Accurate, up-to-date information on land cover and land-use is essential fo
r strategic planning, sustainable resource management and environmental res
earch. A land-cover database for South Africa, Swaziland and Lesotho, desig
ned for use in a variety of regional modelling, monitoring and analytical e
ndeavours, has been created. Database development involved (1) a stratifica
tion of natural land cover and human land-use, (2) a contextual classificat
ion of hardcopy multispectral data derived from Landsat Thematic Mapper (TM
) imagery collected from 1994 to 1996, and (3) a stratified post-classifica
tion accuracy assessment using a large sample of field data. The resultant
database has yielded substantial information to characterize the landscapes
of South Africa. The analysis uses data recorded by province, primary hydr
ological catchment and vegetation biome to explore the land-cover character
istics of South Africa, Swaziland and Lesotho. Aggregated results show that
South Africa comprises the following general surface types: 5.8% forest an
d woodland; 0.3% forest; 17.6% thicket and bushland; 34.1% shrubland and lo
w fynbos; 0.2% herbland; 21.3% grassland; 12.2% cultivated lands; 1.5% fore
st plantations; 0.4% waterbodies; 0.5% wetlands; 0.2% barren rock; 4.9% deg
raded lands; 1.1% urban/built-up lands; and 0.14% mines and quarries. The a
ccuracy of the database ranges from 51% to 93%, depending on geographical a
rea, with corresponding kappa index values from 35 to 88.