Browsing by Author "Mbolambi, Cikizwa"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    Assessment of coastal vegetation degradation using remote sensing in False Bay, South Africa
    (Stellenbosch : Stellenbosch University, 2016-12) Mbolambi, Cikizwa; Luck-Vogel, Melanie; Stellenbosch University. Faculty of Arts and Social Sciences. Dept. of Geography and Environmental Studies.
    ENGLISH ABSTRACT: The coastal zone, the interface between land and sea, faces much pressure from human activities. These coastal pressures make it difficult for the coastal zones to fulfil their natural functions, so threatening the state of coastal environments and making them vulnerable to coastal disasters and degradation. This study aimed to test whether remote sensing techniques can be implemented to assess the intactness of terrestrial coastal vegetation at the high spatial resolution required for coastal management. The study focused on the northern False Bay coast, Western Cape, South Africa. The research used is a modification of the method developed by Lück-Vogel, O’Farrell & Roberts (2013) which involved image segmentation and a habitat intactness index using image derivatives. The procedure used Worldview-2 (WV-2) images of high spatial, spectral and temporal resolution acquired on 25 February 2014 and 11 October 2014. Both images were pre-processed and segmented into meaningful objects using object-based image analysis (OBIA). Five image derivatives and the eight spectral bands were stacked into a single image to extract field-informed training points. Regression analysis was performed on eight spectral bands and five image derivatives to evaluate the most suitable bands to produce a habitat intactness index in a subsequent decision tree classification. Decision tree classification was generated using two spectral bands, namely the RED and NIR1 bands. These bands were chosen because they gave the best regression results and they are available in most sensors. The bands were also chosen because the study deals with vegetation assessment. The overall accuracy of the results was 80.5% which was a satisfactory result with a kappa value of 0.75 (75%) that indicates a substantial agreement between the remotely sensed result and the reference data. A key finding is the importance of seasonality to delineate natural and alien vegetation which is better achieved in the dry season. Validation of the results was done using the field-validation points of a field visit conducted in June 2016. The output maps generated for habitat intactness consisted of five habitat intactness classes namely highly, moderately and lightly degraded, intact vegetation and alien vegetation. The output maps can be used to inform coastal managers about conservation at a local scale. It is recommended that validation of remote sensing results be done in the same season that satellite images were taken.