ESTIMATE OF INDICE OF RISK OF FORESTS FIRES FROM A DATA MODEL OF THE SIG APPLIED TO A COASTAL FOREST OF BAINEM (ALGERIA)

A.Belhadj-Aissa , M. Belhadj-Aissa, A.Belhadj-Aissa, Y. Smara, B Sansal

Image Processing and Radiation Laboratory, Faculty of Electronic and computer science,
University of sciences and technology Houari Boumediene, Algiers (DZ)

I. Introduction
In recent years, Remote Sensing and Geographic Information System (GIS) technologies have been the object of considerable interest to all bodies concerned with space and in particular emergency services and disaster management in Algeria, in collaboration with all other bodies responsible for prevention and management of all major risks in Algeria. Among these risks we can mention the most preoccupying as being fires. Every year, about 30 000 hectares of forests are destroyed by fires in Algeria, fires also take a heavy toll in lives and property and our major preoccupation in to reduce forests fires in the country, thanks to inventory, prevention and management and follow-up. In the last few years, the data acquired by earth observation satellites associated with different types of linked information integrated in a Geographic Information System has proved to constitute a viable option for the follow -up of disasters, the identification of areas at risk, and the mapping out of the extend of fire. The goal of our work is to set up of a GIS integrating remote sensing data for the prevention and management of the fires. The methodology adopted is as follows:
a) Process and structure data, to determine and to establish indicators parameters of the forests fires.
b) Present to decision makers of the sectors concerned, in real time on screen:
· Maps of the state of the surface at any moment in time "t" (maps of plant surface cover, maps of the access roads, maps showing altitude).
· Maps of zoning indicators integrated with meteorological data.
The principal factor of degradation of the Algerian forest is the fire which finds an environment physical and natural favourable to its blossoming and its propagation. The structure and the component of the vegetable formations are as many factors facilitating forests fires.

II. Area study
The figure 1 presents a satellite image extracted from the Landsat scene (resolution 30 m x 30m) representing the west part of the town of Algiers. The forest of Bainem is delimited by the closed curve, and is situated in surroundings of the urban area of Algiers it is limited to the north by the Mediterranean sea. The surface of this forest after 1998 (fired in summer 1998) is about 800 hectares

figure 1: satellite image extracted from the Landsat scene

III. Developed methodology
For the design of the data model and the implementation of methodology, we propose a multilevel step which we schematised by diagram in figure 2 :
These levels lead us to distinguish various elements from the risk. Then each one of these elements will be modeled by selecting the components and the necessary parameters as well as a mode of representation. The interpretation and the analysis of the various elements starting from the crossing of the various layers of data of the SIG allow the estimate of the indice of the risk, the coding of this indice and the realization of the cartography of the risk fires of Bainem forest. In the global diagram, the definition of scenario and the preparation of plan of forecast and intervention require, in addition to the SIG, the contribution of the institutions concerned with these natural phenomena. This part is not approached in this article.

III.1. Modeling of the risk
To estimate the risk it is necessary to model each component of the risk. This stage consists in selecting the parameters specific to each component (combustibility, slope..) then to use a mode of representation of the risk in order to evaluate it. The parameters are the factors of the natural environment and anthropic which influence the blossoming, the propagation and the intensity of fire. The principal ones are: the vegetation (combustibility, infallibility, biomass), the topography (slope, exposure), the climate (moisture of the air, direction of the winds, pluviometry, temperature), the human activities (occupation of the ground, land, not of water...) and the history (burned surface, not of blossoming). These various parameters are correlated and their combination influences or causes the blossoming of the fire. Thus we retained, for this modelling, the most relevant elements, the vegetable cover whose descriptive parameters intervene in the calculation of the index of combustibility, the relief represented by the topographic parameters which intervene in the calculation of the index topomorphologic and the parameter anthropic.

