of a 
data, | INPUT DATA i 
rical : 
  
  
  
  
  
  
  
  
   
     
  
  
  
  
  
  
  
  
  
  
      
  
   
  
  
     
    
   
  
  
   
    
  
  
  
  
   
   
  
  
  
     
  
    
  
  
   
  
  
  
  
  
  
  
  
  
  
  
  
   
    
   
  
     
    
    
  
    
    
  
   
  
  
   
  
  
  
  
   
    
  
       
  
     
    
  
  
  
  
  
  
  
  
  
o: to 
(with Y 
‚if L.S. INTERPOLATION » E 
the 
the 
wise, | SEARCH FOR 
the | OPTIMUM SPACING 
avoid | 
pe FALSE “BiG OUTLIERS RUE | REJECTION 
| M PRESENT OF OUTLIERS 
| EMPIRICAL ESTIMATION 
| OF COVARIANCE AND 
| CROSS-COVERIANCE FUNCTION 
k and 
are FALSE NON-STATIONARY TRUE 
FINITE ELEMENTS COVARIANCE 
d for INTERPOLATION 
n the 
which | 1st STEP ONLY | SEARCH FOR BEST 
erage MODEL FUNCTION 
to 10 
with TRUE COMPUTATIONAL AFTER 1st STEP 
tance | PROBLEMS IN THE 
| FILTERING i 
the | V Y 
f the | FALSE IRRELEVANT 
e and | CROSS-CORRELATIONS 
hould 
quite | FALSE CROSS-CORRELATIONS ^. TRUE FILTERING THE 
ation | TOO HIGH COMPONENTS SEPARATELY 
n the 
on a 
ay to | FILTERING ALL THE 
ments | CONPONENTS TOGETHER 
lent ly | 1 
om a 
duced | FALSE Let SMALL OUTLIERS TRUE 
sides | PRESENT 
rical 
been | 
their x 
SEARCH FOR OUTLIERS < TRUE REJECTION 
5,80. | BREAK LINES RANDOM LOCATED OF OUTLIERS }— 
ation | 
nable | 
mall, | 7 
1 has | START NEW STEP ^ COLOURED 
P OF COLLOCATION RESIDUAL NOISE 
,ional | 
amples 
; look | PREDICTION ON 
hat a | CHEK POINTS AND/OR 
rable | A REGULAR GRID 
ments | 
spline 
w the | 
= than | OUTPUT DATA 
ied by 
ss the 
noîse 
of the Figure 3.1- Flow chart of the system of programs 
or the 
| 
385