The genus Cliffortia ( Rosaceae ) in KwaZulu-Natal

The only area of endemism for Cliffortia L. outside of the Cape Floristic Region (CFR) is centred in the northern KwaZuluNatal Drakensberg. Eleven species of Cliffortia have been recorded from KwaZulu-Natal and distribution maps are provided for each. Clarification of the circumscription of C. browniarui Burtt Davy is discussed and a new species, C dracomontana. is described. C. paucistaminea Weim. is subdivided into two varieties: var. australis and var. paucistaniinea Comments on some of the other species that have frequently been misidentified are also prov ided along w ith a key to all the species in the area.


INTRODUCTION
Cliffortia L. is the largest southern African genus within the Rosaceae.It was last comprehensively revised by Weimarck (1934Weimarck ( , 1948) ) and subsequent published work has focused upon describing new species and clari fying taxonomy (Weimarck 1953, 1959: Oliver & Fellingham 1991, 1994;Fellingham 1993aFellingham , b. 1994Fellingham , 1995Fellingham . 2003)).Recent work by Whitehouse (2003) using mor phological and molecular techniques has indicated that some of the species boundaries need revising to reflect more accurately the diversity found within Cliffortia.
Cliffortia has its centre of diversity in the Fynbos Biome of the Cape Floristic Region (CFR) and over 80% of its species are endemic to that region (Goldblatt & Manning 2000;Whitehouse 2003).Linder (2001) stated several criteria to define areas of endemism: each area should contain at least two endemic species; the areas must be mutually exclusive; the areas must be narrower than the study area and several areas need to be defined for any discussion about their biogeography to be inter esting; and finally, the ranges of the endemic species should be optimized to be maximally congruent.Using these criteria, only a single area of endemism exists out side of the CFR and is centred on the northern KwaZulu-Natal Drakensberg (Whitehouse 2003).
Eleven species of Cliffortia have been recorded with in KwaZulu-Natal.C. odorata L.f. is recorded from a single old collection near Port Shepstone (Alexandra District, Hlokozi, 22 Feb. 1916. Rudatis 2242).The remaining species are either endemic to the Drakensberg or widespread, though scattered, through the province.C. filicauloides Weim.and C. spathulata Weim.are endemic to the northern KwaZulu-Natal Drakensberg, whereas the newly described species C. dracomontana C.M.Whitehouse is also endemic to the main Drakens berg escarpment but is found as far south as Ben Macdhui in Eastern Cape.
Despite the limited number of species present, the tax onomy and identification has often been confused.The flowers are small and remarkably uniform with few diag nostic characters.Therefore, species determination is pri marily based upon vegetative characters, especially leaf form and achenes when present.However, whereas sever al species are easily determined from the leaves of their mature plants, their juvenile foliage (i.e. the first few true leaves of seedlings or of new shoots resprouting after a fire) is morphologically very different.Seedlings and resprouting plants of Cliffortia have frequently been incor rectly labelled as unrelated species and this has confound ed the identification of otherwise easily discernible spe cies.Mature plants also sometimes show ecotypic vari ation in the size and shape of their leaves, especially on the borders of forest or rivers compared with more exposed slopes.Furthermore, the possibility of hybridization and subsequent introgression as found in several Cape species (Weimarck 1934;Fellingham 1993a;Whitehouse 2003) cannot be dismissed, although no confirmed examples have yet been demonstrated within KwaZulu-Natal.A key is presented here that attempts to take account of these variations and to identify species, where possible, even when only juvenile foliage is present.
