was about 3. At epoch 1 the comparable variance factor 
was 1. The results listed in Table 1 are those after the a 
priori standard deviations of the observations of the aerial 
photographs have been increased globally so that the 
resulting variance factor was 1. 
Analysis of plots of residuals on the plate measurements 
for the aerial photography for epoch 2 have shown that 
each plate has a systematic array of residuals but that 
between plates and over the block as a whole the residuals 
are random. The most probable cause of this is effects of 
film unflatness with the cut film stuck to the glass plates. 
Therefore it is apparent that if suitable glass plates could 
be found for the UMK that a high precision could be 
obtained from the aerial photography alone. It is 
fortuitous, therefore, that the decision to take back-up 
photography from the ground was taken as this has 
enabled the original specification to be achieved. 
4. Analysis of deformation 
The next stage in the data processing was to input the GAP 
outputs from epochs 1 and 2 into a program called ANDEF 
(ANalysis of DEFormation). The data from GAP that 
ANDEF uses are the coordinates of both sets of target 
points (x, and x,), the cofactor matrices of these two sets 
(Q, and Q,), the degrees of freedom for each epoch (r, and 
r,) and the estimated variance factors 0,” and 0,7). 
ANDEF checks that the same datum of inner constraints 
has been used in both epochs and also tests that both 
results have the same variance factor ( 5% significance 
level). A common variance factor is then computed from 
Figure 6 300mm UMK photograph taken from high level close in position - epoch1 
(Note obstructions owing to construction plant) 
  
   
   
   
  
   
  
  
  
   
  
   
  
   
  
   
  
   
   
  
  
  
   
  
   
   
  
     
2 2 
o 
Ir +1, 
The matrix of displacement vectors is calculated from :- 
d=x,-x, 
The cofactor matrix for these displacements is :- 
Qu= Qi +Q, 
A global congruency test is then carried out using :- 
Q =d'Q,;* d 
where Q,* is the pseudoinverse of the singular cofactor 
matrix Q, which, in this case, is rank deficient by 7. 
These are then used in a series of congruency tests to 
isolate a set of targets which, at a specific confidence level, 
can be regarded as stable. The displacements of the other 
targets relative to this sub-set are derived. Tests based on 
hypotheses about the geotechnical and structural 
properties of the wall should be included in this process; 
it is intended to do this later in this project.