A Phytosociological Study of the Cape Fynbos and other Vegetation at Jonkershoek , Stellenbosch by

The Braun-Blanquet phytosociological method was tested in the complex Fynbos vegetation o f the South-western Cape Region o f South Africa. In the Swartboschkloof Nature Reserve, Jonkershoek, the Fynbos, riverine scrub and forest vegetation was classified preliminarily into eight com ­ munities, which are described floristically and related to habitat. The results hold promise, and the possibilities o f classifying the Cape Fynbos in a formal phytosociological system are discussed. I n t r o d u c t io n The Braun-Blanquet phytosociological method commonly used in Europe has remained relatively unknown in Southern Africa. Possible reasons for this are language difficulties, the need for more general, rather than more detailed information on the vegetation, and the general non-acceptance of the method by English and American plant ecologists. For many years the only source of information in the English-speaking world was Fuller & Conard’s (1932) authorised translation of Braun-Blanquet’s first edition of Pflanzensoziologie (1928), a work that omitted certain important details of the phytosociological technique. More detailed German works of the phytosociological school were largely inaccessible (for example Ellenberg, 1956; Braun-Blanquet, 1951. 1964). In recent years an English evaluation of the method was given by Poore (1955, 1956), although his main criticisms were shown by Moore (1962) to be largely unfounded. Becking (1957) reviewed the phytosociological school and its concepts, and Kiichler (1967) translated the tabulation techniques from Ellenberg (1956).1 Originally, nearly all vegetation surveys undertaken in Southern Africa were on a physiognomic or on a non-formal descriptive basis, with the exception of Acocks (1953) who used his own floristic technique to construct a system of veld types. Recently, multivariate analysis and ordination techniques have been applied by a number of ecologists in the Republic of Southern Africa (see Killick, 1966—67). In the Portuguese territories physiognomic classifications were created, and in each unit one sample was taken. The species list from each sample was rated on the BraunBlanquet scale of cover-abundance and sociability (e.g. Gomes Pedro & Grandvaux Barbosa, 1955). In Central Africa, Belgian ecologists, traditionally familiar with the Braun-Blanquet method, have set up a hierarchy of syntaxa for the Congolese rain forests (Lebrun & Gilbert, 1954). In Southern Africa it has often been suggested that the flora is too rich to apply successfully a floristic technique. In the Cape Fynbos, particularly, such techniques were thought to be impracticable. Outside Southern Africa also, Donselaar (1965), in a classification of the Northern Surinam savannas, stated that to use successfully the Braun-Blanquet method the number of species must be moderate. * Botanical Research Institute, Department o f Agricultural Technical Services, P.O. Box 994, Pretoria. t Jonkershoek Forest Research Station, Department o f Forestry, Stellenbosch. + Botanical Research Institute, Department o f Agricultural Technical Services, Stellenbosch Unit, P.O. Box 471, Stellenbosch. H'he percentage classes o f Braun-Blanquet as listed in Table 10, p. 231, Kiichler (1967) are not correctly reproduced, as they give the impression o f being based purely on cover. In reality they are based on cover and abundance.


A b s t r a c t
The Braun-Blanquet phytosociological m ethod was tested in the complex Fynbos vegetation o f the South-western Cape Region o f South Africa.In the Swartboschkloof Nature Reserve, Jonkershoek, the Fynbos, riverine scrub and forest vegetation was classified preliminarily into eight com munities, which are described floristically and related to habitat.The results hold promise, and the possibilities o f classifying the Cape Fynbos in a formal phytosociological system are discussed.

