Zacharias, Peter John Kenneth.Le Roux, Noel Peter.2014-01-132014-01-1319951995http://hdl.handle.net/10413/10335Thesis (M.Sc.)-University of Natal, Pietermaritzburg, 1995.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.enGrassland ecology--KwaZulu-Natal.Umtamvuna Nature Reserve (KwaZulu-Natal)Theses--Grassland science.Thesis