REAL TIME FOREST FIRE SIMULATION WITH
EXTINGUISHMENT SUPPORT
Anis Korchi, Aitor Moreno and Álvaro Segura
Vicomtech, Mikeletegi Pasealekua 57, 20009 San Sebastian, Spain
Keywords: Forest Fire, Fire Spreading Simulation, Real Time Algorithm, Environment Process Simulation.
Abstract: Fires and other related disasters provoke great destruction of high valuable environments and economical
losses. In this work, we present a forest fire spreading algorithm to be used in real time and interactive
Virtual Simulations. The main objective is to obtain a fast, interactive and quasi-realistic Virtual Simulation
to be used in the simulation of virtual scenarios where fire-fighters and controllers will be trained. The
algorithm supports the main variables involved in the fire spreading (slope and wind) and the radiation
effect makes the forest fires bypass rivers or firewalls. To complete the simulation, a basic model has been
added to extinguish the fire using water.
1 INTRODUCTION
Fire is one of the most complex and destructive
phenomena in Nature. When they are out of control,
they can devastate large extensions of forest area or
burn buildings provoking economical losses,
environmental impacts and even human casualties
(see Figure 1).
Figure 1: Burnt Areas in ha in Southern Europe between
1980 and 2006. (Forest Fires Website, 2010) (Schmuck,
2006).
The preventive measures are very important, but
eventually, the fire will start. Whether in urban or in
forested areas, we can stop or limit a fire by having a
skilled and experienced Fire-Force, from the firemen
to the management staff who will organize the
available resources to fight the fire.
Emergency Training Centres are a key factor to
fight against fire. One of the most suitable solution
solutions to train the fire-fighters and manager is the
use of Virtual Simulations, where they can safely
collaborate to fight a virtual fire and check if the
concepts learnt in the Training Centre have been
applied correctly.
One of the main algorithmic elements in such
Virtual Simulations is to simulate how fire spreads
as simulation time advances. Although a
simplification of the algorithms is needed in order to
achieve interactive rates, it is desirable that it could
deal with different types of forest fires, react to
different types of terrain, slopes and changing
weather conditions. The simulation should support
the capability of fire to be extinguished by itself
(fuel combustion) or by the fire-fighters (fire
extinguish agent).
In this work, we present a fire spreading
algorithm that can be implemented to be used in real
time and interactive Virtual Simulations. The main
objective is to obtain a fast, interactive and quasi-
realistic system to be used in the simulation of
virtual scenarios where fire-fighters and controllers
will be trained.
2 STATE OF THE ART
There are two major models of fire simulation in
wild land: empirical models and physical models.
The empirical models are made thanks to the
experience of real fire, i.e., those models use
323
Korchi A., Moreno A. and Segura Á. (2010).
REAL TIME FOREST FIRE SIMULATION WITH EXTINGUISHMENT SUPPORT.
In Proceedings of the International Conference on Computer Graphics Theory and Applications, pages 323-326
DOI: 10.5220/0002890703230326
Copyright
c
 SciTePress

statistical relationships found between the fire
evolution and different parameter tested on the field
(Rothermal, 1972). In this case, we can mention
FARSITE (Finney, 1998), which use Huygens
principle of wave propagation.
The second type, physical-based models, use
convection and heat transfer mechanism, but also
computational fluid dynamics methods. The main
mathematical tools used here are partial differential
equations and reaction diffusion systems. Fire
Dynamic Simulator (FDS, National Institute of
Standards and Technology (NIST)) or FIRETEC
(Linn, 2002) follow this approach.
Unlike the two previous models, other research
works have taken a different direction from the
complex mathematical models. Their objective is to
reduce computation time and to implement a real
time simulation.
Gary L. Achtemeier (Achtemeier, 2003)
presented the Rabbit Model, a collection of basic
rules of fire evolution, which are implemented as
autonomous agents (the rabbits). The scope of the
Rabbit Model is limited to the evolution of forest
fire.
The proposed algorithm in this work presents a
forest fire spreading simulation, whose main
characteristics are:
 The fire evolution is based on the terrain
topology, material and weather conditions.
 The fire is allowed to cross rivers, firewall or
other barriers by introducing the radiation effect.
 Very low complexity, allowing real time
simulation even with standard computing power.
In the next sections, the details of the proposed
method will be explained, showing some basic
results of the implementation of the algorithm.
3 PROPOSED METHOD
The proposed algorithm main goal is to simulate
how the fire spreads in a forest environment under
different circumstances. In order to explain the
method, the field where the fire will be must be
defined.