I n t r o d u c t i o n
The Braun-Blanquet phytosociological method commonly used in Europe has remained relatively unknown in Southern Africa.Possible reasons for this are language difficulties, the need for more general, rather than more detailed information on the vegetation, and the general non-acceptance of the method by English and American plant ecologists.For many years the only source of inform ation in the English-speaking world was Fuller & C onard's (1932) authorised translation of Braun-Blanquet's first edition of Pflanzensoziologie (1928), a work that omitted certain im portant details of the phytosociological technique.More detailed German works of the phytosociological school were largely inaccessible (for example Ellenberg, 1956;Braun-Blanquet, 1951. 1964).In recent years an English evaluation of the method was given by Poore (1955,1956), although his main criticisms were shown by Moore (1962) to be largely unfounded.Becking (1957) reviewed the phytosociological school and its concepts, and Kiichler (1967) translated the tabulation techniques from Ellenberg (1956).1 Originally, nearly all vegetation surveys undertaken in Southern Africa were on a physiognomic or on a non-formal descriptive basis, with the exception of Acocks (1953) who used his own floristic technique to construct a system o f veld types.Recently, multivariate analysis and ordination techniques have been applied by a num ber of ecologists in the Republic of Southern Africa (see Killick, 1966-67).In the Portuguese territories physiognomic classifications were created, and in each unit one sample was taken.The species list from each sample was rated on the Braun-Blanquet scale of cover-abundance and sociability (e.g.Gomes Pedro & Grandvaux Barbosa, 1955).In Central Africa, Belgian ecologists, traditionally familiar with the Braun-Blanquet method, have set up a hierarchy of syntaxa for the Congolese rain forests (Lebrun & Gilbert, 1954).
In Southern Africa it has often been suggested that the flora is too rich to apply successfully a floristic technique.In the Cape Fynbos, particularly, such techniques were thought to be impracticable.Outside Southern Africa also, Donselaar (1965), in a classification of the Northern Surinam savannas, stated that to use successfully the Braun-Blanquet method the num ber of species must be moderate.
From systematically distributed quadrats used for an association analysis of the vegetation of the Cape of Good Hope Nature Reserve, Taylor (1969) also prepared a Braun-Blanquet synthesis table.Due to the rigid sampling technique, a num ber of quadrats had to be excluded from the table as they were located on community transitions and were not representative of more or less homogeneous stands of vegetation.Nevertheless, he obtained associations that were recognizable in the field.This stimulated the authors to undertake a survey according to the phytosociological techniques de cribed by Braun-Blanquet (1964) and Ellenberg (1956), in order to test its usefulr iss in the floristically rich Fynbos vegetation.The area chosen was the Swartboschkk of-Sosyskloof N ature Reserve in the Jonkershoek valley near Stellen bosch, as the results could also be used for the International Biological Programme survey of conservation sites.

T h e S t u d y A r e a
The Swartboschkloof-Sosyskloof N ature Reserve, 373 hectares in extent, lies in the Jonkershoek Forest Reserve in the H ottentot Holland M ountains some 15 km from Stellenbosch, at 34°00'S latitude and 18°57'E longitude.It was proclaimed a nature reserve in terms of the Forest Act in 1936, to conserve vegetation representative o f the Jonkershoek valley.
The vegetation is chiefly Fynbos (Acocks 1953, veld type 69).It consists mainly o f sclerophyllous scrub and dw arf scrub, in which amongst other, Restionaceae and Cyperaceae also play an im portant role.A dense scrub grows along the river banks, and upstream, where the valleys narrow, small patches of forest occur.Van der Merwe (1966) analyzed vegetation-site units in the Reserve.

Topography, Geology and Soils
The Reserve is a fan-shaped valley at between 285 m and 1 200 m altitude.The steep slopes average about 30°, ranging from about 5° to 50°.Aspect is largely northerly with only about 5 per cent of southerly facing slopes.
Porphyritic granite of the Pre-Cape System forms the undulating floor o f the valley, but is often buried by sandstone talus fans from the medium-grained hom o geneous sandstones of the Table M ountain Series.The sandstones are very resistant to erosion so that the valley is bounded by precipitous cliffs and knife-edge ridges.Continuous creep characterizes the talus on the slopes below the cliffs.
The valley was formed by a series of secondary faults, roughly at right angles to those which gave rise to the main Jonkershoek valley.Streams follow the courses of these faults.The jointed layers with a northerly tilt store a fair proportion of the precipitation which is released in conspicuous seepage steps.
Soils are generally poorly developed, rocky and acid, the granite soils being less acid than the sandstone.Those soils derived from granite are often over 1 m deep and more fertile than those from sandstone, which are coarse-textured, skeletal and rarely over 1 m deep.Mixtures of granite and sandstone debris give rise to inter mediate soils.Humic and organic soils of varying depth (0,05-2 m) occur locally on wet sites.

