78 
PREDICTION AND SIMULATIONS OF MALAYSIAN FOREST FIRES BY MEANS OF 
RANDOM SPREAD 
Jean Serra 1 , Mohd Dini Hairi Suliman 2 , and Mastura Mahmud 2 
1 Laboratoire A2SI, Groupe ESIEE, BP 99, 93162 Noisy-le-Grand Cedex France. Email: serraj@esiee.fr 
2 Earth Observation Center, Universiti Kebangsaan, 3600 Bangi, Selangor Malaysia 
KEY WORDS: Selangor, risks maps, random sets, forest fires, mathematical morphology, random spread 
ABSTRACT: 
annihilatio 
modynami 
analyses ai 
ically, the 
approach 
factors ini 
practical i 
’his practi 
Stic physic 
model of 
ies can be 
[ay-night 
The paper proposes a stochastic model to study forest fires from risk maps and to predict the location of the scar zones. This model 
holds on both the process (the fire front) and on the result (the scars)of a fire. A new set operator, called ’’random spread” /3, describes 
the front of the fire at each daily step. Under iteration, operator /3 provides the time evolution of the process. Average sizes, and the 
probabilities of extinction at each step are derived. The random spread model is applied to analyse the fires that occurred in the State of 
Selangor (Malaysia) from 2001 to 2004. It proves to able to predict all places where burnt scars actually occurred, which is a strongly 
significant verification. Finally, the model is used for simulating forest fires. 
1 INTRODUCTION 
In a large number of wild forests, such as typically in South East 
Asia, forest fires propagate less under the action of the wind, as 
in Mediterrenan countries, than under almost isotropic irregular 
causes. For instance, the fires that occurred in the state of Selan 
gor (Malaysia, see Figure 1) from 2000 to 2004 seem to evolve 
randomly: the initial seats vanish sometimes, but may also give 
birth to new seats at some distance. 
Figure 1: Malaysia and State of Selangor. 
The foresters use to describe the fire progress by means two key 
maps (Hairi Suliman et al.2005), namely the daily spread rate 
of the fire, i.e. the radius r of the daily circular propagation of 
the fire, and the fuel consumption f w , (a weighted version of 
the vegetation map). Both are depicted in Figure 2. Clearly, a 
straightforward use of such key maps does lead to a pertinent 
description. By starting from any point seat, one always arrives 
to bum the whole country in a finite time, under iteration, as both 
maps are positive. Indeed, one must use the rate information in 
some restrictive way, to be able to reach actual events. It is the 
purpose of the present paper. 
Figure 2: a)map f w of the fuel consumption; b)map of the spread 
rate, i.e.of the radius r of the daily circular propagation of the 
fire. 
1.1 State of the art 
The literature about quantitative modelling of Malaysian type for 
est fires, can be split into two groups. Firstly, the behaviour of a 
forest fire is often described from what is actually observed (Car- 
rega2002), (Wan Ahmad2002). The research centres of forestry, 
in the countries where such fires occur, have analyzed the situa 
tion over long periods of time. They define series of pertinent pa 
rameters, coming either from permanent sources (fuel maps, soil 
nature) or from variable ones (typically weather observations: 
temperature, relative humidity, wind, rain) (Forestry Canada2003). 
Then they propose complex systems for managing this informa 
tion, which result in behaviour predictions of the rate of spread, 
of the fuel consumption, of the temperature at the front of the fire, 
etc. (Colin et al.2002), (Turner et al. 1985), (Zhang et al.2003). 
Their predictions yield daily maps which are often synthesized in 
an overall risk map (Blanchi et al.2002). 
Such descriptions cooperate with satellite data, not only for the 
land cover classification they need, but also because their maps 
have to be read in association with those of the hot spots which 
are detected daily by satellite. 
From a second point of view, the heart of the description lies in 
the equations that underlay a fire. Several phenomena are in 
volved, such as heat transfer (Catchpole et al. 2002), gas proper 
ties (Catchpole et al. 2002), turbulence phenomena (Bak et al. 1990), 
Finally, a 
BEHAVE 
simulations 
we make 
maps of the 
1.2 Fores 
The major 
event of 19 
environmen 
veloped 
to the fo 
lit by fa 
and the 
magnified 
(ADB1999) 
Peninsular Î 
of natural fo 
(Gantz2002 
Most of the 
spells from 
Forest fires 
erally lov 
swamp fc 
cidences 
1991 as a 
ing due t( 
into the 
(Abdullah 
ported froi 
et a/. 2003) 
burning 
of local 
vegetation 
Amongst 
that are 
the slash 
condition 
bated the 
ous contri 
Kalimant 
2.1 Def 
Interpret