomic account of the genus Lacheiuiliii Jacq.f. ex Murray in the Eastern Cape, South Africa, is given. Eight species are recognised, and descriptions of these are amended and elaborated as necessary, three taxa have been reduced to synonymy and five species erroneously recorded within the province are excluded Three of the species are endemic to the province. An identification key is provided.


INTRODUCTION
The genus Lachenalia is confined to southern Africa and consists of a little over 100 species (Arnold & De Wet 1993).The majority of these species occur in the winter rainfall region of the Western Cape in the sclerophyllous shrublands or 'fynbos' ol the Cape floristic region (sensu White 1983).The geographic range of a few species extends from this region into the western parts of the Eastern Cape and one even occurs as far northeast as the Free State.The few remaining species are endemic to the Eastern Cape.The genus has attracted attention as an horticultural subject, and some of the more decorative species have become well known through cultivation.Despite this, no revision of the genus has been attempted since Baker's (1897) treatment in Flora capensis; however the work of Barker (see Arnold & De Wet 1993 for references) and Duncan (1988Duncan ( , 1996Duncan ( , 1997)), has resulted in the species from Western Cape becoming reasonably well known taxonomically.The same cannot be said for the Eastern Cape species.They have mostly been unknown in cultivation, and they are generally represented only by few, rather old herbarium specimens.
In the course of field work in the Eastern Cape it became clear that Baker's treatment was far from ade quate.Observations of populations of certain species indicated that they are very variable, and that many of Baker's diagnostic characters are unreliable.It was clear that a knowledge of the plants in the Field would be essential in producing an effective treatment for the genus in the province, and that a wider range of charac ters should be considered.
Key to the species of Lachenalia in the Eastern Cape la Perianth more than 15 mm long, always yellow-green; flower erecto-patent (45° or less to inflorescence axis) at and after anthesis; confined in Eastern Cape to coastal grassland 1 ■ L. algoensis lb Perianth less than 15 mm long, usually white to deep pink or purple, sometimes dull green with brown markings or rarely yellow-green (see L. bowkeri); flower cernuous to patent at anthesis, sometimes erecto-patent after anthesis; not confined to coastal grassland: 2a Inner perianth segments always recurved at tips, usually exceeding outer by at least 1.4In addition to more conventional macro-morphologi cal characters, Barker (1978) found seed morphology to be particularly useful in delimiting the different taxa in the Western Cape.Omduff & Watters (1978) and Johnson & Brandham (1997) have provided chromosome counts for several species.The value of these characters in understanding the Eastern Cape species has not previ ously been assessed.

MATERIALS AND METHODS
The present account is based on a study of herbarium specimens from relevant herbaria (see acknowledge ments), and of living plants in the field and in cultivation.A total of 16 populations representing five of the Eastern Cape species were located and detailed descriptive notes of living plants were made.Particular attention was paid to differential characters used by previous researchers and to characters that appeared to be variable within each population.Representative specimens were collected for cultivation and as herbarium specimens.Seeds of all species were examined with a JEOL-JSM 840 scanning electron microscope.

Distribution and habitat
In the Eastern Cape it ranges from the Knysna District in the west to the Bushman's River in the east, always near the coast (Figure 3).According to Duncan (1988), it extends to Worcester in the Western Cape.L subspicata Fourc : 79 (1934);Fourc.: 101 (1941).Type: Humansdorp Division, Flats Zuur Anys, on road to Kouga, Fourcade 3044 (BOL, holo.!).
Plant (50-) 110(-210) mm tall.Bulb globose, 8-17 mm diam., inner flesh translucent, white, outer tunics dry, coarsely papery, dark brown, often flaking, inner tunic minutely white-spotted, sometimes forming a short, coarse, brush-like neck of 5-10 mm, just visible at ground level; leaf sheath membranous, translucent, extending above neck by 5-15 mm, loosely clasping.Leaves l-2(-4), (45-)60-120(-300) x (5-)8-10(-30) mm, but frequently eaten by herbivores and thus appearing un naturally short, deeply channelled, erect, slightly fleshy, bracts cupped, deltoid, with slightly winged margins, white to greenish white, 1.1-1.8x 1.1-1.8mm, with basal spur up to 0.7 mm long, apex acute, 0.7-0.9mm long, facing downwards or curling under.Flowers some times (but not always) sweetly scented; pedicels held at 40° to stem, 0.5-2.0x 0.3-0.4mm, white to pale green; perianth patent or more usually somewhat cernuous, oblong campanulate, (6.5-)9.0(-11.0)mm long, (3.0-) 3.6(-4.0)mm diam.at base, slightly swollen at base, varying from white to pale green or very pale purple to purple with green or purple distal keels, becoming dark er with age; inner lobes obovate, 6-10 x 3.0-4.5 mm, spreading outwards, all 3 inner lobes of equal length, exceeding outer by (1.5-)2.0(-3.0)mm; apex obtuse, recurved; base cuneate, sometimes speckled with blue; outer lobes ovate, 4.6-6.0x 2.1-3.Duncan (1988) described L. bowkeri as a poorly known species.We interpret it as an extremely variable species, which is actually well represented in herbaria compared to most of the other Eastern Cape species.The variation in L. bowkeri is so great that during the course of the present study we were convinced for a long time that it represented at least three distinct species.Variation occurs in several characters, notably: overall size, num ber of leaves, flower orientation, pedicel length, perianth colour and degree of exsertion of the stamens.Variation in these characters can be seen among individual plants in a population, and is certainly attributable in part to phenotypic responses to minor variations in microhabitat.Considerable phenotypic plasticity has also been noted in plants brought into cultivation.However, different popu lations of L. bowkeri do show certain relatively constant and distinct features and these appear to some extent to be related to geographical distribution.It may eventually be possible to describe infraspecific taxa within L. bowk eri and thus delimit these variations formally, but at pre sent we understand too little of the nature of the species' variability to enable us to do so confidently.Further cytological studies of the different populations will be done at Kew and may help to clarify this variation.
Interpretation of herbarium material is complicated by the tendency in old flowers for the stamens to be drawn into the perianth and for the inner perianth lobes to become constricted at the mouth (Figure 2C).In this con dition the flowers closely resemble those of certain other species and present a very different facies to that of care fully pressed specimens with freshly opened flowers.At anthesis, stamen exsertion varies from 1 to 3 mm, but the type specimen is a plant with old flowers and thus in Baker's (1897) original description, L. bowkeri is stated to have included stamens.
Baker also states that the leaf is solitary.However, occasional 2-leaved plants were found in the populations we studied.Very rarely, exceptionally vigorous bulbs were found that had three or even four leaves and two inflorescences.Plants of L. bowkeri have regularly been misidentified in herbaria (often as L. orchioides (L.) Aiton), since they are frequently at variance with Baker's description with respect to the exsertion of the stamens and the number of leaves.
Plant height, leaf dimensions and the number of flow ers per inflorescence vary considerably within popula tions and from one population to another depending on exposure.Plants found in open karroid grassland are smaller than those shaded by grass clumps in open coastal grassland.Plants found in salt pans are more robust but not necessarily taller.In several localities robust plants with particularly thick peduncles have been collected.The bulb tunic appears to be influenced by the type of substrate and plants growing in hard, dry, rocky ground will always have thick, dry flaking scales with a short brush-like neck, whereas in loam or sandy soil the tunics will be less scaly and most often without a neck at all.The base of the leaf often tightly clasps the peduncle for up to a third of its length and then opens abruptly.This is seen predominantly in karroid areas, whereas under favourable conditions leaves are broader and sel dom clasping, sometimes in pairs, one always much smaller than the other.From observations in the field and Bothalia 28,2 (1998) from herbarium specimen labels it is also clear that L. bowkeri is variable in coloration.The perianth of plants from the dry Subtropical Thicket of the Fish River valley vary from white, tinged with green, with pale mauve dis tal keels, to entirely purple.Those from Naude's Hoek in the Keiskamma River valley are pale green to dull yellow-green with dull purple markings, while further west at Ghio Marsh Nature Reserve, the perianth is white with green speckling on distal keels and blue speckling on the base of the outer lobes, and older flowers already closed become tinged with dull purple.Within populations the flower colour seems to be more or less uniform, although, as in other species in the Eastern Cape, the individual flowers become darker in colour with age, tending to fade to dull purple as the fruit develops.Flower size, orientation and the extent to which the peri anth lobes are recurved or spread, particularly at the tips, is somewhat variable (Figure 2B-D), but tends to be rel atively constant within populations.Pedicel length shows little variation, from sessile to only 2 mm long.
Seed morphology is variable from one population to another.Seeds from herbarium sheets and live plants have been examined.Figure IB is representative of plants from the Fish and Keiskamma River valleys.Seed from collections further west are of similar size, but tend to have a rather variable arillode, which is often more blunt and less strongly decurrent.
Lachenalia subspicata was described in 1934 by Fourcade, who compared his new species only with L. orchioides and L. youngii Baker.It appears that he over looked the similarity between his plant and L. bowkeri, possibly because Baker (1897) incorrectly described the latter as having 'quite sessile flowers'.On examination of the type, a few' very short pedicels can be seen on one inflorescence.Following 'Pretoria National Herbarium practice ', Arnold & Dc Wet (1993) have placed L. sub spicata into synonymy with L. bowkeri.On examination of the type collections and consideration of the variabili ty of the species we concur with this practice.

