an area 
f river 
f deposits 
he region 
s. On the 
r and more 
temperatu- 
noisture 
is mesic. 
ial valleys, 
butary ri- 
s appear 
ne plain, 
odify the 
pments for 
bbles, fa- 
atrix. 
um and li- 
le, but al- 
»getation 
itions at 
1fficiently 
1cluding 
, 1975, 
ne most crop- 
5 (orchards, 
sen lines 
to UTM coor- 
mapping po- 
80x 80 m. 
] from topo- 
rived from 
maps were 
photogra- 
soil map 
ractive sys- 
ing and ve- 
Seven training sites covering 5 percent of the study area were analy- 
zed to perform an unsupervised classification of the basin of the Tajuña ri- 
ver, using a clustering algorithm from tne ERMAN system. The iterative algo- 
rithm selected found all natural classes existing in the image and divided 
the space into groups of sample points of similar spectral response (Figure 
2). Fifty six natural classes were obtained from training sites and then 
grouped into fourteen spectral classes based upon: 
1. mean spectral response values of all four bands. 
2. configuration er form of spectral curve 
3. ratio between the sums of visible bands and infrared bands. 
4 
total magnitude of the spectral response values of all four bands. 
3. RESULTS AND DISCUSSION 
The fourteen spectral classes (Figure 3), with their mean vector and 
covariance matrix, were used as input to the multivariate normal maximum li- 
kelihood classifier of the ERMAN-II system. 
The classification results differentiate soil surface features into 9 
bare soil classes and 5 classes of soil covered with vegetation. These could 
be identified using the ratio visible bands-infrared bands. The ratio is 
higher than 0.83 for bare soils and lower than this value for soils with ve- 
getation. There was some doubt in a few classes with ratio slightly over 0.83 
but their low reflectance values allowed to identify them wiht the scarpements 
physiographic unit, but scarpments in shadow without vegetation. 
Table I showws: the correspondence between the fourteen spectral classes, 
reflectance values, physiographic units and dominant subgroups of soil which 
have been identified with them. The subgroups of dominant soils which appear 
in the table belong to the Orders Aflsisols, Entisols andInceptisols and to 
the Suborders Xeralfs, Orthents, Fluvents, Ochrepts. 
The identification between spectral classes, physiographic units and 
soil subgroups must be understood only in a general sense. We caanot assure 
that an spectral class corresponds exactly with a specific physiograpbic unit, 
except in the case of scarpments wiht natural vegetation. In general only an 
important proportion of a given spectral class corresponds to the same phy- 
siographic unit. 
Table II contains the result of combining the spectral classes follo- 
wing the criteria of.proximity in the values of the sum of spectral respon- 
ses and the ratio visible bands/infrared bands. This combination simplifies 
the physuigraphic scheme of the classification map. 
Table III summarizes the results of the classifcation, giving the per- 
centage and the number of hectares occupied by each class as well as the soil 
status: bare soil, cultivated, natural vegetation, with or without shadow. The 
area without vegetation represents 92.25% of the total area while the remai- 
ning 7.75% is covered with vegetation. 
Two digital maps of the area were produced, one wiht the fourteen cla-