tion of 
arch. 
k gras- 
shrubs 
; and 
een 40 
of the 
| COVET 
S been 
native 
V CTODS - 
nd beef 
h little 
grostis 
nber of 
which 
2gion. 
o their 
10). À 
a were 
1) and 
3). Six 
e used 
tive of 
  
  
  
sion in TM color composites of single bands (2,4,5 ; 
3,4,5;3,4,7 )or in single indexes such as SAVI (8) or band 
3/4 ratio. 
The objective of this research was to develop an 
enhacement to discriminate the earlier stages of water 
erosion through Landsat Thematic Mapper images in 
this particular region of Argentina because none of the 
usual enhacements were good enough. 
2.1.Background: 
The Miraflores river basin is located between 22 and 25 
Southern latitude and 65 to 67 Western longitude in the 
Argentine Puna region . 
The climate is semi-arid showing wide temperature 
fluctuations between day and night. Winter is the cold and 
almost dry season, with a mean rainfall of 10 mm; while 
summer is relatively warm and rainy (290 mm) 
The arid conditions produce a scant vegetation cover 
which offers negligible protection against erosion. 
The high erodability of the soils, also enhances the action of 
the winds and torrential rains from November to March. 
The semiarid steppe is mainly composed of tussock gras- 
ses (Pennisetum chilense;Festuca scirpifolia) and shrubs 
(Parastrephia | lepydophylla; Baccharis boliviensis and 
Tetraglochin cristatum) with an average height between 40 
and 80 cm. 
In Winter the vegetation covers 40 to 60 % of the 
ground. In Summer ( rainy season) the vegetation cover 
increases up to 80% 
Since the pre-colombian times, this region has been 
subjected to the practices of shifting breeding of native 
camelids,llamas and vicufías in open fields. The few crops 
are grown in narrow terraces on the foot hills . 
Spanish colonization and the beginning of sheep and beef 
cattle augment the impact over this ecosystem with little 
resilience . The introduction of pastures such as Eragrostis 
curvula and the use of fences have increased the number of 
cattle heards, especially beef leading to overgrazing, which 
is considered the principal erosive factor (11) in this region. 
3. METHODS: 
3.1.Identifying water erosion degrees in the field: 
After the stratification of the Miraflores river basin, 
seven sampling areas  (6)were choosen according to their 
representativity , forty three "test sites" were selected (10). A 
correlation test was performed between landfeature reflec- 
tance and TM bands. 
In RUSLE a subfactor method is used to compute SL as a 
function of four subfactors :prior land use,canopy,ground 
cover and within- soil effects. 
Water erosion features were measured in the field according 
to the "Desertification Method" 6). 
3.2. Digital Analysis 
Two sets of Landsat Thematic Mapper (TM) data were 
used, one from the Winter dry season (October,1991) and 
the other from the Summer rainy season (March,1993). Six 
bands (except thermal band) from Landsat TM were used 
for this study. 
A regression analysis was performed with the objective of 
combining three independent bands into one master image 
suitable for visual interpretation of water erosion features. 
Jeffries -Matusita distance (14) was tested to choose the 
optimum bands with the highest accuracy. 
The selected bands were used as input data for the 
indexes and for deriving the Principal components statistics 
(PCA).This PCA is used as a method of data compression 
reducing the dimensionality of the data . The bands of PCA 
data are non correlated and independent and are often more 
interpretable than the source (9). " Selective PCA" (3) 
were applied in this study. Selective means the use of a 
subset of bands as input. 
Principal components are based on the eigenvectors of the 
correlation matrices ,in this study the matrix was extracted 
from a subset area where water erosion is moderate, so that 
the rotated components can highlight features related to. 
4. RESULTS and DISCUSSION: 
The different degrees of water erosion found in the field 
are shown in Tables 1 and 2. 
Table 2 shows soil loss valued according to each method, as 
well as those obtained through direct observation according 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998 385