Climate
Wicht et al. (1969), in describing the climate of the Jonkershoek valley in some detail, notes that " The climate is Etesian of the Mediterranean type,...............with a dry summer and the average tem perature o f the warmest month below 22°C ..............." It conforms to Walter & Leith's (1960) Climate Type IV, and to K oppen's (1931) humid-mesothermal £Wctf-climate.Precipitation is usually associated with strong cyclonic winds from the north-west.In summer, strong anti-cyclonic winds from the south-east prevail.These are generally dry, but frequently deposit moisture at highe altitudes.From a rain-gauge at the mouth of the Swartboschkloof valley the mean annual rainfall over 20 years is about 1 600 mm.Fifty per cent of this fell during the months May to August, and only 12 per cent from December to March, when the greatest moisture deficits occur.Snow falls rarely at higher altitudes.
Colonists first settled in the Valley in the late 17th century, but the effects of their agricultural practices on Swartboschkloof was probably first evident after 1832, when the Reserve formed part of land transferred to a farmer living close by.The early farmers used these lands as grazing for sheep, goats and cattle, exploiting the forests for timber, the Protea arborea stands for firewood, and such shrubs as Agathosma crenulata for medicinal uses.
Early man probably had little effect upon the vegetation, but European coloniza tion would have initiated rapid change.Veld-burning to improve grazing was fairly standard practice, and Swartboschkloof was probably burnt at 4-10 year intervalsmore frequently than probably occurred naturally.Grazing would have been restricted largely to the lower slopes owing to the steep topography.The area does not appear to have been cultivated.Mammal populations diminished or disappeared as a result of hunting and the destruction of habitats in the surrounding lowlands.
The acquisition of Jonkershoek by the Departm ent of Forestry in 1933 and proclam ation of the Nature Reserve introduced radical changes in land-use.A policy o f complete fire protection was adopted, and only recreation and non-destruc tive research were permitted in Swartboschkloof.The Reserve is surrounded by firebreaks, burnt in spring or autum n on a 4-year rotation.In spite of policy, the whole reserve burnt accidentally in December 1942 and February 1958.Two small fires also occurred during this period.

M e t h o d s
Gleason's (1925) regression equation, as reformulated by Goodall (1952), was applied to data from three sets of nested quadrats, 1 m2 to 256 m2 in size, to estimate the quadrat size-information ratio (Werger, 1970).On this basis a quadrat size of 100 m2 was selected.The Braun-Blanquet method does not require uniform sized quadrats, but as many as possible were 100 m2 for the sake of consistency.Only two quadrats (nos.14 and 37) were taken at 50 m2 to avoid obvious heterogeneity due to a sharp change in slope aspect and to a conspicuous vegetational difference.Q uadrat nos.42, 43 and 44 were 128 m2, being a stage in a set of nested quadrats.
Altogether 44 quadrats were layed out at sites selected on the basis of visually asses sed homogeneity of vegetation structure and habitat.Species in each quadrat were listed and rated on the Braun-Blanquet cover-abundance scale.Only permanently recognizable species were recorded, thus omitting annuals and most geophytes.Site features such as slope angle, aspect and altitude were measured, soil depth was estimated, and geological and geomorphological characteristics were noted.
The data were entered in a raw table and from this an association table was compiled (Tables 1 and 2), according to the prescribed methods.Tables 1 and 2 list character species for each community, together with those species with over 50 percent presence in a group of two, three or four related communities.The remaining species are grouped in a " tail" of the tables.O f this " tail" , species occuring only once or twice and with low cover-abundances in the quadrats, are listed for convenience in the Appendix.Complete species-quadrat lists are considered im portant for Braun-Blanquet phytosociological work, because some of the species will assume greater importance when further data are acquired, and are necessary if current concepts of the communities are to be revised.
The communities are named after one or two faithful and conspicuous species, so that they are easily recognizable in the field.
To obtain complete pictures of the communities and their relationships, more data are needed, preferably from other areas.Only then will it be possible to decide with certainty whether some of the below mentioned species are true character species.

T h e C o m m u n it ie s
Based on floristic composition we have distinguished five Fynbos, one riverine scrub and two forest communities.

Fynbos
The Fynbos communities usually consist of two or three layers: a graminoid and dwarf shrub layer, a shrub layer and, in some communities, a tall shrub or small tree layer (Table 1).

