The vegetation of seasonally flooded areas of the Pongolo River Floodplain

The Braun-Blanquet method of phytosociological analysis has been used to identify the plant communities of the seasonally flooded areas of the Pongolo River Floodplain. Six communities and two sub-communities, whose distribution is closely related to relative periods of exposure and inundation, were recognized. The struc­ ture and interrelationships of the communities are considered and the role of human activities in determining their extent is commented upon. Comparison is made with similar situations elsewhere in Africa. RÉSUMÉ VÉGÉTATION DES ZONES PÉRIODIQUEMENTINONDÉES DANS LE LIT MAJEUR DE LA RIVIERE PONGOLO On a utilisé la méthode d ’analyse phytosociologique de Braun-Blanquet pour identifier les communautés végétales des zones périodiquement inondées dans le lit majeur de la riviere Pongolo. On y a reconnu six communautés, dont la distribution est en corrélation étroite avec les périodes relatives d'exposition et d ’inondation. On considêre la structure et les interrelations des communautés et ion commente le róle joué par les activités humaines dans la détermination de leur amplitude. La situation est comparée á d ’autres du même genre qui se présentent ailleurs en Afrique.


INTRODUCTION
The im pou n d m en t o f rivers inevitably results in changed conditions dow nstream and where this involves a floodplain, the effects are likely to be o f considerable im portance (A ttw ell, 1970;Begg, 1973;Phélines, C oke & N ichol, 1973;T ow nsend, 1975).T his m ay be particularly p ertin en t in the A frican context since som e o f the m ost im p o rta n t fisheries are located on floodplains (W elcom m e, 1974).
T he object o f this study was to define and to d eter m ine the factors influencing the distribution o f the p la n t com m unities o f the seasonally flooded areas o f th e Pongolo River F loo d p lain .It form s p a rt o f a m ulti-disciplinary research program m e w hose objec tive is an assessm ent o f th e long-term effects o f im pounding the w aters o f th e P ongolo R iver on the floodplain ecosystem.
T his paper presents the results o f a B raun-B lanquet analysis carried out between June and Septem ber 1974.

THE STUDY AREA
The Pongolo River F loo d p lain (27°S, 37°E) is situated on the low-lying coastal plain (75 m above m ean sea level) in the n orth-eastern co rn er o f N atal, S outh A frica (Fig. 1).It is approxim ately 60 km long and 0,8 km wide.A lthough the floodplain covers 10 416 ha (W elcom m e, 1974), this study only co n sidered 8 800 ha, because the floodplain in the vicinity o f N hlanjane Pan, w hich had n o t been flooded since construction o f the P o n g o lo p o o rt D am (1972) (Fig. 1), was excluded.
Flooding is predom inantly a sum m er phenom enon, and p rior to co nstruction o f the D am , th e m ajor floods were o f relatively sh o rt d u ra tio n : 28 cum ec flows lasted fo r periods o f up to 100 days and caused m inim al flooding; 56 cum ec flows fo r up to 60 days; and flows in excess o f 85 cum ecs fo r only between tw o and ten days (Table 1).T he m ajo r floods, arising from either the very high flows o r low er flows for longer periods, annually recharge a nu m b er o f shallow w aterbodies, locally referred to as pans (H utchinson, Pickford & Schuurm an, 1932).These have a total area o f c. 2 600 ha (M usil, G runow & B ornm an, 1973), but this is reduced considerably by evaporation to less than 1 000 ha during the dry season.
F luctuating w ater levels have been show n to influence b o th the species com position and the d istri bution o f individual species in w etland areas (H ow ard-W illiam s, 1972; H ow ard-W illiam s & W alker, 1974; H ow ard-W illiam s, 1975).On the Pongolo River F loodplain some 8 000 ha lies between high flood level (H F L ) and m axim um retention level (M R L ) o f the pans, i.e. the level at which pan and river lose co n tact (Breen et a l., 1978), and is therefore norm ally only flooded for relatively short periods.A fu rth er ap p ro xi m ately 1 000 ha becomes exposed gradually d uring the dry season as levels d ro p below M R L , and is therefore inundated for longer periods than th a t between H F L and M R L. The n atu ral flooding regime has been altered by the im pounding o f the river at Jozini (Fig. 1), and the effects o f these changed conditions on the floodplain vegetation will be d is cussed in a subsequent paper.
Inhab itan ts o f the higher area im m ediately aro u nd the floodplain m ake extensive use o f the alluvial soils for subsistence agriculture and m ost o f the floodplain vegetation is disturbed (Fig. 2).How ever, since a portion o f the floodplain was incorporated in the N dum u G am e Reserve (Fig. 1) when it was established in 1924, a relatively undisturbed area was available for com parison.

