accuracy, as well as errors of omission and commission for each class 
following Short (1982) (Table 2). The percentage agreement by class was 
determined by dividing the number of correctly classified pixels for each 
class by the number of ground truth test pixels plus the number of 
commissions for that class. The overall mapping accuracy was obtained by 
adding the number of correctly classified pixels in each class and 
dividing by the total number of ground truth test pixels. 
Figure 2. Land development and irrigation potential map centered 
on the Santa Luzia, Paraiba area. Information generated 
from earth resources data derived through computer-assisted 
processing of SPOT digital multi-spectral imagery. 
RESULTS AND DISCUSSIONS 
A more rigorous, quantitative measure of accuracy performed using the 
test areas indicated that some categories were classified and mapped more 
reliably than others. Table 2 presents a suimiary for the maximum 
likelihood classification results of the Santa Luzia SPOT imagery. The 
overall classification accuracy of 92.3% is good with respect to the 
classification categories used for this study. The overall 
classification accuracy, however, is often not a good indicator of 
reliability because it accounts for pixels that were correctly classified 
but does not measure commissions (false inclusion of pixels). Brennan et_ 
al. (1989) explained that to evaluate accurately the reliability of this