Species groups in the genus Ehrharta (Poaceae) in southern Africa

Ehrharta Thunb. is a genus o f Gondwanaland distribution with its centre o f diversity in the winter rainfall Fynbos Biome of southern Africa. In recent subfamily treatments Ehrharta has proved difficult to place satisfactorily, and during the past five years it has been moved between Bambusoideae and Arundinoideae. However, most previous systematic studies using cryptic characters have covered only four taxa out o f about 35. The present study includes all African taxa, and demarcates seven species groups on the basis of both spikelet morphology and leaf blade anatomy. Parallelism and/or convergence in vegetative macromorphology within and between the species groups is widespread, and is similar, in some cases, to adaptations found in other plant families in the Fynbos Biome. However, these macromorphological trends are not reflected in the leaf anatomy. Leaf anatomy is generally consistent with the spikelet morphology. Some anatomical differences between the species groups in Ehrharta appear to be as great as differences between taxa of much higher ranks elsewhere in the Poaceae. This wide range of variability may be related to an early divergence of Ehrharteae from other grasses, as suggested by the Gondwanaland distribution, and may explain the difficulty of placing this fascinating yet baffling genus in a subfamily. UITTREKSEL Ehrharta is ’n genus van Gondwanaland-verspreiding met sy middelpunt van verskeidenheid in die winterreenval-Fynbosbioom van suidelike Afrika. In onlangse subfamilie-behandelings is dit moeilik gevind om Ehrharta bevredigend te plaas, en gedurende die afgelope vyf jaar is dit oor en weer in Bambusoideae en Arundinoideae geplaas. Die meeste vorige sistematiese ondersoeke waarin verskuilde kenmerke gebruik is, het slegs vier taksons uit ongeveer 35 gedek. Die onderhawige ondersoek sluit al die Afrika-taksons in, en baken sewe spesiegroepe af op grond van die morfologie van die blompakkie en die anatomie van die blaarskyf. Parallelisme en/of konvergensie in vegetatiewe makromorfologie binne en tussen spesiegroepe is wydverspreid en is in sommige gevalle soortgelyk aan aanpassings wat by ander plantfamilies in die Fynbosbioom aangetref word. Hierdie makromorfologiese neigings word nie in die blaaranatomie weerspieel nie. Blaaranatomie is oor die algemeen in ooreenstemming met die morfologie van die blompakkie. Sommige anatomiese verskille tussen die spesiegroepe in Ehrharta blyk net so groot te wees soos verskille tussen taksons van ’n veel hoer rang elders in die Poaceae. Hierdie wye verskeidenheid mag verband hou met ’n vroee divergensie van Ehrharteae vanaf ander grasse, soos aangedui deur die Gondwanaland-verspreiding, en mag ’n verklaring bied waarom hierdie bekorende maar raaiselagtige Conclusions................................................................ 64 Acknowledgements................................................... 64 References................................................................. 64


