Sokolic, Franko.
Abstract:
The Eastern Shores of Lake St Lucia forms part of the ecologically important Greater St Lucia Wetland Park, designated a World Heritage Site in 1999. The landscape is characterised by surface water, a high water table and numerous wetlands. Little is known about the distribution and temporal fluctuations of this surface water and its relationship to the wetlands. This study uses remote sensing to examine the relationship by mapping the extent of seasonal, ephemeral and permanent surface water on the Eastern Shores. Much of the surface water occurs in conjunction with emergent vegetation and is not easily mapped using hard classification methods. Neither a cluster analysis nor a maximum likelihood classification were able to map the subtle variations of the water-vegetation mix. Much more successful was the application of spectral mixture analysis using image endmembers of water, woody vegetation and non-woody vegetation. This technique was applied to seven Landsat Thematic Mapper images from 1991, 2001 and 2002. Steep slopes, forests and bare sand were masked out prior to classification. Maps of water extent were produced for each of the seven study dates. Mapping accuracy was verified against rainfall, with high correlations being obtained against rainfall accumulated over six months and longer. Long-term rainfall patterns were reflected in the surface water distribution, with inundation being more extensive when accumulated rainfall was high. Fire scars reduced the accuracy of the spectral mixture analysis but these scars could be identified from the thermal image bands. The largest open water body in the study area was Lake Bhangazi. Large extents of surface water were also found in the Mfabeni swamp and the wilderness area to the north where water concentrations of 90% were measured during wet periods. Surface water present near Brodies Crossing during wet periods was less evident when rainfall was lower. No inundation was recorded in the areas to the west and south-west of the Mfabeni swamp or in the southern parts of the study area. The techniques used in this study were developed into a water mapping protocol that uses image endmembers and spectral mixture analysis to measure water concentration.
