FE N PANE NE
interpretation errors which are naturally attached to image interpretation.
As the equations are strongly depending on evaluation scale,constants were re-
lated which allows the transition of the equations from one to another scale.
The lowest constant errors and the highest reliability of the river basin equa-
tions are delivered in scales from 1:100 000 to 1:500 000 as far as machine pro-
cessed images are concerned.
On 1:1 000 000 scale the errors are increasing rapidly and the characteristic
equations are becoming unreliable.
4. Vegetation and moisture zone analysis(Map 1,2,3)
Based on the assumption that a scale of 1:100 000 is suited for further studies
of large area characteristics the vegetation,moisture and dry zones were analy-
sed and mapped.
While false colour composite is delivering a general impression of the vegeta-
tion distribution a more detailed interpretation has been carried out on a blue
print of band 6 superimposed with an equidensite.
The dry zones are appearing in green blueish colour,while the drainage system
is shining light and the forest and vegetation areas are dark to light violett.
This method is of some advantage in cloudy zones as these are often met at Fast-
African coastal regions and as all minor clouds can be discovered immediately
and separated from vegetation and small agricultural plots.
In case of the analysed zone with a structure of scattered and mixed subsistance
farming the method gives additional security from misinterpretation.
Further experiments have shown that except an enhancement all other ways of
image processing like clustering,histogram analysis or classification either
supervised or unsupervised are of low gain in case of subsistance farming struc-
tures.
This favours the application of inexpensive manual processing and lowers down
the image processing costs.
5. Rainfall,runoff and evaporation(Fig. 5,6,7)
These environmental factors are not determinable directely by image interpreta-
tion. Rainfall intensities are correlated with soils,moisture conditions and
vegetation coverage or biomass. The first can be detected by colour or photo-
graphic density profiling,the latter by superimposing band 5 and 6 and by area
measurements and mapping.
As the East-African coastal zones are characterized by a high erosion of sandy
soils,a change of erosional channels and splash erosion areas are detectable on
successive images which leads to some information of rain storm intensities,
from which the runoff can be calculated.
The relationships are different for each landscape and geomorphological struc-
ture. The parameters determined are valid for the particular area and can serve
as further inputs for monitoring operations.
Evaporation datas are obtainable from terrestrial,aerial and satellite measure-
ments by reflectances especially in the near infrared region together with tem-
perature sensing.
Interception,infiltration and percolation and subsurface runoff are difficult
to determine as they are acting under the vegetation canopy.
A way to approach a solution is to collect large samples of spectral reflectan-
ces from reflection mesurements from various soil and vegetation status by ter-
restrial radiometric measurements and climatic datas.Correlations with satelli-
te image reflectances lead to some results usable within the demand of foreca-
sting and planning.Infrared aerial photography with simple camera arangements
is indispensable. Within a five hour flight in height of 720 m all datas,pho-
tographic infrared and radiometric observations and descriptions from landscape
could be obtained. This method is fully replacing a difficult ground truth mis-
sion in an inaccessible area.
454
o
z
<
9
z
Brio)
Cz
z
9
er
Eu
w
Qu. >
o +
ww
ou x
£2 ©
ee eo
ze £
ER ©
a à
(MÁS
=:
4
o
uu
ad
io
ul
ac
o
zo
2z
e<
ez
0 a0
z< °
HE A
© 4
=
Ez
£3 >
i
a> «
zx ll
25 z
o
ez 5
=< <
Sz ©
gS >
«r- o
-— a
s:
