Remotely sensed imagery serves as a tool for accelerating soil survey
activity to the extent that features in the imagery can be related to
soils in their landscape. This has been clearly proven for discriminat
ion of broad geomorphic units in Landsat and SPOT imagery. However, fa
rm-level soil mapping is likely to require discrimination of soil unit
s at a scale where topographic clues are not enough for class separati
on. The capability to distinguish soil classes in remote imagery is ve
ry much related to inherent spectral behavoir of diagnostic soil compo
nents. This is especially true in Brazil where soil color is used for
class definition, principally as a covarying property with iron oxide
form and content. This study investigated the potencial for soil spect
ral discrimination of a chromosequence of latosols (oxisols), from the
Central Plateau of Brazil, using a portable field radiometer sensitiv
e in the 380-875 nm reflective wavelength range. Pairwise comparison o
f spectral reflectance of soils reveals wavelength regions of maximum
spectral discrimination, accentuating the 450-520 nm (blue) region of
opaque behavior of the hematite-rich Dark-Red Latosol (LE) in contrast
to the goethite rich Red-Yellow Latossol (LV).