Protea arborea-Rhus angustifolia Community
This community consists of many of the more common Fynbos species.The vegetation usually consists of three layers: a tall shrub and small tree layer from 2-3 m high; a shrub layer 1-1,5 m high; and an undergrowth of sedges, grasses, restionaceous plants, other herbs and dw arf shrubs from about 0,10-0,60 m high.The middle layer usually has the highest cover, although sometimes the upper layer may dominate.
There are a number of possible character species, which are, however, not fully constant, like Podalyria m yrtillifolia, Euphorbia genistoides and Helichrysum zeyheri.Constant but not strictly faithful character species appear to be Diospyros glabra and Rhus angustifolia.A number o f species that clearly typify this community and the Restio perplexus Community are Protea arborea (which has its optimum in this community), Psoralea obliqua, Watsonia pyramidata, Themeda triandra and Ursinia filiformis.
The Protea arborea-Rhus angustifolia Community is found on relatively deep, detrital, sandy loam soils; slope varies, but aspect is generally between north and east.It is limited to lower altitudes (up to ca. 600 m) in Swartboschkloof.
The community occurs in a very distinct habitat, with generally steep slopes (16°-32°) and relatively deep detrital soils, with a definite south-eastern aspect.It occurs mainly at altitudes of 300-500 m.

Tham nochortus gracilis-Hypodiscus aristatus Community
The shrub layer (0,75-1,50 m high) is the most im portant stratum in this com m u nity.There is an undergrowth o f about 0,20-0,50 m high, but the upper layer of ca. 2 m high tall shrubs and small trees if often wanting.1).Restio sieberi is highly constant but com m on also to the Restio perplexus Community.
M any of the species com m on to the Protea arborea-Rhus angustifolia and the Brunia nodiflora-Psoralea rotundifolia Communities, are rare or absent here, but as Table 1 shows, there are many others, especially Leucadendron adscendens and Tetraria bromoides, that are common to all these communities.
The community occurs at fairly high altitudes (± 500-800 m), on Table M ountain Sandstone soils o f various depths.Slopes are m oderate to steep (13°-34°).There is no prevalent aspect.

Restio perplexus Community
A lthough this community is only represented by three samples, it is, nevertheless, quite distinct from the other communities.The community is usually two-layered (undergrowth layer of herbs, sedges, grasses and dw arf shrubs 0,20-0,40 m high; shrub layer 0,60-1,50 m high), although there may be an open small tree layer 3-4 m high.
R estio perplexus appears to be a good character species.Cliffortia polygonoides and Euryops abrotanifolius are abundant and/or constant, but not strictly faithful (see Table 1).
Again, some species, for example Clutia alaternoides, are common to all four comm unities so far described, while many of the species generally occurring in either the first two communities or the first three communities are absent in this Restio perplexus Community.As already pointed out, this latter community has a num ber o f species in common with the first Protea arborea-Rhus angustifolia Community.The relationship between these two communities is, however, not yet clear.
Releve 34 cannot be unequivocally placed in this Community, because it contains species typical of the Thamnochortus gracilis-Hypodiscus aristatus Community.This releve may be more heterogeneous than was thought, due to its position in a firebreak.

Berzelia lanuginosa-Osmitopsis asteriscoides Community
This Fynbos community on seepage areas along drainage lines is very different from all others.The vegetation is dense, with an undergrowth of ca.0,50 m high, a middle layer o f sedges and shrubs 1-2 m high, and an open tree layer 3-5 m high.The permanently moist soil often contains a high percentage of organic matter.

Riverine Scrub and Forest
The stream channels are generally fringed by woody vegetation.This is largely a dense scrub, but locally real forest occurs, especially where the valley narrows.Small patches of forest also occur on coarse, loose scree.These woody communities have a few faithful species in com m on (e.g.Secamone alpini and Restio subverticillatus) and there is a mutual overlap in species between the different comm unities (Table 2).

Brabeium stellatifolium Community
This dense scrub up to 5 m high fringes the lower, less steep parts of the streams.Character species include Brabeium stellatifolium, Podalyria calyptrata, Halleria elliptica, Blechnum capense, Pentameris thuarii, Freylinia oppositifolia and others.M ore information might necessitate subdivision into two communities, one with the character species M eterosideros angustiolia, and the other lacking it.

Rapanea melanophloeos Community
This community constitutes the largest part o f the forest vegetation.It occurs at higher altitudes along stream channels, with steep gradients.Further downstream the rivers are fringed by the Brabeium stellatifolium Community.
There are two tree layers.The highest reaches 10-12 m, the other 5-8 m, beneath which is a " layer" largely formed by saplings.There is no shrub layer and very little undergrowth.Character species of the community are Rapanea melanophloeos and, possibly, Kiggelaria africana, Pellaea viridis and Asplenium aethiopicum.

