Skip to main content

A GIS decision support system for the prevention of desertification resulting from forest fire


To develop a knowledge based decision support system which brings
to bear ground and spaceborne multisensor data, together with
expert knowledge about desertification processes, on the problem
of desertification risk assessment after forest fires, with
particular relevance to the Mediterranean countries of the

The basic approch is to develop a system using a modern GIS shell
for the integration of various moduli. Input to the system will
be the major parameters that govern the natural forest
regeneration, and soil erosion, remotely sensed data and field
observations. The output of the system will consist of a set of
inferences and maps showing the degree of risk of
desertification associated with each area.

A rule base will be incorporated into the system, constructed
from modelling the physical processes that contribute to soil
This rule database used with the ground and remotely sensed data
will give the probabilities for soil erosion and
desertificaion, and of natural regeneration. The ground data
used will refer to 1 : 50 000 topographic maps and will include
vegetation cover, soil, erosion and land use data. Further, data
from existing GIS will include topographical, geological and
elevation data. A database will be constructed to incorporate the
above mentioned data .

The relevance of the remotely sensed data (ERS--1 SAR, SPOT
panchromatic, Landsat Thematic Mapper)
will be studied by investigating the statistical correlations
between reflectance and backscattering values with specific
states of soil erosion and natural regeneration. A library of
functions will be build for
image segmentation and relevant feature extraction. Segmentation
techniques combining edge and region information will be
investigated. Multiband texture analysis will be used . The
invocation of the appropriate algorithms will be knowledge
controlled, using information about the time of image
acquisition. The image classification
will be done using Bayesian decision theory. Two approaches will
be investigated: either combining all sources of information at
the data level, or interpreting the data from each source first
and combining them at the symbolic level afterwards.

Finally, the performance of the system will be evaluated by
comparing its output risk maps with conventional maps, to assess
positional and labelling accuracy as well as cost effectiveness.

Funding Scheme

CSC - Cost-sharing contracts


Terma Alkmanos Street
11528 Athens

Participants (2)

Wastiangasse 6
8010 Graz
National Technical University of Athens
9,Iroon Polytechniou Avenue
15780 Athens