Evaluation
Excellent
Good
Possible
Unsuitable
Urban
Water
Impossible (Non-soil)
Impossible (Stcep)
Figure 4. The assessment map for grassland establishment. (*Color Printed)
CASE STUDY 2 : Evaluation of Land
Degradation in Northeastern Syria
In the Abdal Aziz region in northeastern Syria, animal
grazing on native pasture has occurred for a long time.
Grazing has been the typical land-use system in this
region; however, land-use is changing due to an
expansion of cultivation and forestation recently. These
human activities have caused difficulties affecting
environmental changes. To maintain Sustainable
agricultural activity in dry regions, regional resource
management in consideration of environmental
conservation is necessary.
This study aims to evaluate hazards of land degradation,
using NN and a map database of the Abdal Aziz region.
(1) Outline of Evaluation
Figure 5 shows a flow chart for production of a hazard
map of land degradation. The operation was divided to two
parts; first, to construct NN using NEURO92, and
Second, to draft a hazard map of land degradation using
the constructed network. The supervisor for NEURO92
was made as follows.
First, 214 points data were extracted at randomfromthe
map database. From these point data, elevation, slope,
direction of aspect, soil category and vegetation
coverage were input to the input layer of thes upervisor.
Output data of the supervisor were degree of degradation
and extent scale of degradation. They were input s ingly.
Thus, two supervisor files were prepared, having the
same input layer and different output layer.
Using the two supervisors, NN were constructed with
NEUROS2. To s elect the most suitable network, 15trials
with different numbers of hidden layer units were
performed. The learning cycle was repeated 10,000
times.
After the s election of networks, the output of the whole
points in map database were calculated using the network
parameters with GIS. The outputs were mapped to
express the degree and extent scale of land degradation.
These two outputs were integrated by multiplication to
produce the final hazard map of land degradation.
Factors = Input
Elevation
Slope
Direction of aspect
Soil category
Vegetation coverage
br
Results = Output
Degree of land degradation
Extent of land degradation
Y
Data Extraction for Supervisor (N=214 points)
Learning by NEURO92
=
NN for Degree NN for Extent
Map Calculation Map Calculation
(Estimation Map for Degree ) X ( Estimation Map for Extent )
S
f ERE,
f : e
e Hazard Map of land Degradation P
PIERII 7
Figure 5. The flow to draft the hazard map
of land degradation.
791
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996