Distribution and habitat
L achenalia bowkeri is virtually confined to the Eastern Cape, with only two known collections from the Western Cape.The type locality is 'Somerset' (presum ably Somerset East), but this is the only collection known from there, the majority of collections arc from the Albany District, with others from the Alexandria and Port Elizabeth Districts (Figure 4).All collections arc from below 850 tn and by far the majority of these arc found at altitudes below 600 m.It usually occurs in open grass land and often on margins of salt pans usually partially shaded under grass clumps, or, less commonly, in drier, rocky areas of Subtropical Thicket on south-facing slopes.Populations arc small and localised and plants occur singly or in small groups of two or three in areas not exceeding 10 nr.
Lachenalia campanulata can be distinguished by the combination of its spreading to horizontal flowers about 5.6 mm long with exserted stamens and subequal peri anth lobes, and its (usually) paired leaves.In the Eastern Cape it most closely resembles L. convallarioides Baker which has flowers of similar size and also has subequal perianth lobes.The two species differ in several charac ters.L. campanulata has a relatively dense inflorescence and the thickest point of its peduncle is at the base; in L. convallarioides the inflorescence is rather lax and the thickest point of its peduncle is immediately below the first flower.In L. campanulata the perianth is campanu late and the lobes are reflexed, whereas in L. convallari oides it is cup-shaped and the lobes are not reflexed.In L. campanulata seeds are ± 1.7 mm long and have a short, tapering, assymmetrical, decurrent arillode, and a reticu late testa; those of L. convallarioides are ± 2 mm long, and have a rounded, symmetrical, terminal, arillode, and a smooth testa.
Plants of L. campanulata are variable in size depend ing on the degree of exposure and individuals 230 mm tall can be found in sheltered positions in close proximi ty to others only 60 mm tall in exposed positions.Colour is also very variable from pure white to pink to deep red at anthesis, darkening to purple.
Baker's description of L. rhodantha is based only on the type collection.Bolus 719.It was said to be distin guished from L. campanulata by having a single leaf rather than two, lower flowers cernuous rather than hori zontal, stamens exserted, not included, and being bright red rather than white tinged with red.Not having seen these plants in their natural habitat Baker could not have known of the variability of these supposedly diagnostic characters, and plants closely matching the type speci mens of both species can be found growing together.In all three populations we have studied in the field, most of the plants were 2-leaved, whereas occasional 1-leaved and even 3-leaved plants were found in each.All of the specimens we have examined (living and preserved), including the type of L. rhodantha, have spreading to horizontal flowers in the lower part of the inflorescence.Baker's description of the orientation of the lower flow ers of L. rhodantha is an error, as they are clearly hori zontal.The type collection of L. campanulata, MacOwan 1836, consists of plants past anthesis which have their stamens included and style excluded, dark perianths and ovaries obviously swollen.In the field we have recorded that young plants consistently have spreading flowers with a slightly swollen base, an open mouth and exserted stamens.After anthesis the flower becomes horizontal and darker in colour, the perianth lobes close at the mouth and the stamens become drawn into the perianth and the base of the flower swells with the ovary.