INTRODUCTION
In the most recent treatment (Willemse 1982), the tribe Ehrharteae Nevski (1937) consists of only a single genus, Ehrharta Thunb.(1779) which includes the gen era Microlaena R. Br., Petriella Zotov and Tetrarhena R. Br.Its distribution area in southern Africa, south western, southern and eastern Australia, Tasmania, New Zealand and Malesia falls within the cool-temperate Gondwanaland region (Figure 1).The centre of diversity for the genus is the winter rainfall Cape Floristic Region (Goldblatt 1978) or Fynbos Biome (Rutherford & West fall 1986) of southern Africa, with about 35 taxa, the other 10 taxa being distributed unevenly elsewhere in the range of the genus.In southern Africa, Ehrharta can be divided into seven species groups demarcated on the basis of both spikelet morphology and leaf blade anat omy.
An understanding of the variation within Ehrharta in southern Africa is necessary, not only to improve the classification of the local taxa, but also to show the relationships between the bulk of the genus in southern regions.Ehrharta may therefore serve as an additional plant group to indicate the floristic history of the Gondwanaland continents, and is the first representative of the Poaceae with this potential to be studied with this aim.
Additional information on Ehrharta is required because its position within the Poaceae is not yet settled.Since the time of Bentham & Hooker (1883), the group has been shuffled between the Phalarideae and the Oryzeae of the Pooideae (Festucoideae), the Ehrharteae and Arundineae of the Arundinoideae, the Ehrharteae of the Oryzoideae, and the Ehrharteae and the Oryzanae of the Bambusoideae.Modem studies based on cryptic charac ters such as leaf anatomy, chromosomes and embryo types have enabled most grass tribes to be clearly placed in one of about five subfamilies, but there is still no general agreement regarding the affinities of the Ehrhar teae.Prat (1960) and Clifford & Watson (1977) place the tribe in an unclassified residue that does not fit neatly into any of the modem subfamilial concepts of the Poa ceae.Most recently, Ehrharteae has been removed from the Arundinoideae (Renvoize 1981) to the Bambusoi deae (Renvoize 1985;Watson et al. 1985).
The instability of the higher classification of the Ehrharteae may be partly a result of the high degree of variation in its anatomical characters.Differences between species and species groups in the southern Afri can species appear to be as great as differences between taxa of much higher rank elsewhere in the Poaceae.Overseas workers who have examined only a few taxa may have erroneously concluded that they are represen tative of the entire genus or tribe.This wide range of variability may be a result of the early divergence of the Ehrharteae from other grasses, as suggested by the ancient Gondwanaland distribution pattern.
The representatives of Ehrharta in southern Africa may help to provide a better understanding both of the history of the flowering plants and of the early evolution of the Poaceae.They will therefore be treated in some detail.We propose to set out the species groups infor mally in this introductory paper, and thereafter to publish a parallel series of papers covering the taxonomy and anatomy of each species group.A concluding paper will present a formal generic and infrageneric classification and will develop hypotheses about the phylogeny of the species groups, and their phytogeographical signifi cance.

MATERIALS AND METHODS
For the morphological investigations, herbarium specimens were examined from BOL, JF, K, NBG, PRE, SAM and STE.The PRE material includes the anatomical voucher specimens of Ellis, described below.Field observations were made for about 30 of the taxa.Spikelets were dissected without special preparation, and were observed with a dissecting microscope.
For the anatomical investigations, specimens of Ehrharta plants were collected in the field throughout South Africa during the period 1974 to 1984.A total of 160 specimens were collected representing most of the taxa recognized in this genus.This sample was designed to incorporate considerable morphological and geo graphical variation in many taxa.Herbarium voucher specimens were prepared for identification by the Natio nal Herbarium (PRE) after segments of leaf blade mate rial had been removed and fixed in FAA (Johansen 1940).
Transverse sections, 10 /xm thick, were prepared after desilicification in 30% hydrofluoric acid (Breakwell 1914), dehydration using the method of Feder & O 'Brien (1968) and infiltration and embedding in Tissue Prep (Fisher Scientific).The sections were stained in safranin and fast green (Johansen 1940).The manual scraping method of Metcalfe (1960) was used to prepare scrapes of the abaxial epidermis.These were either double stained in methylene blue and ruthenium red or only in safranin.The anatomical detail was recorded photo graphically using a Reicherdt Univar microscope and Ilford Pan F film (ASA 50).
The material fixed in FAA was also used for ultrastructural studies with a scanning electron microscope.The leaf blade segments were dehydrated in 2,2dimethoxypropane (Merck) for 2 hours and then placed in 100% acetone for 5 minutes following the method described by Neumann, Rushing & Mueller (1982).The material was critical point dried with liquid C 0 2 at 85 atmospheres at 40° C after which it was mounted on aluminium stubs with double sided tape.It was then glow discharge coated with a thin (± 400 A0) layer of metallic gold in a sputter coater.Preparations were either stored in a dessicator with silica gel or observed directly with an ISI SX-25 scanning electron microscope ope rated at 25 Kv accelerating voltage.Cuticular structure was photographed with a 6 x 7 cm camera on Ilford FP4 120 film at varying magnification.

