. The 
nt 
0 
can 
lls. 
cen 
xted 
raphic 
ween 
rder 
ce to 
| in the 
  
model. Distances related each predictor map were within the range of the influencing distance for each map 
pattern as described in table 1. For a dialated lineament with 5 distance classes, the association has been 
calculated and obtained values for C are given in table 2. The weights and contrast ,C, for the other spatial 
features such as distance to valley, distance to tank, et. have also been calculated. For geological units, only 
positive weights were calculated as these units are mutually exclusive and the Table 3 shows calculated 
weights for different rock types. 
  
  
Distance 10 W W C Rock type W 
lineament (m) - 
  
  
  
  
50 263381 0.490, 2.825 Charnockite me 
100 1,123 0550-42 293 Granite or | 
150 1,383 0,510 94,893 Granitic gneiss 0.319 
20 1.985 0,483 1.548 Hornblende 
250 Q. O66 0. 534. 1.520 biotite gneiss —0, 034 
Table 2: Weights for lineament Table 3: Weights for rock types 
After obtaining the weights and contrast, C, for each predictor maps, the following equation has been used 
to integrate binary predictor maps in order to optimize the probability of obtaining high yielding wells. 
O post = exp {In (O prior y+ Tw } 
j=l 
Where, O odds, either prior or posterior, and is related to the probability P by O = P/(1-P). 
"d W for map pattern j present 
) W for map pattern j absent 
Finally a map of posterior probability indicating potential groundwater zones is created by calculating the 
posterior probability P ,. as seen in figure 7. 
  
a a 
y . u... x. 
aN aa NaN 
Cait em etn hn LM aS 
* at t >. 
+e "ran at 
«-O low potential 
O1 O10 low to moderate 
0:10 03 moderate 
0-3 0-6 moderate to high 
>0'6 high 
Valleys: considered as high 
Well 0 lkm 
rre 
  
  
  
Fgure 7: Posterior probability of groundwater occurences 
363 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996