—
VOTRE MAIL TAR RE y —À— m 1
T +
"n
(6) Aggregating similar physiographic—
'egard for ; such as : soil units i i
: suc: agor S in the sian ; ;
where it volley UE of the western sand island and the atum f Prior to classification
Ww 0
1 have no Fig. 5. gu shale and that of ihe remaining upland did t e Benue banks or the
ter would not improve classification.
ited maps.
ı would be :
duce more |
nnaissance
that would
Es:
; hs N th y NY
RM
in “we
LE gel
+
p
7
m
Rye
i 8 isis) (Ion by maximum likelihood algorithm after aggregating 2
irs of units. Note that there is no improve gi i
eum aru p ment over the original grouping
ys
«AANA A
at
‘Etat
i (7) The computer is confused by the reflectance characteristics into grouping dissimilar
i atures together as a class such as the various misgroupings of settlements, sand bars, the ai
ield and even cloud together as a group because of their high reflectances T
: (8) The classification signatures identified at training sites as physiograp hic-soil unit |
tremendousoverstr etching problems, For example, even though many features such RIS i
sand island ridges, depressions and levees and most of the physiographic soil units ay
classified, other units with no soil, morphological or geographical affinity to hr
iii For this reason, many of the smaller physiographic-soil units have C lr
ce area values vis-a-vis the large units. This highlights the danger of misinventorvi i
by the computer if the classification was completely unsupervised An Sens