.odel. 
ming 
and 
Yes, a 
kage, 
ting 
ind a 
| was 
was 
onds 
reas. 
sins) 
e of 
ond. 
nage 
The 
one 
each 
age 
age 
ling 
dow 
ass, 
(3), 
nber 
0 is 
don. 
xels 
urs, 
the 
ing 
vest 
| on 
The 
sin 
ited 
ally 
ded 
A: If the centre pixel of the three-by-three pixel window 
is a sink and the elevations of the surrounding pixels 
results in a defined aspect. 
B: If the centre pixel of the three-by-three pixel window 
is assigned an aspect value that gives a drainage 
direction to a pixel with a higher elevation. 
In the first case, the problem was solved by giving all 
single sinks the drainage direction code 0. The second 
case was solved by demanding that the drainage 
direction always point towards a pixel with an 
elevation less than or equal to the centre pixel's. 
The second stage of the definitions of drainage 
directions, after each pixel had been assigned a 
drainage direction code (0-8), was to solve the problem 
of drainage directions for flat regions. Adjacent 
drainage directions, both upstream and downstream 
from flat regions, were used as a basis for determining 
drainage directions for each flat region exceeding one 
pixel in size; these were assigned drainage direction 
codes 1-8. To solve the problem of assigning correct 
drainage direction codes to these pixels, every flat area 
consisting of more than one pixel was examined a 
second time. The position (row and column) of each 0 
coded pixel in an area exceeding one pixel in size was 
stored during the execution of the program. Each of 
these pixels was then assigned the mean drainage 
direction of its neighbours with codes other than 0. 
First, all pixels with seven neighbours were assigned a 
new direction code, then all pixels with six neighbours 
were assigned a new direction code, and so on, until all 
the pixels had been assigned new drainage direction 
values. 
2.2 Definition of Drainage Areas 
After calculating drainage directions, automated 
drainage detection was performed. The coordinates (x, 
y) of the outflow of each potential wetland were 
imported to the program, and the ponds and their 
drainage basins were delineated. 
The program starts by constructing an 'imaginary 
barrier’, with the height of 1.0 metre, at the outflow 
(pixel) of each potential wetland. All pixels that supply 
water (drainage direction) to this pixel, and that have 
an elevation less than or equal to the outflow pixel plus 
1.0 metre, form the pond. The area and volume of each 
819 
pond was calculated according to general statistics. 
Independently of elevation, all pixels that supply water 
(drainage direction) to the outflow pixel form the 
drainage basin to the pond. 
2.3 Comparison of Wetland Areas 
The results of the calculation of the ponds and their 
drainage basins were compared with the results 
determined by Wessling (1991), where areas of the 
same 30 wetlands were determined manually using 
aerial photographs. Assuming normally distributed 
data, a paired sample t-test (Williams, 1984), not 
assuming the same population variances, can be carried 
out. This was done using the area values calculated by 
the different methods (DEM and aerial photographs) in 
order to test the hypothesis that the differences in areas 
represent a sample drawn from a population of 
normally distributed differences whose mean is zero. 
The hypothesis was: 
Ho: Vaerial photographs 7 HDEM 
versus 
HA: Haerial photographs # HDEM 
This test was performed both for the pond areas and 
the areas of their drainage basins. Additionally, a 
simple correlaton analysis between the results derived 
from the different methods was performed. 
3. RESULTS 
The 30 ponds and their drainage basins were identified 
and the areas were calculated (Section 2). A section of 
the DEM, the ponds located within it along with their 
drainage basins are presented as a three-dimensional 
plot in Figure 1. It should be noted that a large portion 
of the wetlands are rather small and thus only 
represented by a few pixels. 
3.1 Comparisons of Wetland Areas 
The correlation coefficient between the 30 pond areas 
derived by the two methods (DEM and aerial 
photographs) was 0.263, with a 95% confidence interval 
between -0.108 and 0.569. A plot of the areas derived