The conservation status of all the KwaZulu-Natal species of Cliffortia should be regarded as "lower risk, least concern' for 1UCN Red Data List assessments (Golding 2002).The widespread species are common and often weedy.In particular.C. linearifolia and C. nitidula subsp.pilosa are sometimes dominant and have then been used in defining certain vegetation types (e.g.Killick 1963;Edwards 1967).The endemic species are more localized but all grow within the protected area of the uKhahlamba-Drakensberg Park, generally on the higher slopes and in more inaccessible areas.As a result, although currently only recorded from a few scattered localities, they are probably present and common in the interv ening parts of their ranges too.One species that has been particularly misunderstood in the past is C. browniana Burtt Davy.Its type collec tion was from the Mpumalanga Drakensberg escarpment near Lydenburg.from where several similar collections have subsequently been made.Morphologically, C. browni ana has no unique diagnostic character and is very diffi cult to separate from some small elliptic-leaved forms of C. serpyllifolia Cham.& Schltdl., with which it has often been confused.Within KwaZulu-Natal the two species need not be confused, as their distribution (see Figure 2) and altitude ranges do not overlap.However, the C. ser pyllifolia grows on the Silotwane Hills of Swaziland, closer geographically to the type locality populations of * C. rarnosissima has not been recorded from KwaZulu-Natal, but is included in the key because it is the only species that has also been recorded from the Free State.Mpumalanga.Limpopo and Lesotho.The key is therefore made more widely applicable by its inclusion.
C. browniana than those in the KwaZulu-Natal Drakensberg.Consequently, the exact nature of the rela tionship between the Mpumalanga populations and the specimens also attributed to C. browniana in the KwaZulu-Natal Drakensberg needs to be clarified by molecular techniques.
The identity of C. browniana has been further con fused by two distinct species being included under the same name within the KwaZulu-Natal Drakensberg.Originally Weimarck recognized the two entities as dis tinct, describing the specimen Hutchinson 102 separate ly in the footnotes to C. browniana.However, when he later described C. spathulata he also included a picture of the latter entity (Esterhuysen 101 S3) under the name C. browniana.Consequently the two species have been included under the same name (e.g.Hilliard & Burtt 1987).They are in fact easily distinguishable both in the herbarium and the field and the latter species is here described as C. dracomontana.
Etymology: spathulata means spoon-or spatula-shaped, referring to the leaflets that have a broad apex, then taper down into a narrow stalk.
The type locality of C. spathulata is recorded as Mahwaqa Mtn above Bulwer by Medley Wood.How ever, this mountain is relatively far from the currently known distribution of the species on the Drakensberg escarpment to the north.Furthermore, C. spathulata has not been recollected from that mountain and was not reported in a recent survey (Meter et al. 2002).It is there fore probable that the type locality has been erroneously recorded.

C. linearifolia Eckl. & Zeyh.
gate shrubs with unifoliate foliage and the other shorter and densely branched with trifoliate leaves.The unifoli ate form is found in wetter areas, such as along water courses, whereas the trifoliate form is found on rocky outcrops and drier slopes.However, both forms have similar distribution patterns (Figure 6) and it is not pos sible to determine from morphology alone whether they constitute distinct varieties or just ecotypes.Molecular work and/or transplantation experiments are needed to examine these two entities to establish the degree of phe notypic plasticity within the species.One of the most marked occurrences of dimorphism between juvenile and mature leaves is found in C. niti dula (Engl.)R.E.Fr.& T.C.E.Fr.subsp.pilosa Weim.(Figure 7).Whereas the mature foliage has needle-shaped leaflets typical of many species of Cliffortia, the leaves of seedlings and resprouting growth after fire are trifoli ate with a petiole and broadly toothed leaflets.It has therefore been frequently confused with the northern Hilliard & Burtt (1987) note that this species has two  Drakensberg endemic C. filicauloides (Figure 8), a species whose closest relatives are in fact found in the CFR (Whitehouse 2003).In the field, confusion is un likely as the habit of C. nitidula subsp.pilosa is always erect, whereas C. filicauloides forms arching stems that sprawl across boulders and down slopes, and the juvenile nature of the shoots is usually evident.Furthermore, shoots with juvenile leaves appear to be always sterile, thus flowering material will belong to C. filicauloides.
Sterile herbarium material can best be distinguished by examining the lateral leaflets, which are toothed and simi lar in size and shape to the middle leaflet in C. nitidula subsp.pilosa, but generally untoothed and narrower than the middle leaflet in C. filicauloides.