Distribution and habitat
Lachenalia campanulata is confined to the mountains of the interior of the Eastern Cape between 24° and 29° E and 30° to 33° S, and has been recorded at altitudes from 1 200 to 2 400 m on steep, open slopes (Figure 5).
Lachenalia convallarioides is recognised by its rather lax inflorescence of nodding flowers with included sta mens and subequal perianth lobes, and its (usually) sin gle leaf.Other characters that serve to distinguish it from L. campanulata, the Eastern Cape species it most closely resembles, are given under that species.L. convallarioides is very variable in leaf and inflor escence size depending on the plant's degree of exposure to the sun.The most exposed plants are the most con spicuous in natural populations but as a result of their exposure are also the smallest, and these match the type collection.Plants growing in shaded places, partly hid den by other vegetation are rather inconspicuous, and can be much larger.Baker, not having seen this variation within populations, assigned varietal status to William son's collection (T.Williamson s.n.) which clearly match es shade-growing plants in populations we have seen.The chromosome number of 2n = 30 is unusual in the genus Lachenalia, where the common basic number is x = 7 or 8, although x = 5,9, 10, 11, 12, 13 and 15 have also been found.Johnson & Brandham (pers. comm.), have suggested that the 2n = 30 (x = 15) could be an allote traploid derived from taxa with x = 7 and x = 8 following hybridisation and chromosome doubling.

Distribution and habitat
Lachenalia convallarioides is only known from the Albany District near Grahamstown in the Eastern Cape, where it grows on rocky outcrops of Witteberg Quartzite, and from the Mount Arthur Range about 180 km to the northeast, at altitudes of around 17 to 1 800 m (Figure 5).Populations are very localised resulting in their being easily overlooked even in well-explored localities such as on Mountain Drive, Grahamstown.This locality is within walking distance from the Botany Department at Rhodes University but, remarkably, only three collec tions have ever been made from this generally well-col lected area.
Bowker's type collection from Kreli's Country, Caffraria, and Barber's Zuurberg collection, both num bered 444, are almost certainly the same collection from what was known in 1856 as the Zuurberg in the extreme northwest of Chief Kreli's territory in Caffraria (Hall 1856).These mountains are now known as the Mount Arthur Range.Bowker, a soldier, was stationed in this area and collected plant specimens for his sister, Mrs M.E.Barber (née Bowker), (Thorpe 1977;Gunn & Codd 1981), who sent them to Harvey at TCD. Baker saw this material much later (Dr John Parnell, Trinity College, Dublin, pers. comm.).The Bowker/Barber collection is the only known collection from outside the Albany District.
Lachenalia latimerae was described by Ms W.F. Barker in 1979 from the type and from cultivated mater ial collected near Patensie (Bayliss 7102).Recently we have located two new populations of this species.Barker's description needs no further elaboration.In common with most of the other species in the Eastern Cape it should be noted that overall size varies consider ably.Seed morphology of our collections conforms to that of the type collection.Figures IE; 6

