The main aim of this study was to classify and map the threatened coastal grassland communities of the 3 257 ha Umtamvuna Nature Reserve (UNR) in KwaZulu-Natal (30°07'30" to 300 11'05"E; 30°55'00" to 31°04'30"S). Secondary aims were to relate past management and selected environmental variables to community composition and to develop guidelines for monitoring. Alpha diversity was measured using a Whittaker plot and revealed 119 species. A pilot study to test the efficiency of botanical techniques showed that a point based technique (nearest plant method in a 20 X 20 m plot) was efficient (52 minutes for recording 200 points), but recorded only 23% of the species. By increasing the number of points to six hundred, 34% of the species were recorded in 178 minutes; the same time was required to randomly place 30 quadrats (50 X 50 cm), which revealed 80% of the species. Tests for replicate similarity showed a high retrieval of internal association (PS = 86%), using abundant species only and 100 points per plot. The point based technique was thus efficient in detecting abundant species and was acceptable for producing a classification, especially in this case where a comprehensive species list already existed. Indirect gradient analysis (TWINSPAN) identified six grassland communities. An ordination using detrended correspondence analysis (DCA) contributed towards the community classification and grazing and fire frequency gradients were inferred from this ordination. Protea roupelliae communities were common but did not influence grass species composition . Canonical ordination revealed that, of the eight environmental variables measured, 'distance from the sea' strongly affected species distribution (r= 0.83). Cost effectiveness was considered in the development of a monitoring programme. Point based monitoring techniques favoured by sourveld researchers in KwaZulu-Natal were found to be inefficient, particularly for studies requiring the measurement of both species richness and community composition. Randomly located 100 X 100 cm quadrats, located in selected sites which represent previously identified communities, was more efficient. This study contributed towards a refinement of information on the grassland communities of KwaZulu-Natal and supported the use of point data for the classification of grasslands not previously studied. It also demonstrated that point based techniques were not suitable for meeting all grassland monitoring requirements. / Thesis (M.Sc.)-University of Natal, Pietermaritzburg, 1995.

Climatic oscillation during the Quaternary resulted in fragmentation of once more continuous ancient floras and a series of invasions of different floras into the Pondoland Centre in response to climatic change, with some elements invading more than once. This implies both a temporally complex and a floristically complex origin for the extant flora of the Centre. Data derived from analysis of the melange of extant floristic elements in the Pondoland Centre is presented in support of this hypothesis. A synfloristic comparison of Oribi Gorge (OGNR) and Umtamvuna Nature Reserves (UNR) is the basis for this study. A comparison of the species lists generated for OGNR and UNR reveals that 24% of the 1514 angiosperm species are shared. The familial composition of the reserves is similar, with eight of the ten most diverse families contributing a similar proportion of species to the respective floras, with the exception of Acanthaceae. The ten most diverse families comprise a comparatively small proportion of the respective floras; this is indicative of high diversity over long geological periods, i.e. of refugia. Analysis at the generic level revealed similar consistancy between the two gorge floras. Approximately 4% of the UNR species and 2.3% of OGNR species are Pondoland Centre endemics. Approximately 40% of the endemic species are shared by the gorges. Data reveals that both palaeoendemic (predominantly woody, forest taxa) and neoendemic (predominantly herbaceous or suffrutescent, grassland taxa) species occur. The Pondoland Centre is thus a refugium for species trapped on the Msikaba Group sandstones as a result of climatic oscillation during the Quaternary, and a centre of neoendemism. OGNR and UNR floras include Cape, Afromontane and tropical elements (11.3%, 2.8% and 19.1% respectively for OGNR and 16%,3.4% and 15.3% respectively for UNR). Seventy - two percent of Afromontane species are shared, indicating a relatively recent invasion(s) and lor the relative proximity of the gorges to the Afromontane flora. The lower species overlap in the Cape element (39.2%) of the two gorges implies that the invasion of the element is ancient, with subsequent extinction of many of the taxa from OGNR in response to climatic change. It is also possible that this invasion was initally less successful; fewer species found refuge in OGNR. The tropical element comprises the largest proportion of the flora in both gorges and many (ca. 50%) of the species are shared. The degree of species overlap indicates that the invasive flora was either initially more similar or that it is tess prone to extinction. The level of overlap could also suggest that the invasion was more recent than that of the Cape taxa. The tropical element is larger in OGNR and the Cape element is larger in UNR. This is partly due to the gorges' respective proximities to the tropical and Cape floras. This trend is echoed in the endemic data. iv The conservation status of the endemics and of the Pondoland vegetation types is established and recommendations for further research are made. The data support the establishment of a larger UNR, the maintenance of both Umtamvuna and Oribi Gorge as formal nature reserves and the establishment of a new reserve (or reserves) within the Pondoland Centre. / Thesis (M.Sc.)-University of Natal, Pietermarizburg, 1998.