47°20' 47°} 
22°03' 
1 
4 
S 
4 
1 
22°10’ 
LEGEND 
1=LV — Main River ESCALE 
2- LR TH : o I 2Km 
3- PV ~~ Tributaries EE 
4-Hi Q Soil Limits 
5-TE 
6— LE 
  
  
Fig. 4 - Spectral map of soils of the study area + field work map. 
patches of soils. Nevertheless, the new 
map presents a greater richness of 
details, has better definition between 
the soil boundaries, and reveals 
important transitions between soils. 
A comparison between the two 
soils maps was made with a 2x2 mm grid 
sample (5425 grid cells, each 
representing 200 x 200 meters on the 
earth surface). The fieldwork map is 
used as the reference. Comparing the 
homologous squares, the percentage of 
coincidence (86.9%) was calculated with 
the following formula: 
Percent = Number of correct predictions 
of map units x 100 
Corret Total number of map units in 
the sample 
Figures 2 and 4 indicate that the 
main discrepancies appear in the 
spectral-data map in patches of 
homogeneous or associated soils that are 
absent from the map based on field work. 
Subsequent field checks have verified the 
existence of the patches. Of the total 
13.1% error in coincidence, 7.8% resulted 
from deficiencies in the map based on 
field work, whereas 5.3% were from 
insufficiencies of the spectral map. of 
the 5.3%, 1.7% derived from patches of 
300 
hydromorphic soils whose reflectances are 
masked by dense natural vegetation. We 
note that both types of errors are from 
omission of detail, not from excessive 
detail that was incorrect. 
The results of the study match 
well the existing literature. The 
spectral differences between soil types 
are derived mainly from the different 
percentages of organic matter and iron 
oxide (Montgomery and Baungardner, 1974, 
and Stoner and Baungardner, 1981). In 
this case, the soils in the area are 
quite distinct between themselves, 
especially in relation to iron oxide. 
When the best spectral 
combinations are analyzed, the TM4 and 
TM5 bands were found to be present in the 
composites considered to be the most 
informative. The spectral curves shown 
in. Figure 1 confirm that the greatest 
differentiations occur precisely in the 
intervals that correspond to bands 4 and 
5. Because the soils of the study area 
have quite similar concentrations of 
organic matter (C/N relation from 9 to 
12), the spectral differences are more 
dependent on the quality and quantity of 
iron oxides present (Fe; 03). The average 
percentages of iron oxides in the soil 
types are: LR = 34%, TE = 26%, LE = 
22%, PV = 10%, LV = 5%, AQ = 2.0%, and 
  
HOMO I 
HON QE Hct Hh 
moo Qr~Q8 tt 
Qc ct N HS Oo 
not 
m SCH 
B.