The classification o/Ehrharta within the Poaceae
The placement of Ehrharta in the grass family has varied at both the tribal and subfamilial level (Table 1).Most older systems, based only on spikelet morphology, place Ehrharta in Phalarideae because of the two sterile florets below the single fertile floret.Modem treatments based on a number of cryptic characters place the genus in a tribe of its own, Ehrharteae.Although first distin guished as an entity by Link (1827), his authorship of the name cannot be accepted for nomenclatural reasons (Willemse 1982).Thus, Nevski (1937) is the recognized author for Ehrharteae, which was also described in error as new by Tateoka (1957).Nees (1841), in the earliest treatment of southern African grasses that adopted a suprageneric classification, placed Ehrharta in the Oryzeae, although Trinius (1839), working in co-operation with Nees, had earlier placed it in the Phalarideae, where it remained until the modem system of Stebbins & Crampton (1961), who placed the tribe back into the Oryzoideae.
Although there is now agreement on tribal composi tion, the question of subfamily classification is not yet settled.The most recent treatments include Ehrharteae in Bambusoideae (Renvoize 1985;Watson etal. 1985), but both discuss the difficulty of including Ehrharteae in the subfamily as they constitute it.Renvoize (1985) states that there is no clear correlation between anatomical and other characters, such as embryo type and lodicules, and Watson et al. (1985) indicate that, although they decided to treat the group as bambusoid, other numerical analyses either place it as arundinoid or lose it alto gether.Soderstrom & Ellis (in prep.), who classify bam busoid genera and allies, place the Ehrharteae outside the Bambusoideae, closer to Arundinoideae.This confused situation may be explained partly because, as Table 1 indicates, the modem subfamily classifications are based on a very small set of data derived from only a few species.In this century, only Tateoka (1963) has examined more than three taxa, and only four taxa in all have been examined either for chro mosomes, embryos or leaf blade anatomy.The detailed study of the southern African taxa of Ehrharta presented in this and following papers should provide the breadth of information needed to indicate the natural affinities of this fascinating but baffling genus.

The southern African species o f Ehrharta
The study of African Ehrharta falls into three phases, which correspond roughly to the 18th, 19th and 20th centuries.The earliest phase was characterized by the description of species in four different genera.L. C. M. Richard (1779) described Trochera striata, and a few months later Thunberg (1779) described Ehrharta capensis.These names were based on different types but represent the same taxon.Soon after, Linnaeus (1791) described two more species, but placed them in the genus Aira.Lamarck (1786), J. E. Smith (1789Smith ( , 1790) ) and J. F. Gmelin (1791) each described species in Thunberg's genus Ehrharta.However, Thunberg himself (1794) accounted for his specimens collected from the Cape in Melica, so that his Prodromusplantarum capensium treatment comprised two genuine Melica species and four species of Ehrharta.
The second phase, one of consolidation, began in the next century with Swartz's (1802) thorough and beauti fully illustrated study of the genus, in which he recog nized nine species, each with synonyms.He described no new taxa, but moved Thunberg's 'Melicas' to Ehrharta.He was aware of, but did not take up, Tro chera as the name for the genus.He also did not always follow priority of publication in applying the specific epithets, although his synonymy within each species is otherwise sound.Palisot de Beauvois (1812) alone took up the name Trochera, to which he attributed two species.
Schrader (1821) covered 17 species, including all those in Swartz's (1802) treatment as well as eight new species that were based on the specimens of Hesse.He was the first to divide the genus into sections, distin guishing two entities on the character of bulbous versus fibrous 'roots'.Although published two years later, Thunberg's (1823) Flora capensis treatment was merely a summary of Swartz (1802), and did not include Schrader's species.During this period of consolidation, a few new species were described by Kunth (1829), by Schultes (1830) and by Steudel (1853).
Most ambitious of the nineteenth-century studies of Ehrharta were those of Nees ab Esenbeck (1832Esenbeck ( , 1839Esenbeck ( , 1841)).His Florae Africae Australioris (1841) recog nized 25 species, of which nine were described by him, using collections of Drege and Ecklon.Nees also put forward a multitude of varieties (nine in E. calycina alone), which, however were not validly published be cause he did not consistently apply Articles 24-27 of the ICBN (Linder 1985).He adopted Schrader's basic divi sion of the genus by 'root' type, and in addition sub divided the fibrous-rooted species according to size and hairiness of the sterile lemmas.These subdivisions are  Mez (1921).Since 1955, new taxa have been described by Launert (1961) and Gibbs Russell (1984a, 1984b).
Stapf did not follow Kuntze (1891) in taking up the name Trocher a, and soon after the conservation of generic names was permitted, Ehrharta of Thunberg was conserved over Trochera of L. C. M. Richard, which was rejected in ICBN (Voss et al. 1983).