Heeria argentea Community
This forest community occurs in small patches on coarse, loose screes.There is usually a singly tree layer, with an average height of 7-8 m, sometimes reaching 10 m.The undergrowth consists of herbs and small shrubs up to 0,60 m high with a low cover.The trees are covered with lichens and mosses, indicating that moisture con densation often occurs.
Character species are Heeria argentea, which has a high cover, and possibly Rumohria adiantiformis.The species Aloe mitriformis and Oftia africana may also be considered character species of the undergrowth where the tree canopy is open.
The two Rapanea and Heeria forest communities in Swartboschkloof have some species in common of which Podocarpus elongatus is the most abundant.This species and other common species, such as Olea africana, Maytenus acuminatus, Hartogia schinoides and Olinia cymosa, all form canopy trees, while Halleria lucida and Myrsine africana occur regularly in the understory.

D is c u s s io n
The main purpose was to test the Braun-Blanquet phytosociological method on the floristically rich Fynbos vegetation.In that a practical classification into com munities based on floristic criteria has been possible, and because these communities can be correlated with definite environmental factors, we may conclude that the Braun-Blanquet method can be successfully applied to the analysis of Fynbos vegeta tion.This contradicts D onselaar's statement (1965) that the species num ber must be moderate if the method is to be successful.
W alter (1968) states that, on the South-west Cape mountains near the coast, aspect plays a more im portant role than altitude.The rather strict aspect preferences of the Protea arborea-Rhus angustifolia and the Brunia nodiflora-Psoralea rotun difolia Communities seem to confirm this, although the Thamnochortus gracilis-Hypodiscus aristatus Community does not show any aspect preference, and is corre lated in Swartboschkloof with altitude and geological substrate.However, it is difficult to distinguish whether altitude or geological substrate is the more im portant, since these two factors are correlated with each other.Adamson (1931) sampled the vegetation o f north and west slopes of Table M ountain between 300 and 360 m, on granite with overlying Table M ountain Sand stone talus, in order to find " some explanation of the apparent variability o f the floras" of the communities which he floristically distinguished.Although he noted that " the floristic composition of a community must ultimately be its most im portant characteristic" , he could not prove this when he tried to express it in terms o f species dominance, concluding that there was life-form dominance rather than species dominance.Nevertheless, our results show that when floristic composition is used as a criterion, it is indeed possible to characterize communities.
It is remarkable that conspicuous vegetational structures do not always coincide with the boundaries of communities defined upon floristic composition.We find the dense sclerophyllous Protea neriifolia and Protea repens dom inated scrub distributed in the three most common Fynbos communities, although mainly concentrated in the Protea arborea-Rhus angustifolia and the Brunia nodiflora-Psoralea rotundifolia communities.The relationships between physiognomic structure and phytosociological classification have been discussed by Westhoff (1968).He uses the term " Twin Form a tion" for " strata which vary independently on [of] each other" , or " stands which are floristically hardly, but structurally considerably different."These situations occur where extreme habitat factors prevail.Burning is considered to be such an extreme factor.Donselaar (1965) found that in the Surinam savannas the floristic composition of some treeless areas was so similar to the undergrowth o f a certain type o f tree savanna, that they should both be placed in the same association, arguing that the floristic principle should predom inate over the structural one.Dansereau & Arros (1959), who favour the principle o f structure, argue that decisions of the kind similar to the above are too arbitrary and they doubt the reality of such associations.
Since Fynbos is recurrently burned and the structure of the Fynbos is very often closely related to the elapse of time since the last burning, we feel, like Donselaar (1965), that a floristically based system of communities is more consistent and practical.
It is interesting to note that Tables 1 and 2 show the intrusion into the Brabeium stellatifolium and Berzelia lanuginosa-Osmitopsis asteriscoides Communities o f some species that usually occur in other Fynbos.Such species are Asparagus thunbergianus, Rhus angustifolia, Rhus rosmar inifolia, Diospyros glabra, A r is tea thysiflora, Cliffortia cuneata, Erica hispidula, Protea neriifolia, Anthospermum aethiopicum and Ehrharta ramosa.Both the Brabeium stellatifolium and the Berzelia lanuginosa-Osmitopsis asteriscoides Communities occur only as long narrow strips o f vegetation along stream channels and seepage lines.The Communities are thus characterized by a high ratio o f margin to surface area, or a large " edge effect."Because of this large " edge effect," intruding species from surrounding communities are more likely to be found throughout such long narrow communities compared with communities of another shape.Species that are most likely to intrude are those