5 
  
| the 
2 30 
.059 
and 
area 
  
25 4 
20 4 ° 
Drainage area (DEM) 
© 
  
0 + 3 9 + + = 
0 0.5 1 1.5 2 2.5 3 
Drainage area (aerial photographs) 
  
  
  
Figure 4. A plot of the drainage basin areas (ha) derived from the 
two methods. n=30, r=-0.059. 
  
25 4 
20 4 
Frequency 
  
  
-18. -15  -12 -9 -6 -3 0 3 6 9 12 15 18 
Differences in drainage area (aerial photographs-DEM) 
  
  
  
Figure 5. The differences in drainage basin area (ha) derived from 
the two methods. mean=-0.627, n=30. 
The influence of the method on the estimation of the 
areas of the ponds was analyzed using a Student's t- 
test. The paired sample t-test gave a t-value of -0.72, 
which resulted in that the Hp hypothesis (Haerial 
photographs = HDEM) cannot be rejected at the 99% 
confidence level. 
4. DISCUSSION AND CONCLUSIONS 
4.1 Comparison of Wetland and Drainage Areas 
The results of the t-test used to compare the wetland 
areas derived from both automatic and manual 
(Wessling, 1991) methods indicates significant 
differences. These may be caused by the following: 
- the automatic methods are properly calibrated while 
the manual methods provide unrealistic results. 
- the automatic methods are poorly calibrated while 
the manual methods provide more realistic results. 
- neither automatic nor manual methods provide 
realistic results. 
821 
There is, at present, no field data that can be used to 
evaluate the results generated using the two methods; 
it is thus difficult to quantitatively determine which 
method provides the most realistic results. It should, 
however, be stressed that the automatic methods used 
in this study have eliminated much source of error and, 
therefore, would logically seem to provide more 
accurate and detailed results. 
Possible sources of error that may have influenced the 
results of the analysis of pond and catchment areas, 
and, consequently, the results of the t-tests as well, are: 
- The pond locations determined using the manual 
methods were originally located on aerial photos 
which were not geometrically corrected to existing 
maps. These locations were then transferred 
manually to topographical maps. Some of the pond 
locations may have been transferred with their 
outflow points resultingly not being located in a 
stream or valley bottom. 
- Since the pond coordinates used with the automatic 
methods consisted of digitalized outflow points from 
the above mentioned maps, a number of these may 
have been improperly located when imported to the 
DEM. An error of only one pixel from a stream or 
valley bottom can significantly alter the pond and 
drainage area characteristics calculated from the 
given outflow pixel. 
- The DEM and above mentioned maps may not have 
been geometrically compatible, resulting in outflow 
pixels being improperly located. The potential of this 
influencing the results of the automatic analysis is 
seen as being minimal. 
The results of the t-test used to compare the ponds 
drainage areas indicate no significant differences 
between results derived from automatic and manual 
methods. Since drainage areas can be delineated fairly 
confidently using manual methods, it can be concluded 
that the automatic methods provide realistic and 
accurate results, provided the pixel on which all 
computations are based is properly located.