3.1 Introduction and Field Definition
The forest fire simulation method is based on a
divided field in autonomous cells. The field is
regularly divided in a grid of square cells and is
defined by its geometrical information (origin, size
and height map).
Every square cell contains all the necessary
information to support the fire spreading, being
divided in two main categories: i) passive
information, which will be used to initialize the field
and ii) runtime information, which will be modified
or calculated during the simulation.
The passive information of each cell corresponds
with the geometrical data (position, size and height
of the cell in the field) and the characterization of
the cell, i.e., type of cell, which will determine the
nature of the cell (dry grass, tall trees…), and the
starting quantity of combustible.
The fire-spreading algorithm is a step-by-step
process, i.e., every time that the algorithm is run; a
new state of the fire evolution is calculated.
Thus, the runtime information will be calculated
during each simulation step, and includes: the
quantity of combustible, the power or intensity of
the fire and the state of the fire.
3.2 Main Considered Physical
Variables
The objective is to reduce the algorithmic
complexity and the processing time while keeping a
correct behaviour of the spreading of the fire. To
reach this goal, among the whole set of variables and
physical parameters which can influence the fire
behaviour, we have chosen the most significant
ones.
Figure 2: The different states of a cell and the available
transitions.
The most relevant external parameters are the
wind and the slope (Weise, 1996). Other variables,
although necessary for precise simulations or
predictions have been discarded.
3.3 The States of the Fire Evolution
All the cells in the field have an internal state which
corresponds with the state of the fire that exists in
such cell. The different states are Safe, Activated,
Burnt, Survive, RiverCrossing and FireStopped (See
Figure 2).
GRAPP 2010 - International Conference on Computer Graphics Theory and Applications
324

When a cell is in the Safe state, there is no fire in
it. The Activated state indicates that there is an
active fire in the cell.
The Burnt state is the final state of a cell after
being completely consumed by the fire. The
Survived state is pseudo-final state, when the cell is
burnt, but still has residual heat. Even being
completely consumed by the fire, the cell can
irradiate some heat to others cells. Eventually, the
cell will pass to the Burnt state.
The RiverCrossing state is a hidden state used in
the computation of the fire spreading through a river
or firewall, using the radiation effect.
Finally, the FireStopped state represents an
intermediate state of a cell, when some external fire
extinguisher has stopped the fire.
3.4 Fire Spreading Algorithm
The algorithm that defines how the fire spreads in
wild land is based on the cells’ initial data of the
whole field (combustible, states ...).
Following a similar approach to Gary L.
Achtemeier in the “Rabbit Model” (Achtemeier,
2003), we define the fire spread as a displacement of
mice. The combustible on the field is considered as
the mice’s food (cheese). Thus, the basic concept is
that when a mouse eats a cheese, it is equivalent to
when a combustible unit is burnt by the fire.
A fire is started by positioning a mouse in a
square cell, changing its corresponding internal state
to Activated. The added mouse will interact with the
neighbour cells and will follow some rules.
Every mouse is born in a square with a given
power to eat a quantity of cheese per simulation step
(fire intensity), which changes depending on the
type of the cell. When a mouse eats a quantity of
cheese, its power is increased in the same amount,
which simulates how the fire intensity is
continuously growing while there is available
combustible.
If there is no cheese remaining, the mouse dies.
When a mouse dies, up to 8 new mice can be born,
since there are 8 potential neighbour squares. The
algorithm uses the local slope and wind in order to
determine which neighbours will be the destiny of
the new born mice.
The mouse will give birth to all the squares
which are in an area of +/- 45° of the wind angle. If
there is no wind (or it is too slow), the mouse will
give birth to other mice in all the 8 neighbours.
Also, the mouse will always try to give birth to
other mice in squares that are in higher altitude. This
behaviour tries to simulate the fact that the fire goes
up if there is a slope. Only if there is not a non-
burned square at a higher altitude, the mouse can go
down (it has no option to go up).
Every square cell has a parameter named Fire
Power, which is in fact the sum of the mice’s power
of the cell. Similarly, the Fire Power of a given cell
will be stronger or more intense with a higher
number of mice.
3.5 Radiation Effect
The radiation effect is added to support the fact that
a fire can cross a river, firewall or other barriers like
roads. It is based on the heat radiation.
In the proposed algorithm, a fire is stopped when
it encounters a barrier. Depending on the Fire Power
in the area, and considering the width and type
(streets, road, river [...]) of the barrier, it can be
overridden.