with a wide ecological amplitude, as appears to be the case for some o f the species listed above.O ther intruding species, like Pteridium aquilinum and Cassytha ciliolata, seem to be ones with their optimal occurrence in ecotones, especially ecotones o f a shrubby nature.In Leeuwen's (1966) terminology they can be called species of the " limes convergens" , that is species living under unstable conditions, usually as large numbers of individuals.Adamson (1927) andW alter (1968) point out that on the Table M ountain Plateau Schizaea pectinata can be very abundant and proves to be a good indicator o f fire.In Swartboschkloof this species was not abundant, however, and was not present in the quadrats sampled on the firebreaks.Adamson (1935) found that after a fire in the vegetation on the slopes of Table M ountain, Euryops abrotanifolius increased quickly during the first three years and was dom inant for the following two years.From then on the species decreased rapidly.O ur results from Swartboschkloof also link Euryops abrotanifolius with the fire factor.In quadrats 34 and 35, situated on the firebreak, it scored cover-abundance values of 2 (5-25 %), whereas it occurred in only two other quadrats, in old stands, with low values.
Floristic data on Fynbos vegetation in the literature are limited, and complete species lists of quadrat samples are unavailable.It is, therefore, difficult to link the Communities presently distinguished, with communities described in other Fynbos literature.As early as 1908, M arloth presented an outstanding description of the flora and vegetation o f the Cape, describing the vegetation on a structural and geogra phical basis.In his general descriptive account of the vegetation of Table M ountain, Adamson (1927) mentions a num ber o f species from the .. .plateaucommunities on sandy soils. . . ." , which are typical o f the Thamnochortus gracilis-Hypodiscus aristatus Community described here.The habitat and altitude of these two com m uni ties also show similarities.Further study may show closer correlations.In addition, the Communities he briefly described from seepage areas and along stream channels, show floristic similarities, through species such as Berzelia lanuginosa and Osmitopsis asteriscoides, with those of Swartboschkloof.The same is true for gorge forest and riverine scrub communities in both areas.
Taylor (1969) mentions, among others, Pentaschistis colorat a and Staberoha cernua as preferential species for one of two communities forming a mosaic in the " plateau fynbos association" in the Cape of Good Hope Nature Reserve.It is probable that this community is related to the Thamnochortus gracilis-Hypodiscus aristatus Community.Certainly, his " Osmitopsis seepage scrub association" is closely related to our Berzelia lanuginosa-Osmitopsis asteriscoides Community.His description of the Protea arborea pseudo savannah association" (waboomveld; Taylor, 1963), with typical species Protea arborea, Montinia caryophyllaceae, Rhus rosmarinifolia, Leucospermum conocarpodendron, Themeda triandra, Hermannia cuneifolia, Cymbopogon marginatus and Bobartia incica, strongly suggests it belongs to the Protea arborea-Rhus angustifolia Community.
Releves from the Fynbos are expected to accumulate and will enable ecologists to classify the vegetation in a hierarchical system of associations, alliances, orders and classes.A special difficulty in classifying the complete Fynbos vegetation lies in the fact that many species have a very limited distribution.This will make it necessary to distinguish geographical races of an association, or regional associations with a limited geographical extension.Communities earn the status of an independent regional association only if they have their own character species and alliance charac ter species, as pointed out by Oberdorfer (1968).If they differ only in the accom pany ing species, they are geographical races of one association.Problems of this nature will need to be handled very carefully in classifying Fynbos vegetation.
It is inevitable in such a survey that some quadrats will be badly sited.They may either fall on an ecotone (probably Q uadrat 34), or give an unrepresentative picture of the vegetation type to which they belong (Quadrat 27), or represent a different vegetation which is clearly undersampled (Quadrat 22).All such samples are usually left out of the association table, because they do not contribute to an under standing of the vegetation.Quadrats which are wrongly sited cannot be used, but if they represent an undersampled community, as in Quadrat 22, they should be kept until more information becomes available.We have here included these stands in Table 1 to give an example of sampling errors.
Many more releves, especially from elsewhere in the Fynbos are required for a successful phytosociological classification (see Tiixen, 1970).We have not presented a formal hierarchical system here because of a lack of data; this report should be regarded as a challenge to South African ecologists to prove and improve the classifi cation and so to build up an understanding of the vegetation.

A c k n o w l e d g e m e n t s
We thank the Secretary of Forestry for allowing publication of this paper and for permitting the investigations on a State Forest Reserve.We are indebted to Dr. D. Edwards and Prof. Dr. R. Tiixen for their critical comments and to Mr. C. van der Merwe for assistance in the field.