TAXONOMIC CHARACTERS
The taxa described by previous workers can be grouped into seven easily recognizable groups on the basis of morphological and anatomical characters as shown in Table 2.A brief discussion of those characters found to be useful in this regard follows.

Macromorphological characters
For a genus of moderate size, Ehrharta has a wide range of macromorphological characters, a selection of which are shown in Table 2.They are reasonably well known, from the detailed descriptions of Stapf (1900) and the briefer but more image-creating descriptions of Chippindall (1955).However, unless the species are placed in groups based on spikelet morphology and leaf blade anatomy, the diversity of vegetative characters is confusing.With these groupings, it can be seen that still recognized, although a number of the species have been re-aligned.
The last of the comprehensive nineteenth-century accounts of Ehrharta was that of Steudel (1853).He compiled the species known at that time, 32 in all for southern Africa.Of greatest significance is his basic division of these species into two groups, not by the possession of a bulbous base, as previously, but by the separation of E. setacea and E. rupestris into a group ('Racemosae') distinct from the rest of the species ('Paniculatae').It is unfortunate that he did not elaborate on the characters on which he based his decision, because this fundamental division in Ehrharta is supported by the present study.After 1855, no further work was done in the genus except that Kuntze (1891), rediscovering the priority of the genus name Trochera, transferred all the then-accepted taxa to Trochera.
The third phase in the taxonomic study of Ehrharta has been dominated by Stapf s (1900) Flora capensis treatment, which is the most recent critical study of the southern African taxa.He also recognized 25 species, of which six were newly described by him, and he placed a number of previously recognized species into synonymy.Chippindall (1955)

C apensis 1. M IDR IB A N D KEEL: Keel absent (A ) A
a. Habit in the majority of Ehrharta species is peren nial, but several are annual.The annual habit seems related to unpredictable moisture availability.All the annual species, except one, occur in the drier north western part of the range of the genus, centred on the Succulent Karoo, from about Clanwilliam north to southern SWA/Namibia.These species fall only in the Erecta and Calycina groups, and each has closely related perennial species in the more mesic central part of the distribution of the genus.The other annual species, E. longiflora, grows in seasonally wet places, and it has, for an Ehrharta, a wide distribution, occurring not only in arid areas to the northwest, but as far east as Port Alfred.It is the only species in which the spikelet morphology is not consistent with the leaf blade anatomy.
b. Rhizomes occur in all perennial species, but the form of the rhizome varies widely, and is correlated with the substrate.Species that grow between rocks have tough, branched, naked rhizomes that are hardly diffe rent from the lower part of the culm.These rhizomes can creep for several meters in deep cracks, so that the plant forms a dense linear cushion.Species that occur in sand have long branching rhizomes that run 0,3 m or more below the soil surface and send up culms at the tips, so that a single plant can appear to be a loose colony several meters across.These rhizomes may be naked or densely clothed by hairy cataphylls.Species that grow in clay or loam soils usually have short knotted rhizomes that do not spread far away from the base of the individual plant, although in a few species unbranched rhizomes a few centimeters long and with papery cataphylls may occur.c.Culm characteristics are the most important in determining the overall appearance of a grass plant, and in Ehrharta the culms are very diverse.In different taxa the culms vary from solitary to numerous, from herba ceous to wiry or woody, from erect to procumbent, from unbranched to simply or verticillately branched.Length varies between taxa from less than 100 mm to over 1,5 m in flowering individuals.
Taxa in the Capensis groups have the lowest (some times the two lowest) intemodes modified into an extremely hard cylindrical or globose bulb-like structure.The occurrence of this 'bulb' may be associated with the periodic fires in the fynbos.'Bulbous' species are much in evidence for a year or two after a fire, but apparently do not occur in long-unbumed situations.However, immediately following the next bum, flowering plants may be found with many old dead 'bulbs' hidden under ground, indicating that the plant is several years old, having survived the fireless period in this form.
Stolons are known only in a local limestone variant of the widespread species E. calycina.d.Leaves, as in other grasses, are composed of sheath, ligule and blade.The sheaths are split to the base and usually cylindrical, or rarely keeled.In some species the upper sheaths may be longer than the intemodes, so the leaves are closely imbricate.In suffrutescent species with reduced leaf blades the sheaths may be persistent around the culm, or be held outward at an angle, or they may be lost.The basal sheaths may be persistent, densely clothing the base of the culm, or they may slip away from the culm with age, leaving it exposed.A few taxa may be distinguished by a characteristic colour of the basal sheaths.
The ligules are short, usually less than 3 mm long, and take the form of a ciliate rim in all species groups except the Erecta and Calycina groups, in which most taxa have glabrous ligules.
The leaf blades vary from broad and flat, often with one or both margins with a heavy undulate marginal vein, to folded, rolled, setaceous or absent.Length varies between taxa, from less than 100 mm to over 1 m.Loss or reduction of leaf blades is correlated with suffru tescence.In these taxa, the phenology of leaf blade development should be studied in the field, because it is possible that striking differences in appearance may be due to the age of a plant rather than to genetic dif ferences.
e. Inflorescences are most commonly paniculate, but in some taxa may be reduced to a raceme with few spikelets, and there is often a tendency for the narrower inflo rescences to be secund.Generally the inflorescence is exserted far above the highest leaf sheath, but in a few taxa (especially annuals) it is closely subtended or even enveloped below by an inflated leaf sheath.In some taxa with racemose inflorescences the main axis curves sinu ously around the appressed spikelets.