Figure 3: The Fire has bypassed successfully a river. The
sticks in the water represent the Fire Power and it can be
seen that it is being reduced gradually. In the presented
case, the fire reached the opposite bank, spreading it.
The fire is able to bypass the river only if the
existing vegetation is composed of trees or any other
high vegetation. The main idea is the difference of
size: the more a tree is high, the more the wind can
spread heat/radiation to the other side of the barrier
(see Figure 3).
As the fire is stopped when the power is zero or
negative, there are not too many possibilities to
reach the other side of the barrier if it is wide
enough.
4 FIRE EXTINGUISHER MODEL
The proposed method allows trying to extinguish an
ongoing forest fire. Its main purpose is to allow the
fire-fighters to interact with the fire evolution and
see the consequences of their decisions on the field.
REAL TIME FOREST FIRE SIMULATION WITH EXTINGUISHMENT SUPPORT
325

This model is conceptually the antagonist of the
fire spreading algorithm, where the Fire Power
amount is replaced by Fire Extinguisher Agent
amount (e.g., water). When, a given quantity of
water is thrown into a cell, the same quantity of Fire
Power is decreased.
Thus, as long as there is water in a cell, its Fire
Power will be equal to zero and the fire will be
stopped (changing to the state FireStopped).
A cell being in the state FireStopped still has
food (combustible), therefore, the fire in that area
could be restarted due to the surrounding burning
areas.
5 RESULTS
A C++ implementation of the algorithm has been
done to test the algorithm behaviour and outputs.
The tests have been performed using a field of
2.25 hectares, choosing a tile side size equals to 3
meters, which is half the distance of possible spread
of fire when there is no wind (Breton, 2008).
To simplify the tests, the wind is uniform and
constant in the entire field and during all the
simulation.
The first test is to run the fire simulator without
any external factor which can influence the fire
spread (Wind, Slope). The fire evolves in a circular
way, as expected (see Figure 4, top-left).
To test the wind effect, another test is performed
with a uniform wind blowing from North-West to
South-East.
The slope tests have been divided in two, one to
test how the fire can evolve towards higher positions
(see Figure 4, top-right) and to test how the fire
avoids go towards lower positions (see Figure 4,
bottom-left and right).
6 CONCLUSIONS AND FUTURE
WORK
In this paper, a pedagogical real-time fire simulation
algorithm has been presented. Its main purpose is to
be integrated into interactive Virtual Simulations
where fire-fighter and managers can train their
skills.
Although the forest fire spreading is a very
complex phenomenon, we tried to simulate the most
common characteristics of its behaviour by
simplifying the model. Only two main physical
variables have been used in the algorithm: direction
of the wind, and the slope.
For the future work, a comprehensive
comparison between different algorithms should be
done. These tests will try to measure analytically the
efficiency and other related factors. Also, a graphical
comparison of the output should be done, since the
ultimate objective is to be integrated in a VR
environment. In this case, user tests will be done in
order to validate the algorithm output.
Figure 4: Some results of the Fire Spreading algorithm.
Grey level shows the height map of the field.
ACKNOWLEDGEMENTS
This work was carried out in the context of project
SIGEM, funded by the Spanish Industry Ministry
through its Avanza I+D Programme.
REFERENCES
Achtemeier, G. L., 2003. An Application of Stephen
Wolfram’s “New Kind of Science” to Fire Spread
Modeling, USDA Forest Service, Athens, GA.
Breton, T., Duthen, Y., 2008. Les simulations de
propagation de feu en milieu urbain.
Finney, M. A., 1998. FARSITE: Fire area simulator-
model development and evaluation.
Forest Fires Website, 2010. Statistics on Forest Fires,
http://library.thinkquest.org/07aug/01254/statistics.ht
ml . Last time visited: 15/01/2010.
Linn, R., Reisner, J., Colman, J. J., Winterkamp, J., 2002.
Studying wildfire behavior using FIRETEC.
Rocky Mountain website, 2010. Aircraft Rescue and Fire
Fighting Basic Training, http://www.rmestc.com/.
Last time visited: 15/01/2010.
Schmuck, G., 2006. Forest Fires in Europe, Report No7 /
2006, European Commission, Joint Research Centre
Institute for Environment and Sustainability.
Weise, D. R., Biging, G. S., 1996. Effects of Wind velocity
and slope on flame properties.
GRAPP 2010 - International Conference on Computer Graphics Theory and Applications
326