METHODS
The Braun-Blanquet m ethod o f sam pling and synthesis as described by W erger (1974) was adopted.Using aerial photographs, physiognom ically distinct vegetation areas were dem arcated and field checks were carried out to ensure th at they were correctly assigned.O ptim al plot size was determ ined by increas ing the area until the num ber o f species recorded rem ained m ore or less constant.Plot size was d eter m ined as 100 m 2 (5 m x 20 m) for the forest com m unity and, although it may be adapted to vegetation type (W erger, 1974), all com m unities were sam pled by 100 m 2 plots.A to tal o f 106 plots was investigated during A ugust and Septem ber 1974.All species present were listed and, since m ost annuals only appear after the floods have receded (usually in A pril), it is probable th at m ost were still recognizable at the time o f the survey.
A fter com m unities had been checked and dem ar cated on the aerial photographs ( 1 :25 000), they were subjectively transcribed onto a m ap o f the sam e scale.
The following site factors, soil pH in 1 N KC1 (Jackson, 1958), and proportions o f clay, silt and sand (Black, 1965), were determ ined.W ith the aid of 1: 6 000 contour m aps the height o f each plot above M R L and below H F L was estim ated.

THE PLANT COMMUNITIES
Six com m unities have been recognized and they may be grouped according to their relative periods of exposure and in undation: (1) the com m unities (two) o f high-lying areas, which are only inundated for short periods; (2) those (three) o f low-lying gently sloping areas, which only becom e exposed as w ater approaches and drops below M R L , i.e. those having the longest subm ergence and shortest period of exposure; and (3) the com m unity which occupies the interm ediate areas, where the slope is slightly steeper  O utside the N dum u G am e Reserve (Table 3: relevés 90,65,96,81,80,87,83 & 104), the com m unity usually has only tw o strata: a tall tree stratum (12-15 m), form ed principally by Rauvolfia caffra, Ficus sycomorus and Trichilia emetica with cover of 60-80% , and a 2-3 m shrub stratum o f Allophylus decipiens, Grewia caffra, M onanthotaxis caffra, Ficus capreifolia, Syzygium guineense and Adina microcephala with cover o f 10-20% .W ithin the reserve a further two strata o f shade to leran t species are evident (relevés 73, 72, 68, 69, 70, 71 & 74), the taller (0,75-1,25 m, cover 15-20% ) o f Dicliptera heterostegia and Setaria chevalieri, while in some relevés (81 & 87) the grass Oplismenus hirtellus form s a low stratum (0 ,1 -0 ,2 5 m) w ith cover n o t exceeding 10%.The tendency for these tw o strata to be developed within the Reserve reflects the agricultural disturbance on the levees outside the Reserve.The tw o sites where this sub-com m unity has developed, to the south-w est o f K hangazini and west o f M engu Pans (Fig. 2), are areas where the river levees and the m argin o f the floodplain are close together.Such a situation w ould facilitate an intergrading o f the Ficus sycomorus-Rauvolfia caffra and the Acacia xanthophloea-Dyschoriste depressa C om m unities.This sub-com m unity is therefore regarded as transitional.