Spikelet morphology
The spikelets are usually laterally compressed, although in some taxa the sides are rounded.At matu rity, the spikelet is shed as a unit above the glumes, which are persistent on the pedicel.A spikelet consists of a pair of glumes, two empty sterile lemmas, and, at the tip of the rachis, a fertile floret composed of lemma, palea and bisexual flower.Each species group has a characteristic range of spikelet sizes.a. Glumes are subequal, and may be shorter than, equal to or longer than the whole spikelet, and relative glume length is a useful character for distinguishing between species in a group.The glumes are more papery in texture than the lemmas, and are unomamented.
b. Sterile lemmas are the single most striking spikelet feature of Ehrharta.Their curious sculpturing in some species is unknown elsewhere in the grass family, and renders a detached spikelet immediately recognizable.Shape, relative size and ornamentation of the sterile lem mas are the most useful characters in separating the species groups.In three of the species groups, Erecta, Calycina and Dura, the tips of the sterile lemmas can have long awns, and in the Capensis and Villosa groups the bases are shortly stipitate.In most Ehrhartas the ster ile lemmas are of similar size and the smaller fertile lemma differs from them, but in the Setacea group the first sterile lemma is short and glume-like and the fertile lemma is similar to the second sterile lemma.
The lemma surfaces may be dull or shining, scaberulous or smooth.The sides and margins may be glabrous or have hairs of various lengths, and the bases may be glabrous or bearded.The bases of the second sterile lemma and the fertile lemma come together in a hinge like joint that resembles an earlobe, and which may have a membranous appendage.All these characteristics are used in separating the species groups.Differences in transverse and longitudinal ribs and veins distinguish species within the groups.c.Fertile florets have a lemma that is smaller and more laterally compressed than the sterile lemmas, and is unomamented.The palea is sickle-shaped or straight, is much smaller then the lemma and is usually hidden within it.The fertile florets of most species are similar, and are not useful for distinguishing species or groups.d.Flower structure is remarkable in Ehrharta because of the presence of six stamens in most species, although some of the taxa with small spikelets may have five, four, three or one stamens.The two lodicules are relati vely large and flat, and are usually ovate or 2-lobed.e. Caryopsis information for comparison is lacking because of the small number of spikelets that are found to have mature fruits.It appears that the spikelet is shed as soon as the fruit is mature.The lack of this informa tion, as well as data about embryos, is a serious defect in our knowledge of the genus.