C. paucistaminea Weim. is a widespread species from the Outeniqua Mountains to the KwaZulu-Natal Drakens berg.It is easily recognizable over most of its range by its four-sepalled flowers and needle-shaped leaves that occur on closely overlapping brachyblasts.However, two distinct forms can be recognized.Between George and Port Elizabeth, the leaves are strongly curved up wards and towards the stem, accentuating the overlap ping nature of the brachyblasts.In addition, the leaflets are quite fine, 0.3-0.5 mm wide, and the young stems as a whole have a feathery appearance, resembling the Western Cape species C. exilifolia Weim.(which has only three sepals).To the north and east of Port Elizabeth the leaves are straighter or sometimes recurved away from the stem, when the brachyblasts then appear starlike, and the leaflets are generally broader.
Unlike C. linearifolia Eckl.& Zeyh.above, the two forms also have a geographical separation and therefore attributing them a taxonomic rank is appropriate.However, the variation could be the result of clinal dif ferentiation associated with climatic factors: the southern populations are subject to year-round rainfall, whereas those further north have an increasingly dominant sum mer rainfall pattern.Morphometric studies along with population-level molecular work are needed across the range of the species to determine if there is continuous gene flow between the two varieties or if they would be better regarded as distinct species.In this case, particular attention needs to be focused on the border between the two varieties in the mountains around Port Elizabeth.C.juniperina auct.non L.f.: Jacot-Guillarmod: 186 (1971).
Flowering time: predominantly November to April. Figure 11.
Habitat: humus-rich where it grows are subject to frequent rain and cloud; altitude 50-1 550 m.
Distribution: southern Cape mountains and lowlands be tween George and Uitenhage.Figure 10.
Etymology: australis means southern, referring to the distribution of the variety in the southern Cape.Weimarck described C. erectisepala from a single col lection from Paarl Mountain.He subsequently comment ed on collections from Franschhoek (Weimarck 1940) and Swellendam (Weimarck 1946).C. repens Schltr. he applied to unifoliate, needle-leaved collections from Uiten hage to Pietersburg.However, morphologically the collec tions from Uitenhage to Grahamstown are closer to C. erectisepala, having shorter, narrower leaves, upright erect habit, and non-spreading sepals.More recent collec tions have extended the range of C. erectisepala from Gifberg in the north to Rooiberg in the Little Karoo.The species in general is very inconspicuous and the current study has revealed that it is in fact much more common and widespread with many new localities being recorded.Consequently, the disjunction between Western Cape populations and the narrow-leaved Eastern Cape popula tions of C. repens is not as marked as originally thought.Therefore, these populations are now included in C. erec tisepala, whereas the range of true C. repens is reduced and extends only as far south as Baziya Mtn near Umtata.This change is significant biogeographically, as C. repens is no longer part of the Cape Flora and is better regarded as a broad Drakensberg endemic, whereas C. erectisepa la can be counted as a Cape Ubiquist (sensu Weimarck 1941) (see Figure 12).
A single collection is incongruous with this re-organi zation.The sheets are labelled as collected by Thomcroft (in Herb.Rogers 19187) from Barberton and are typical of C. erectisepala, although only male flowers are pre sent.However, no other collections have been made in this well-collected area that resemble C. erectisepala and it is therefore presumed that the locality is erroneous.

FIGURE 7 -
FIGURE 7 -Known distribution of Cliffortia nitidula subsp.pilosa.forms in the southern Drakensberg, one forming tall vir-I .and Cliffortia odorata, A. in southern Africa.
6a. var.paucistamineaLeaflets 0.4-0.8mm wide, held straight or curved downwards and away from the stem.Flowering time: predominantly September to February.Figure9.Habitat: found in humus-rich soil over sandstone rocks in full sun; altitude 0-2 350 m.Distribution: widespread from the Suurberg Mountains along the Drakensberg escarpment as far as Giant's Castle, with outlying populations along the Wild Coast and KwaZulu-Natal lowlands as far north as Nkandla.
Table Mountain Sandstone derived soils, in full sun and on well-drained soils, but areas \ m i n i FIGURE 11.-Holotype of Cliffortia paucistaminea var.australis.Witte Els Bosch, flats, Fourcade 2114 (BOL).