The Acacia xanthophloea-D yschoriste depressa
Community (Fig. 5) Acacia xanthophloea and Dyschoriste depressa are confined to this com m unity and the Ficus sycomorus-Acacia xanthophloea Sub-com m unity and they may, therefore, be regarded as either character or diffe rential species (Table 2: relevés 99,18,53,23,16,11,13,15,17,45,102,30,26,32,36,100,98 & 75).A striking feature o f this com m unity is th a t it is form ed o f two strata only, the tree stratum being com posed entirely o f Acacia xanthophloea (8-12 m, cover 2 0 -50%).The second stratum is m ade up o f low-growing (0,15-0,45 m) herbs with sparse cover (usually 7-10%).Dyschoriste depressa is the m ost prom inent, others being Sida alba, Heliotropium ovalifolium, Heliotropium indicum, Cynodon dactylon, Am brosia artemisiifolia and Cardiospermum halicacabum.These latter species tend to be widely distributed and have low cover-abundance values (r or + ) .M any may be regarded as opportunists, reflecting the unstable conditions brought ab o u t by flooding and grazing.In some areas outside the relevés investigated a few well-established Ficus sycomorus are present.These stands are, however, not considered typical.They do, however, serve to indicate the close relationships between this com m unity and the Ficus sycomorus-Rauvolfia caffra Com m unity.

The communities o f low-lying areas
Three com m unities were recognized (Table 3)

The Phragm ites australis C om m unity
This com m unity is dom inated by Phragmites australis, which may be regarded as the character/ differential species (Table 2).U nder protection in the N dum u G am e Reserve (Table 3: relevés 76 & 77), it form s dense stands grow ing to a height o f 2,0-3,0 m with high cover-abundance values and few other species (relevés 76 & 77).W here its vitality is reduced by cutting and burning (relevés 92, 94 and 95), invasion by o th er species occurs, principa'ly Echino chloa pyram idalis and Eriochloa meyeriana th at form a stratum betw een 0,3 and 0,45 m.

The Phragm ites m auritianus Community
Phragmites m auritianus is the dom inant and dif ferential species (Table 2), growing to a height o f 1,5-3,5 m and form ing dense stands with high coverabundance values (Table 3).It is usually associated

Interrelationships
The three com m unities o f the low-lying areas all develop on acidic soils with a pH o f 3,3-6,4 and o f rather variable texture (Table 3).It seems th a t the Phragmites australis C om m unity develops on soils th at are m ore sandy (27-73% ) th an those o f the Phragmites mauritianus and Cyperus fa stig ia tu s-Echinochloa pyramidalis C om m unities (10-57% ).
T he developm ent o f all three com m unities is closely associated w ith M R L (Fig. 6b).This suggests th a t it is n o t so m uch inundation as the presence o f a m ore stable w ater supply th at controls the position o f these com m unities.The Phragmites C om m unities tend to occupy the lower positions, P. australis favouring sw am py areas as opposed to P. m auritianus, which prefers sites where there is w ater m ovem ent, such as river banks.The Cyperus fastigiatus-Echinochloa pyram idalis C om m unity develops only on flat and gently sloping draining areas adjacent to pans, and in depressions th a t rem ain wet for m ost o f the dry season (Fig. 4).
The Cyperus fastigiatus-Echinochloa pyramidalis Com m unity shows m arked affinities w ith the Phrag m ites C om m unities through the extension o f the distri bution o f Echinochloa pyramidalis into these semiaquatic habitats, and with the Cynodon dactylon and Acacia xanthophloea-Dyschoriste depressa C om m uni ties because of the presence o f Cyperus fastigiatus in these higher-lying areas.The degree to which these rather broad distributional patterns reflect the distri bution under natural and undisturbed conditions is debateable, because reduction in the cover o f the Phragmites Com m unities m ight favour intrusion o f Echinochloa pyram idalis, whereas grazing o f Cyperus fastigiatus seems to favour an increase in the Cynodon dactylon com ponent.W here both differential species are grazed, particularly in the areas west o f Sivunguvungu and to the south and south-w est o f K hangazini Pans, their cover-abundance values are lower and a wider variety o f species is present (Table 3: relevé 46, 38 & 39).

Communities o f intermediate areas
Only one com m unity, the Cynodon dactylon C om m unity, has been recognized, covering 171 ha o f the floodplain.It is generally found on gently sloping areas th at become exposed gradually as the floodwaters recede (Table 3, Fig. 6).The soils vary from acidic (pH 4,5) to alm ost neutral (pH 6,8), w ith very variable texture: clay 29-73% , silt 6-3 6 % and sand 11-63% .Cynodon dactylon is extrem ely tolerant of extended periods of dry conditions foi owing exposure, and o f submergence.N o t surprisingly, therefore, the distribution o f the C. dictylon C om m unity does not show m arked relatior ships with either height above M R L or below H F L (Fig. 6a & b).It therefore extends from below m axim um retention level to above high flood level (Fig. 4).C. dactylon may be found in all the com m unities, particularly where disturbance has occurred.

