S. Ringrose et al., USE OF LANDSAT THEMATIC MAPPER DATA TO ASSESS SEASONAL RANGELAND CHANGES IN THE SOUTHEAST KALAHARI, BOTSWANA, Environmental management, 23(1), 1999, pp. 125-138
Management problems arise in semiarid rangeland that are characterized
by marked wet and dry seasons because or forage deficiencies in the d
ry season. These natural vegetation rangelands can sustain livestock a
ll year long when forage and senesced grass are available into the dry
season. Seasonal range condition data are required to provide a basis
for pasture management to help locate dry season cover and thereby mi
nimize overstocking and degradation. The generation of seasonal data u
sing Thematic Mapper (TM) imagery was undertaken to assess changes in
natural vegetation cover in the southern Botswana Kalahari. Visual ana
lysis oi spectral reflectance curves, the development of spectral sepa
rability indexes, and conventional classification analysis techniques
were used to identify and differentiate rangeland features. Results fr
om reflectance curves indicated that most rangeland cover types could
be preferentially distinguished using mainly wet season data, especial
ly on the longer TM wavebands, and that range feature differentiation
was more problematic on darker soils than on lighter soils. Spectral s
eparability indexes (SSIs) confirmed that range feature separation var
ied considerably as a function of waveband and was more effective in t
he wet than the dry season. The SSIs also showed that range feature di
fferentiation in both seasons was most effective using a combination o
f the chlorophyll absorpance band (TM3) and two mid-infrared bands (TM
5 and TM7). Wet season data were more effectively classified in terms
of range features than dry season data although some class similarity
was inferred across the two classified data sets. The work shows that
overall trends may be generated by comparing seasonal data sets, there
by providing an overall basis for dry season decision making. However,
particular problems arise within the dry season data sets probably be
cause of spectral similarities between shadow and darkened vegetation
cover, thereby implying that further work is needed.