352 
Table 1. Santa Luzia area land use and land cover classification 
w. 
Symbol 
Description 
Pixels 
% 
1. 
W1 
Deep to very deep water 
3,137 
0.3 
2. 
W2 
Moderately deep to deep water 
2,994 
0.3 
3. 
W3 
Very shallow to shallow water 
5,275 
0.5 
4. 
U1 
Dense urban area 
11,469 
1.2 
5. 
U2 
Sparse urban and barren rocky land 
23,013 
2.3 
6. 
CF1 
Cotton cultivated and fallow land 
29,571 
3.0 
7. 
MF1 
Mixed cultivated and fallow land 
19,821 
2.0 
8. 
CPI 
Mixed cultivated and pasture 
106,459 
10.7 
9. 
Al 
Alluvial with dense shrubs/trees 
17,943 
1.8 
10. 
A2 
Alluvial cultivated and eroded land 
42,429 
4.2 
11. 
FI 
Sparse caatinga forest and rocky 
49,884 
5.0 
12. 
F2 
Sparse to moderately dense caatinga forest 
285,754 
28.6 
13. 
F3 
Moderately dense to dense caatinga forest 
84,347 
8.4 
14. 
F4 
Dense caatinga/mixed forest/undulating land 
147,773 
14.8 
15. 
F5 
Dense caatinga and mixed forest on hills 
170,111 
17.0 
Total 
1,000,000 
100 
The first six maps were prepared by recoding the digital land use and 
land cover classification. The recoding was possible because of the high 
degree of correlation of land use and land cover with the features of 
other maps. Field investigations conducted at the sites in March-April 
and October-November 1988 confirmed this relationship. The first four 
earth resources information layers derived from analysis of the SPOT 
imagery and collateral data were weighted for their relative importance. 
Finally, these weighted variables were combined in a logical model to 
generate a suitability classification for land development and irrigation 
potential (Figure 2). 
Accuracy Evaluation 
Accuracy assessments of the transformed and non-transformed SPOT image 
were conducted to compare the test areas of known reference data with the 
same areas on Level II land use and land cover classification on a pixel- 
by-pixel base produced by supervised classification. All the preselected 
polygons were displayed over the tor. Then the test areas were selected 
to allow the data file coordinates for polygons to be digitized directly 
from the image. Three to four test areas (polygons) were selected for 
each original class. The resulting test area file was then digitally 
overlaid on the classified Santa Luzia land use and land cover map and a 
pixel-by-pixel comparison was performed. Accuracy was assessed by 
intersecting the maximum likelihood classification results with their 
respective ground truth digital map which revealed the per-category 
agreement and disagreement. The data for the supervised (original ) 
interpretation and verified interpretation were then put in a table 
showing row-column totals, percent agreement by class and overall mapping