3.1
The C ynodon dactylon Comm unity (Fig. 8) Cynodon dactylon, the differential species (Table 2) for this com m unity, form s dense, alm ost pure, stands  o f considerable extent a ro u n d som e pans (e.g.N am anini-B um be 42 ha, east o f M thikeni and N sim bi 27 ha).It form s a single stratu m (0,2 m) w ith up to 90% cover where conditions are m ost favourable (Table 3: relevés 5, 8 & 10).Elsewhere, where cover is reduced (relevés 27 & 28), and grazing m ore intense, the height m ay be less (0,05 m) and the weeds (A m brosia artem isiifolia, Conyza bonariensis and Poly gonum aviculare) becom e m ore prom inent.These species form an ill-defined stratu m o f 0,2-0,3 m in height.

Interrelationships
The close relationships o f the Cynodon dactylon C om m unity w ith o th er com m unities on the floodplain are em phasized n o t only by the d istrib u tio n o f C. dactylon, b u t also by five relevés (56, 21, 19, 35 & 20) th a t form a nodum o f indeterm in ate ran k betw een the C. dactylon and Cyperus fastigiatus-E chinochloa pyram idalis C om m unities.
It seems probable th a t this com m unity w ould norm ally develop in areas w here th e in u n d n period was to o long for the w oody com m unities and w here it becam e to o dry for com m unities ch aracteris tic o f the low-lying areas (Fig. 6).Since it presently extends into areas th a t clearly show relics o f other com m unities, it m ust be concluded th a t d isturbance has increased the extent o f this com m unity.A t the low er levels it is principally by replacem ent o f Cyperus fastigiatus, which is adversely affected by grazing and tram pling and, a t higher levels, by clearing, w hich reduces com petition and allow s m ore direct illu m in a tion, both o f which favour Cynodon dactylon.

Relevés o f undeterm ined affinity
Six relevés (50, 12, 47, 41, 60 & 62) do n o t a p p ear to fit satisfactorily into the com m unities o utlined above.A lthough they could have been allocated to specific com m unities, thereby im proving th e in fo rm atio n on to tal floristics, the degree o f affinity did no t, in o u r opinion, justify this action.They are included in the table because an advantage o f the B raun-B lanquet m ethod is th at new relevé d a ta m ay co ntinuously be added, thereby facilitating recognition o f com m unities th a t m ay n o t be presently recognizable (W erger 1973).(1967), recognized six form ations (Table 4) o f which five corresponded well w ith those recognized in this study.The m icrophyllous deciduous shrub savanna is, however, difficult to relate to our com m unities, because it includes species such as Sesbania sesban, Echinochloa pyram idalis and Phragmites australis, which appear to exhibit m arkedly different responses to inundation, and it therefore probably form s a tran si tion between " forest and grasses" as has been sug gested by Pooley (1978). N either De M oor et al. (1977) nor Pooley (1978) have, however, attem pted to relate the p lan t com m unites to flooding regimes.