Anatomical characters
A suite of leaf blade anatomical characters separate and define the various species groups in Ehrharta.These cryptic features are manifested in the leaf blade as seen in transverse section, in the abaxial epidermis in surface view and ultrastructurally (Table 2).Some of these characters distinguishing the species groups recognized in Ehrharta are generally considered to be features of high taxonomic value, important at the subfamily level in the Poaceae.In Ehrharta, however, they separate taxa below the generic level.Examples are differences in mesophyll structure and microhair shape.
The 10 characters which were scored for all the southern African Ehrharta taxa are diagramatically represented in Table 4.This suite of anatomical charac ters in combination serves to define and diagnose each of the seven species groups as well as two subgroups within one of the groups.The variation encountered in most of the characters in most of the groups will be fully described and discussed in subsequent papers.The character states used to define each of the groups here are, therefore, somewhat generalized and may vary wi thin certain limits and intergrade between certain groups.In the cases where variation was observed, definite lines of development could be traced and the variation could usually be interpreted on this basis.

Midrib or keel
The vascular structure of the keel or midrib is an important character in grass taxonomy at the subfamily level.Complex vasculature of the keel would definitely support bambusoid affinities as suggested by many authors.This type of keel does not occur in any of the species groups from South Africa but the presence of only a median vascular bundle or the presence of a defi nite keel with additional parenchyma tissue differs between the species groups recognized, and only varies within the Calycina group.

Ribs and furrows
Adaxial ribs and furrows, particularly the massive ribs of the Longifolia group, distinguish certain groups.The raised, inflated abaxial epidermal cells in the mid-inter costal zones of the Calycina group are diagnostic and of considerable phylogenetic interest because this charac teristic is shared with the Phalarideae with which the Ehrharteae has been linked by earlier authors.These epidermal cells are also evident in the epidermis as seen in surface view.

Mesophyll
The chlorenchyma cells of Ehrharta are surprisingly variable in structure and arrangement.In the Setacea group arm cells definitely are present in some taxa.This is a bambusoid characteristic but is also known in some taxa without bambusoid or oryzoid affinities (Watson et al. 1985).In some groups the chlorenchyma cells are compact and angular with minute air spaces but, in other groups this tissue is very diffuse, of irregular, rounded cells with air spaces clearly visible.In the Villosa group the abaxial layer of chlorenchyma is very regular and almost palisade-like.

Epidermal zonation
Some groups are distinguished by costal and intercos tal zones on the abaxial surface.Zonation is not evident in the Setacea and Longifolia groups and this is asso ciated with an absence of abaxial stomata as observed in these two groups.This lack of zonation appears to be associated with the ecological conditions of the Moun tain Fynbos with very low soil nutrition and also occurs in many of the danthonoid grasses from the same habi tats.Species from the Lowland Fynbos, on the other hand, have clear epidermal zonation and abaxial sto mata.These characters tend to vary in groups which have a wide ecological tolerance and occur in both these two major habitat types.

Intercostal long cells
Hexagonal or inflated but rather short long cells with sinuous walls distinguish some groups, whereas others have elongate, fusiform, hexagonal intercostal long cells with straight walls.In the Calycina group with this latter type of long cell, the mid-intercostal cells are much longer than the lateral ones and these cells are also the raised, inflated cells as seen in transverse section.In other groups the intercostal long cells are rectangular with sinuous walls.

Abaxial stomata
The stomata of the various species groups of Ehrharta are also of taxonomic interest.They may be absent on the abaxial surface, as in the Longifolia group and in many taxa of the Setacea group.In the Capensis group the guard cells are clearly visible but the subsidiary cells are covered with wax platelets, as seen with the scanning electron microscope.In the Erecta group the stomata are often obscured by large, solid wax plugs, whereas, in the Calycina group cubical wax granules often overlie the stomata.In the Villosa group the stomata are sunken and overarched by four characteristic papilla-like flanges from the adjoining long cells.In the Ramosa group the stomata have no associated wax deposits whereas in the Dura group the pores are obscured by dense accumula tions of wax platelets.Stomatal structure, and associated epicuticular wax, therefore, serves to distinguish all but two of the species groups in Ehrharta and provides use ful characters at the species level.The low dome-shaped subsidiary cells, which are present throughout Ehrharta, are typical of, but not exclusive to, the Arundinoideae (Renvoize 1981).

