654 - 
Useful quantitative information was nevertheless obtained. The instrument quickly 
indicated that the surface area of the wet, intermediate, and dry zones amounts 
respectively to 39, 33, and 27% of the total area; that the % area sinkhole in the 
three zones is almost equal and amounts to 26. 6, 26. 5, and 24. 2%. The full- 
feature count indicated that the number of sinkholes in the three zones are 52,99 
and 29 respectively. It thus appears that the sinkholes in the intermediate zone 
are smaller and in the dry zone larger than those in the wet zone. This conclusion 
is confirmed by statistical analysis. A Kruskall and Wallis-test has been applied 
in order to test the hypothesis that at least the average size of the sinkholes of 
one of the three measured areas differs significantly from the others. The null- 
hypothesis - there is no difference in average size of sinkholes in the three tested 
areas - can be rejected if the statistic H èX ^^(df = 2). Execution of the test 
showed that the null hypothesis can indeed be rejected for a very low level of Π. 
Since the full-feature count relates to sinkhole complexes rather than to individual 
sinkholes the most likely explanation for the differences in average sinkhole size 
between the wet and the intermediate zone is that the degree of coalescence in the 
wet zone is higher than in the intermediate zone. The small number of sinkholes 
observed in the dry zone may have been affected by their difficult identification 
that has led the interpreter to a generalization resulting in an omission or a com 
bination of certain sinkholes. ' 4 . 
It was subsequently attempted to discover whether the sinkholes show a tendency 
to elongation by measuring the number of intercepts in two perpendicular direc 
tions. For statistical reasons it would have been better to carry out these mea 
surements in all directions. However, for reasons of convenience and because it 
was felt that elongation - if any - would be parallel to the topographic contours, 
the number of intercepts was only determined in the direction of and perpendicular 
to the lower left-top right diagonal. The results obtained are given in fig. 1 that 
relates the counted number of intercepts to the interval in picture points (pp) of 
125 fil and thus gives the frequency distribution of sinkholes of various sizes parallel 
and perpendicular to the topographic contours. The graphs of fig. 1 clearly indicate 
that parallel to the topographic contours the number of small sinkholes ( < 10 pp) 
is considerably less than perpendicular to the contours, whereas a few large-sized 
(coalescent) sinkholes measuring 60/70 to 110 pp occur parallel to the topographic 
contours that are absent in the perpendicular direction.