DISCUSSION
The studies by Tinley in 1958Tinley in (published in 1976) ) are the earliest reports on the P ongolo R iver F loodplain vegetation.He recognized tw o fo rm ations, the W erger (1974a) has suggested th a t in areas exposed to extrem e conditions, em phasis should be placed on perm anently recognizable species in phytosociological analysis.This concept was applied to the floodplain vegetation because, particularly above M R L where both in u n d atio n and exposure have to be tolerated, conditions m ay be considered to be extrem e.It reduces the num ber o f species from 64 to 25, b u t does not alter the definitions o f the com m unities (Table 5).
N one o f the com m unities recognized on the Pon golo River Floodplain is unique.They have been recorded from widely separated areas on the A frican continent, on floodplains and areas o f fluctuating w ater levels.
Phragmites australis and P. mauritianus have a world-w ide d istribution (C layton, 1967;Fernandes et al., 1971) V esey-Fitzgerald (1963) recorded Phragmites in Riverine G rasslands and " lakes" where it form ed dense stands on silt banks, sand bars and in lagoons.He did n o t record it from " pans" or " alkaline swam ps and flats" , b u t this may, in the form er instance, reflect the fact th a t the pans being investigated were shallow depressions th at norm ally dried up durin g the dry season.They are clearly different from those o f the Pongolo system.V esey-Fitzgerald did n o t record the specific nam es o f Phragmites and from the d istri bution (C layton, 1967;Fernandes et al., 1971) it seems likely th a t, although b oth P. australis and P. mauritianus could have been present, P. mauritianus w ould be the m ore com m on.
Cyperus fastigiatus is w idespread in the province o f N atal in S outh A frica (Ross, 1972), b u t has n o t been recorded in tropical areas (V esey-Fitzgerald, 1955;1963;D ean, 1967;C ook, 1968;H ow ard-W illiam s & W alker, 1974;Im evbore & Bakare, 1974;R zóska, 1974).Ross (1972) claims th a t C. fastigiatus is closely allied to the m ore tropical C. auricomus Sieber ex Spreng., w hich is also placed w ith C. digitatus Roxb. subsp. auricomus by K iikenthal (in Ross, 1972).H ow ard-W illiam s & W alker (1974) reported C. digitatus from relatively acidic soils (pH below 6,0) in their neutral to acidic m arsh vegetation type.These conditions are sim ilar to those in which C. fastigiatus is found on the Pongolo system.U nlike Cyperus fastigiatus, Echinochloa pyram idalis is w idespread in wet areas where it is often associated w ith a variety o f other species (V esey-Fitzgerald, 1955;1963;D ean, 1967;C ook, 1968;Im evbore & Bakare, 1974;H ow ard-W illiam s & W alker, 1974).The description o f its grow th in floodplain grassland by V esey-Fitzgerald (1963) aptly describes its beha viour in the Pongolo system: " G ro w th starts at the onset o f the rains but the stature depends on the extent o f flooding.U nder optim um conditions the previous season's accum ulation o f rough rots away in the w ater and the new grow th is very vigorous" .A lthough E. pyram idalis is grazed by hippopotam us (Scotcher et al., 1978), the num bers o f hippopotam us outside the N dum u G am e Reserve on the Pongolo Floodplain are small and they do n o t exert a m arked effect.However, as soon as the floods recede and the m arshy areas becom e dry enough for cattle, both E. pyramidalis and C. fastigiatus are grazed.As in tropical areas, even when drying o u t and being grazed, node shoots rem ain green until quite late in the season.In areas where grazing is particularly heavy and where drainage is slightly m ore rapid follow ing the floods, E. pyramidalis may form a m osaic w ith Cynodon dactylon as has been observed by V esey-Fitzgerald (1955) and D ean (1967).
The developm ent o f Cynodon dactylon Com m unities under conditions o f fluctuating w ater levels is well docum ented, both as pure stands (Lea & Van V. W ebb, 1939;Vesey-Fitzgerald, 1955;D ean, 1967;G reenway & Vesey-Fitzgerald, 1969) and in associa tion with other species (Lea & V an V. W ebb, 1939;Burnett, 1951;A nderson & H erlocker, 1973;H ow ard-W illiams & W alker, 1974).These short-grass lawns develop on a wide range o f soils from acid to alkaline in areas where prolonged flooding is not experienced. Dean (1967) noted th a t C. dactylon was easily killed by flooding.On the Pongolo, it regularly tolerates periods o f subm ergence o f up to 150 days, apparently w ithout m uch detrim ental effect.D uring sum m er the C. dactylon aro u n d the pans may be inundated, thereby becom ing an integral part of the aquatic system.Even if it is n o t com pletely inundated, the substrate is generally too wet for access by cattle and goats.Thus it is only during the drier parts o f the year th a t these lawns becom e accessible to terrestrially based grazers, for w hich it provides a valuable source or pasturage.
The distribution o f Acacia xanthophloea along tropical and sub-tropical river courses and in dam p depressions is well docum ented (A nderson & H er locker, 1973;G reenw ay & V esey-Fitzgerald, 1969).Vesey-Fitzgerald (1974)  Ficus sycomorus occurs th ro u g h o u t A frica along river courses, swampy areas and even arid areas where the water table is high (Palm er & Pitm an, 1972).Site preference seems to be for those areas w here drainage is quite good and yet w ater is freely available, because the best stands on both the Pongolo F loodplain (in the N dum u G am e Reserve) and in Lake M anyara N ational Park (G reenw ay & V esey-Fitzgerald, 1969) are along the tops o f the river banks and levees.
This study has illustrated the im portance o f the flooding regime in the developm ent o f the vegetational com m unities.As a result, changes in the p attern o f flooding, both of frequency and perhaps m ore im po rt antly of duration o f inundation and exposure, can be expected to exert a profound effect on the com m uni ties, as has been observed elsewhere (D ean, 1967;Townsend, 1975, Attwell, 1970).W ith the very sandy nature of the soils o f the floodplain, particularly of the levees and around the pans, destruction o f vegeta tion, resulting either directly from changes in the flooding pattern, or indirectly as a result o f allowing cultivation and grazing in areas th a t were form erly too wet, could increase erosion, thereby having a profound im pact on the floodplain system as a whole.

