classifi
-aphical
'gement
Univer-
ountries
ce such
‘he base
required
pe ready
ouple of
the ex is-
erplanes
st 1982.
resented
an Nov.,
hode der
Society
stitution
nki, July
ee
"Im
QUALITATIVE AND QUANTITATIVE ASSESSMENT OF OIGITAL SEPARATION OF SOILS
By
Ayodele Fagbami,
Dept. of Agnonomy, University of Ibadan
Ibadan, Nigeria.
ABSTRACT
A ground truth (GT) map, produced by the interpretation of large scale 8 & W aerial photo-
graphs was compared with digital maps produced from the LANDSAT data of the savanna
around Makurdi in Nigeria. The GT maps were digitized, grid — converted and then aggregated
while the LANDSAT data was resampled for the purpose of rectification on ERDAS 400 micro-
computer. A visual comparison of the maps was done by overlaying the digital maps on the GT
map on the colour monitor while the histogram listing provided a rough quantitative compari-
son.
°
The two algorithms of supervised classification (Maximum Likelihood and Minimum Distan-
ce) produced similar results but the third unsupervised classifier algorithm, Cluster analysis, pro-
duced a far simpler map that is ideal for reconnaissance soil/resource survey. The major landforms
were recognized by the computer and the reflectance — based classification resulted in misgrou-
pings because of (1) the predominant well drained soils tend to be grouped together just as were
poorly narrow units were masked by surrounding pixels and were therefore wrongly classified.
The vegetation cover of the tropical savanna is a major problem in digital classification of
soils,
INTRODUCTION
Most of the previous works :eported on the application of LANDSAT to mapping of the
tropical savanna soils has been done by monoscopic interpretation of the photographically pro-
cessed Landsat imageries (Elbersen, 1973; Fagbami, 1980). However, extensive investigation of
digital analysis in soil mapping has been done in the industrialised countries of Europe and North
America (Weismiller et al, 1977, Kristof et al, 1977, 1980, Hinzel et al, 1980). In Nigeria this is
the first attempt to apply digital analysis to the problem of soil survey in the savanna which are
the only parts of the country that have good (i.e. cloud-free) LANDSAT imageries.
THE METHODS
1. Production of ground truth map: A soil map to which digitally produced maps from LAND-
SAT are to be registered and compared was produced by a combination of aerial photo interpre-
tation of B & W aerial photographs and field checking soil examination. The aerial photographs
flown in 1978 at a scale of 1: 10,00U were interpreted and the lines were drawn directly on
overlaps over the 1: 15,000 controlled mosaics.The entire area was mapped into six sheets,
3 in the north and 3 on the south of the Benue River Fig. 1.