Costal silica bodies
The silica bodies of Ehrharta are somewhat variable and rather difficult to use as diagnostic characters.Single, rounded, or paired bodies characterize some groups whereas in others this type of body is more dumb bell-shaped.In the Erecta and Calycina groups the costal silica bodies are in short chains and of the dumbbell type but this shape can only be determined by varying the focus.Silica body shape and arrangement is often a character differentiating taxa at higher taxonomic levels than that of genus.

Prickles
The costal macrohair-like type of prickle hair is only found in the Erecta and Calycina groups.This type ol hair is common in the pooid and bambusoid grasses but is much rarer in the arundinoids.In the other species groups of Ehrharta, only shortly barbed prickles occur, except in the Longifolia subgroup of the Capensis group where prickle hairs are absent.

Microhairs
Microhairs in Ehrharta are very small and difficult to observe with the light microscope.However, they are clearly visible with the SEM and clear differences between the species groups are visible.The hairs may have short, truncated distal cells or this cell may taper to an acute apex.The length of the hairs also differs between the species groups.The Longifolia subgroup is the only grouping that does not possess abaxial micro hairs.Microhair shape is a very important character separating the subfamilies of the Poaceae and the degree of variation observed in Ehrharta is most unusual.

Epicuticular wax
The occurrence and nature of the epicuticular wax deposits also appears to be a useful taxonomic character in Ehrharta, serving to distinguish between the species groups.The wax may be absent or present either as fine rods or as heavier platelets.The significance of this wax as a feature of higher taxonomic application is unknown.
The suite of leaf anatomical characters used to distinguish the species groups in Ehrharta, therefore, includes a wide spectrum of attributes and, in combina tion, these features serve to assign any given specimen to a particular group.However, it is not easy to identify to the species level using leaf anatomical criteria because the species are not so distinct anatomically.Neverthe less, the recognition and definition of the seven groups of species in this very difficult genus will definitely help in providing new insights into the phylogenetic relation ships of the genus, until such time as sufficient embryo and chromosome data are available.

OUTLINE OF SPECIES GROUPS
Each of the species groups in Ehrharta can be described on the basis of a set of morphological and anatomical characters as summarized in Tables 3 & 4 respectively.A brief synopsis of characters and diagno sis is then given for the constituent taxa of each group.2. Adaxial ribs and furrows present; ribs rounded and well developed (except i n f .setacea subsp.disticha and subsp.uniflora).

Mesophyll compact, of small rounded or isodiametric cells;
tendency for arm cell-like invaginations in all taxa; E. setacea subsp.scabra with typical bambusoid-like arm cells.
5. Leaf blades expanded or rolled (but very short in E. barbinodis).

CONCLUSIONS
A better understanding of the taxonomy and systematics of Ehrharta is needed because its morphological and anatomical variation and its geographic distribution indi cate that the genus may provide important clues towards understanding the natural relationships of the Poaceae.The southern African species fall into seven groups based on both spikelet morphology and vegetative ana tomy, while the vegetative macromorphology exhibits parallelism and/or convergence between the groups.The series of papers to follow will relate the details of mor phological and anatomical structure of each species group to the interpretation of relationships within Ehrharta, as it is presently circumscribed and within the rest of the grass family.

FIGURE 1 .
FIGURE 1.-Worldwide distribution of Ehrharta, showing the number o f species and infraspecific taxa reported for various areas.The number o f southern African species naturalized in India and Australia are indicated with an asterisk.
keel developed; only median vascular bundle present.
sterile lemma not reduced, veined.10. Second sterile lemma not stipitate.11.Sterile lemma surface rough.12. Sterile lemma sides glabrous (few short hairs in E. rehmannii var.and sterile lemmas well developed, with tips rounded, sides rough and glabrous, bases glabrous and with basal appendages.Stomata with distinct rims and no wax deposits.and intercostal zones differentiated.5. Intercostal long cells rectangular with slightly sinuous walls; tanniniferous cells.6. Abaxial stomata present; pore obscured by dense accumulation of wax platelets

TABLE 3 .
-Morphological characters diagnostic for the species groups in Ehrharta

TABLE 4 .
-Anatomical characters diagnostic for the species groups in Ehrharta