F
i g . 2 .-Anaerial view of part of the Pongolo River Floodplain showing the extensive cultivation along the levees and riverbank.Only small pockets of the Ficus sycomorus -Rauvolfia caffra Community remain outside the Ndumu Game Reserve.so th a t w ater drains off fairly rapidly thereby pre venting the developm ent o f com m unities typical o f the low-lying areas.1. Communities o f the high-lying areas (Fig. 2) Tw o com m unities m ay be recognized: the Ficus sycomorus-Rauvolfia caffra C om m unity w ith two subcom m unities, occupying 406 ha, which is restricted to levees on either side o f the m ain river channel, and the Acacia xanthophloea-Dyschoriste depressa C om m unity, 128 ha in extent, w hich is confined to narrow bands alo n g the m argin o f the floodplain (Figs 3 & 4).1.1 The Ficus sycom orus-R auvolfia caffra Com m unity This com m unity form s a d istinct vegetational unit.Since m ost o f the species present do not occur in any o th er com m unity, they m ay be regarded as character o r differential species ( representation of a cross-section of the Pongolo River Floodplain showing the relative positions of the different plant communities.

F
i g .5 .-Aview of the Acacia xanthophloea -Dyscho riste depressa Community taken during the extended summer floods of 1976.At this time the herb layer becomes incorporated in to the aquatic system.
RELEVËS OF TABLE 3 .AMBROSIA ARTEM ISIIFOLIA L .( 9 5 : +) AZIMA TETRACANTHA LAM. ( 1 3 : + , 1 5 : + , 1 7 : r ) BARLERIA S P .( 1 1 : r ) CENTEMA SUBFUSCA (MOQ.)LOPR.( 3 2 : 1) COTULA AUSTRALIS (SPRENG.)HOOK.f .( 8 9 0 :+ , 74 : EUCLEA UNDULATA THUNB.( 1 5 : r ) GNAPHALIUM S P .(96 : + , 9 2 : + , 94 :+) GREWIA SUBSPATHULATA N .E .B r .(16 : 1 , 13 : r ) HYGROPHILA AURICULATA (SCHUMACH.)HEINE (75 : + , ' Fig. 6a reveals th a t although the distribution o f both com m unities is relative to height below H F L , and therefore to the period o f in u n d atio n while the river is in flood, the Ficus sycom orus-Rauvolfia caffra C om m unity is the m ore sensitive to flooding in th a t it develops on the highest areas.Clearly, however, d istribution o f the com m unity is n o t only determ ined by period o f inundation, because it is absent from areas aroun d the m argin o f the floodplain, even w here they m ay have an alm ost identical flooding regime to th a t o f the levees.It seems likely th a t season al availability o f soil m oisture acts in conjunction w ith inundation, because proxim ity to the river w ould prevent the developm ent o f the very dry conditions w hich arise along the edge o f the floodplain during the dry w inter m onths (unpublished data).The seasonally d rier conditions along the m argin o f the floodplain elim inate M onanthotaxis caffra and other hygrophilous w oody species, while the presence o f occa sional Ficus sycomorus suggests th a t it m ay be som e w hat m ore to leran t.A It lough m ore favourable soil m oisture conditions m ight be found a t low er eleva tions along the floodplai 1 m argin, developm ent o f the w oody com ponent is prevented by the longer period o f inundation.The com bination o f these factors is probably responsible fo r the d istinct separation o f the Ficus sycomorus-Rauvolfia caffra and Acacia xanthoph'oea-Dyschoriste depressa C om m unities.T he Acacia xanthophloea-D yschoriste depressa C om m unity tends to develop in slightly lower-lying areas than the Ficus sycom orus-R auvolfia caffra C om m unity, b u t above M R L (Fig.4).It is, therefore, inundated only while the river is in flood, an d the . The Phragmites australis and the P. mauritianus C om munities together occupy an area o f c. 234 ha, m ost of which (65% ) is in the N dum u G am e Reserve.The Cyperus fastigiatus-Echinochloa pyramidalis C om m unity is one o f the largest, covering c. 2 471 ha, with particularly extensive stands occurring west o f Tete and N sim bi Pans and in the N dum u G am e Reserve (Figs 3 & 7).
-S e q u e n c e o f R e le v é s in P h y to s o c io lo g ic a l T ab le -F ig .6.-Position of the relevés relative to: (a) height below high flood level; (b) height above maximum reten tion level.Relevés are ordered in the sequence defined by the phytosociological Table 3. *Nodum of indeterminate rank.

F
stand on the north-western edge of Tete Pan has been reduced in size and in height.Most of the area is now a mixture of Cyperus fastigiatus, Echinochloa pyramidalis and Cynodon dactylon forming a stra tum of about 50 cm in height.

Fig
Fig. 8.-The Cynodon dacty lon Community is grazed extensively during the late winter period when pasturage elsewhere is low.

R
iparian F orest, w hich included the disturbed and undisturbed vegetation o f the levees and high-lying m argins o f the floodplain, and the A quatic and M argi nal Pan V egetation.This broad classification, which does n o t take in to account the observed m arked influence o f the flooding regime, does n o t agree well with the com m unity distinctions draw n up in this study.M ore recently, D e M o o r et al. (1977) and Pooley (1978) have reported on the vegetation w ithin the N dum u G am e Reserve (Fig. 1).O nly a p o rtion o f this Reserve includes the floodplain o f the Pongolo River.De M oor et al. (1977) using the system o f Fosberg concluded th a t the cyclic and serai status o f groves o f A. xanthophloea were substantially influenced by drainage conditions.A consequence o f this is th a t trees along the lake-shore may succum b during periods o f high lake levels (G reenw ay & Vesey-Fitzgerald, 1969), a situation observed on the Pongolo F loodplain particularly in the vicinity of M zinyeni Pan (Figs 1 & 5) during the unnaturally extended floods caused by discharge o f w ater from the P ongolopoort D am after its co nstruc tion.

TABLE 4 .-Comparison of the formations determined by De Moor et al. (1977) with communities of this study De Moor et al. (1977) Present study Formation 1 Dry season deciduous forest 4 Tall evergreen graminoid marsh 4a Tall evergreen graminoid marsh 4b Seasonal orthophyll tall grass 4c Microphyllous shrub savanna deciduous 5 Seasonal orthophyll meadow Community Ficus sycomorus-Rauvolfia caffra Phragmites mauritianus Phragmites australis Cyperus fastigiatus-Echi nochloa pyramidalis Probably transitional be tween Cyperus fastigia tus-Echinochloa pyra midalis and Ficus syco morus-Rauvolfia caffra Cynodon dactylon
, although P. mauritianus tends to be m ore tropical and is frequently encountered in swam py and seasonally flooded areas (G o rd o n -G ray & W ard, 1971; H ow ard-W illiam s & W alker, 1974; R zóska, 1974).T he distinction in ecological preference between P. australis, which favours standing w ater, and P. mauritianus which prefers m oving w ater, supports the observations o f G ord o n -G ray & W ard (1971).H ow ard-W illiam s & W alker (1974) reported sim ilar environm ental conditions in stands o f P. mauritianus in Lake C hilw a, although it was also present in alkaline sw am ps.In other parts o f C entral A frica,