A benchmark approach to map rangeland condition from Landsat TM imagery in the Mpumalanga Province of South Africa

摘要

A procedure to test the ability of various image processing parameters for mapping land degradation is presented. Besides mapping areas of low vegetation cover, the Landsat TM data is challenged with mapping areas of negative species compositional change in the grassland biome of Mpumalanga, South Africa. Floristic data were collected during extensive surveys as part of a natural resource monitoring system. The study area was stratified into floristically uniform environmental units (i.e. veld type habitats). The rangeland condition mapping from Landsat images followed a benchmark approach by using fence line and water point effects. The applicability of various image processing techniques (e.g. Bare Soil Index (BSI), Mean Standard Deviation Index (MSDI) and Normalized Difference Green Red index (NDGR)), in assessing raugeland condition (including species change) were evaluated. Within every veld type habitat, multivariate analyses were employed to place sampling sites along a degradation gradient based on their floristic composition.