Distribution and habitat
L. latimerae is found in the Kouga Mountains near Patensie in the Eastern Cape and it has also been collect ed at the Cango Caves near Oudsthoom in the Western Cape.It has been found at altitudes from 320 to 1 000 m (Figure 4).The two populations of L latimerae that have been examined in the field were both very localised, with less than 50 plants occurring in an area of ± 8 m2 in each case.Both were situated at the base of a very steep southfacing rocky slope with dense pockets of bush.The plants occur in open vegetation, often in moss-covered patches of about 1 m2 in full shade.
Lachenalia youngii was described by Baker (1897) and by Duncan (1988).No further elaboration is neces sary.Baker (1897) only cited a single sheet of Young 5545 under L youngii.He identified a second sheet labelled Young 5545 as L. unicolor.We have examined both sheets, and conclude that they are without any doubt duplicates and they conform perfectly to Baker's description of L. youngii.Baker's misidentification of the sheet accounts for erroneous distribution records for L. unicolor in the literature.It is uncertain whether seed morphology is diagnostic or not as only a single specimen has been examined (Figure IF).

Distribution and habitat
Lachenalia youngii has been recorded from numerous localities in the mountains of the southeastern part of the Western Cape, and extends into the Eastern Cape near Humansdorp (Figure 5), growing in fynbos.

Distribution and habitat
The species is essentially a Western Cape species with the majority of collections from the Karoo National Botanical Garden at Worcester.It has been recorded from Worcester District south to Port Beaufort and eastwards to Albertinia (Duncan 1996).One isolated record from Kommadagga in the Eastern Cape is a wide disjunction (Figure 3).The occurrence of the plant in the Eastern Cape needs further study, especially since the species appears to be quite closely related to L. bowkeri.

Distribution and habitat
Lachenalia karooica has been recorded from scattered localities in the Great Karoo in the Western Cape, the Northern Cape and the Free State, and extends into north ern parts of the Eastern Cape just south of the Orange River (Figure 3).At the type locality plants grow on south-facing aspects on dolomite outcrops.Lachenalia abides (L.f.) Engl.
Baker (1897) cited a specimen from Hill Park, Port Elizabeth (East South Central African Herbarium 49 K!) as L. aloides.The specimen was wrongly identified by Baker-it is L. algoensis.Duncan (1988) regards L. aloides as restricted to the Western Cape, a view with which we agree.
Lachenalia orchioides (L.) Aiton Baker (1897) cited two specimens from what is now the Eastern Cape, both from Grahamstown: MacOwan 1337 (GRA!) and Galpin 228 (PRE!).On examination of the specimens it is immediately clear that they are both L. bowkeri.A view shared by Barker (determinavit labels, no dates).We have not seen any collections of true L. orchioides from our area, and this species appears to be confined to the extreme southwestern part of the Western Cape.A large number of other specimens of L. bowkeri have been misidentified in herbaria as this species.

Lachenalia trichophylla Baker
Baker originally cited the type of L. trichophylla as MacOwan 2197 from Somerset East, i.e. within the Eastern Cape, and cited no other specimens.In Flora capensis (Baker 1897).however, he cited a single speci men for the species: Mader sub herb.MacOwan 2167 from Clanwilliam (in the Western Cape).The reason for this change is unknown but MacOwan in his list of dis tributed collections recorded MacOwan 2197 as Catha edulis, whereas MacOwan 2167 is recorded as Gladiolus spathaceus.No specimens of a Lachenalia bearing the collection number 2197 have been traced in relevant herbaria; however, specimens of Mader sub herb.MacOwan 2167 are present in K and SAM.The descrip tion of L. trichophylla does not match any known Eastern Cape material, but does agree with the Mader specimens, and other material from the Western Cape referred to as L. trichophylla.It is concluded that the original citation contains two errors, firstly the collection number should have been 2167 not 2197, and the specimen was not col lected in Somerset East.