For our application, considering the absence of a model of calculation of the risk of fire of forest in our country, we used the model of indice of risk developed at the point by A. DAGORNE and al [DAG. 1993], which we adapted to the type of risk in Algeria. Its expression is given by:

IR = 5 IC + IM + 2 IA

where IC, account for the combustibility indice, IM: the topomorphologic indice and IA: the anthropic indice. The characterization of this indices is based on the space variability of the fire hazard from which the determination results from the physical parameters intervening in the model chosen for our application.

Figure 2 : block diagram for the mapping, the prevention and management of forests fires by GIS

III.1.1. Combustibility indice Model (IC)
The principal criteria which intervene in the appearance of a fire in a vegetable space are the space structure of fuel (biomass, horizontal covering and vertical stratification) and the nature of the dominant species. To this end the IC are expressed by the following relation:

IC = 39 + 0.23 BV (E - 7.18)

it is an empirical model where BV is calculated starting from the bio combustible volume, E represents the average combustibility of each type of settlement. For our application we determined these parameters starting from the investigations and of the observations as well as measurements on the site. Then, we supplemented information by spectral analysis of the behaviour of the species starting from the images of the satellites Landsat TM5 and TM7+ of the area

III.1.1. Combustibility indice Model (IC)
The principal criteria which intervene in the appearance of a fire in a vegetable space are the space structure of fuel (biomass, horizontal covering and vertical stratification) and the nature of the dominant species. To this end the IC are expressed by the following relation:

IM = 3p + m*e

it is an empirical model where BV is calculated starting from the bio combustible volume, E represents the average combustibility of each type of settlement. For our application we determined these parameters starting from the investigations and of the observations as well as measurements on the site. Then, we supplemented information by spectral analysis of the behaviour of the species starting from the images of the satellites Landsat TM5 and TM7+ of the area

figure 3: digital terrain model (DTM) of the forest of Bainem

figure 4: slope layer

III.1.3. Anthropic indice Model (IA)
The presence of human beings and habitation near the forests constitutes an important parameter which determines the degree of vulnerability of the medium. Thus the parameter anthropic is the principal term in the model of the IA. Two situations can arise: either the drill is approximity of the infrastructures, or of the inhabited zones are inside the forest. To this end the risk is quantified by an indice of vicinity IV (approximity or included) and a factor of urban density. The indice anthropic is then expressed by the linear combination of the two parameters:

IA = IV + D

Where: IV: indice of vicinity and D: density. We used for our site of study, the panchromatic images of satellite SPOT (10 m) and IRS (5,8 m), the aerial photographs at the 1/40.000 scale and the maps of city of Algiers at 1/7500 scale for the realisation of the layers representing the infrastructures, the urban areas as well as the linear structures. The figure 5 represents the urban layer.

figure 5: urban layer

III.2. Risk Mapping
We proposed a procedure of estimate of the indice of risk which we developed by using software ENVI and MAPINFO environment. It consists on overlapping, in raster mode, a different thematic layers. Result image represents a map of risk indices coded on levels of degree of setting off just as spreading the fire in study area.

IV. Conclusion
Analysis of hazard is a complex task, as many factors can play important role in the occurrence of the disastrous event. Therefore, analysis requires a large number of input parameters, and techniques of analysis may be very costly and time consuming. The increased availability of Remote Sensing data and GIS during recent decades has created opportunities for a more detailed and rapid analysis of natural hazards. The proper structure of information system for disaster management should be present to tackle the disaster and to manage it.
· In this communication, we present our contribution of the Algerian experience in the use of Space technologies for fire forest management. In effect, Remote Sensing and GIS can provide useful information, and create disaster awareness with politicians, concerned decision makers and the public, so that on a national level decisions are taken to set up disaster management organisations.
The remote sensing and GIS database can be used to create elaborate and effective Disaster Management Information System (DMIS). An integrated approach using scientific and technological advances should be adopted to mitigate and to manage natural hazards. Moreover there should be a national policy for natural disaster management.

References