(4) The retlectance classification of the terrain by the computer had no apparent regard for 
soil similarities or dissimilarities. It went by the reflectance of the open soil surface, where it 
existed, or by the reflectance of cover types as was most often the case, which often have no 
ditferentiating relationships to the soils. Certainly, if the soils were exposed, the computer would 
probably do a classification based on their reflectance characteristics — and do it well. 
(5) On the basis of the above, the GT map was simpler than the computer — generated maps. 
Of the classifier algorithms, the clustering algorithms produced the simplest map. Fig. 4 
For a first generation (reconnaissance) soil map theretore, the clustering algorithm would be 
an ideal one to use; it is faster than the conventional method and perhaps would produce more 
leads to follow up. Without interaction with the computer therefore, a tentative reconnaissance 
terrain (soil) map may be produced with the help of the computer, in a few hours that would 
make the job of field checking, shorter and faster. 
m P DELE 
PN RA P RN MR: ven E 
A NAA 
NV CAS 
N AAA) 
a uf Pd 
MA A IRONIA 
  
Fig, 4. Note the relative simplicity of this map by clustering algorithm. Al- 
though 33 clusters were classified, a few were of large extent wh ile the majori- 
ty are of very minor extent. 
   
    
  
  
  
    
   
   
   
  
   
   
   
    
Fi 
pa 
(se