ARTICLE IN PRESS

Journal of Environmental Management ] (]]]]) ]]]]]]

www.elsevier.com/locate/jenvman

A decision support system for managing forest re casualties

Marc Bonazountasa, Despina Kallidromitoub, Pavlos Kassomenosc,, Nikos Passasd
a

National Technical University of Athens, Faculty of Civil Engineering, Department of Water Resources Hydraulics & Maritime Engineering, Iroon
Polytechniou 5, GR-15780 Zografos, Greece
b
Epsilon International SA, Monemvasias 27, GR-15125 Marousi, Greece
c
University of Ioannina, Department of Physics, Laboratory of Meteorology, GR-45110 Ioannina, Greece
d
Technical University of Crete, Department of Environmental Engineering, GR-73100 Chania, Greece
Received 2 July 2005; received in revised form 15 June 2006; accepted 20 June 2006

Abstract
Southern Europe is exposed to anthropogenic and natural forest res. These result in loss of lives, goods and infrastructure, but also
deteriorate the natural environment and degrade ecosystems. The early detection and combating of such catastrophes requires the use of
a decision support system (DSS) for emergency management. The current literature reports on a series of efforts aimed to deliver DSSs
for the management of the forest res by utilising technologies like remote sensing and geographical information systems (GIS), yet no
integrated system exists.
This manuscript presents the results of scientic research aiming to the development of a DSS for managing forest res. The system
provides a series of software tools for the assessment of the propagation and combating of forest res based on Arc/Info, ArcView, Arc
Spatial Analyst, Arc Avenue, and Visual C++ technologies. The system integrates GIS technologies under the same data environment
and utilises a common user interface to produce an integrated computer system based on semi-automatic satellite image processing (fuel
maps), socio-economic risk modelling and probabilistic models that would serve as a useful tool for forest re prevention, planning and
management. Its performance has been demonstrated via real time up-to-date accurate information on the position and evolution of the
re. The system can assist emergency assessment, management and combating of the incident. A site demonstration and validation has
been accomplished for the island of Evoia, Greece, an area particularly vulnerable to forest res due to its ecological characteristics and
prevailing wind patterns.
r 2006 Elsevier Ltd. All rights reserved.
Keywords: Forest re; Decision support system; Southern Europe; GIS

1. Introduction
Every year, wild-land res reduce the forested areas across
southern Europe at a remarkable rate. Despite forest
managers having increased their efforts in forest re ghting,
thousands of hectares of forests are lost. The consequences
are loss of lives and goods, soil erosion, damage to wildlife
habitats and degradation of watersheds. Thus, it is
mandatory for the forest managers to improve their
efciency by investing in forest re ghting technology.
Corresponding author. Fax: +30 26510 98699.

E-mail addresses: bonazoun@central.ntua.gr (M. Bonazountas),

kallidromitou@epsilon.gr (D. Kallidromitou), pkassom@uoi.gr,
pkassom@cc.uoi.gr (P. Kassomenos).

In the recent years, a substantial effort has been made

towards characterising, forecasting, modelling, planning,
and managing forest res in several Mediterranean
countries. In this context, Illera et al. (1996) examined
the possibility of estimating forest re-danger by means of
analysing the temporal evolution of the Normalised
Difference Vegetation Index (NDVI) using National
Oceanic and Atmospheric Administration (NOAA) Advanced Very High-Resolution Radiometer (AVHRR)
satellite images. Gonzales et al. (1997) used data from
AVHRR, NOAA satellites to develop a conceptual risk
assessment analysis for the territory of Valencia, Spain.
Nunez et al. (2000) developed a risk index mapping model
based on forest biomass, chemical, and meteorological
factors in Galicia, Spain. In addition to Europe many areas

0301-4797/$ - see front matter r 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jenvman.2006.06.016
Please cite this article as: Marc Bonazountas, et al., A decision support system for managing forest re casualties, Journal of Environmental
Management (2006), doi:10.1016/j.jenvman.2006.06.016

ARTICLE IN PRESS
2

M. Bonazountas et al. / Journal of Environmental Management ] (]]]]) ]]]]]]

worldwide face the same problem, such as Australia, the

USA, Russia and Canada, which lose thousands of
hectares of forests every year. Thomson et al. (2000)
compared eight forest management plans to determine the
most appropriate for forest re hazard modelling in British
Columbia, Canada.
Greece has experienced intense forest res with an
increasing frequency since the early 1970s. Recently, there
were some efforts aiming at studying forest res in Greece
in the context of re management. Specically, Illiadis et al.
(2002) developed a system aimed at classifying the
prefectures of Greece into forest re risk zones clustering
areas according to similar weights in forest re risk, using
fuzzy techniques, while Kalabokidis et al. (2002) presented
an analytical and operational inventory procedure aimed at
appraising forest re potential.
Forest re management involves fuel management, re
detection and communications and reporting systems, re
weather forecasting, re danger and behaviour indices,
initial attack systems, identication of re sensitive
resource areas, standby systems, training and pre-organisation of co-operators, suppression resources and capability,
and knowledge of re and its ecological implications. The
complexity of re management and the factors involved
require more systematic and sophisticated decision-making
techniques such as those based on operational research and
decision information systems (Vakalis et al., 2004).
Based on the above analysis and in order to ll the gap the
objective of this study is to introduce a decision support
system (DSS) utilising LANDSAT and SPOT imagery. The
DSS aims at providing quantitative estimation of the forest
re consequences, while helping users to evaluate options
and effectively implement decisions. Its architecture involves
an integrated framework of image processing and geographical information systems (GIS) coupled with a number of
regional data (geographical, socio-economic and meteorological). The system introduced is a re risk model and it is
used before the emergence of a re event for prevention
planning and during a re crisis for supporting the
management of resources as well as for the assessment of
possible re combating strategies and actions. It provides
the user with maps (fuel map, NDVI map, risk estimates,
map of minimum time path needed to confront a wild re,
aggregated map integrated in an ARC VIEW 3.0 platform)
and a software tool integrating the data received from
satellite and in situ observations and producing the abovementioned maps. This software is based on Visual C++
computer language. A case study has been developed for the
Prefecture of Evoia, which is particularly vulnerable to
forest res due to the high mountains, the anomalous
landscape and the prevailing high wind intensities.
2. The DSS structure and functionality
The DDS provides a tool aimed at managing, analysing
and optimising the forest re ghting strategies in use
(Fig. 1). The system consists of eight modules:

1. The Data Acquisition (DA) module: It provides the

suitable mechanisms to transform the raw data coming
from different sources into a common format using
transformation packages such as ERDAS IMAGINE,
for the satellite images and common packages (Microsoft Excel) for the meteorological data to be used by the
integrated system.
2. These data consist of satellite images in the visible part
of the solar spectrum from LANDSAT an SPOT
satellites and meteorological data from monitoring
networks operating in the area of the application.
3. The Fuel Mapping (FM) module: It translates satellite
images into a GIS raster-based map that illustrates the
different types of fuels in the areas under study and it
illustrates a set of homogeneous fuel regions, each
containing one type of fuel, used to deliver a physical
natural re risk-map of the area.
4. The Scenarios Generation (SG) module: It provides
scenario data structures generated either by a probabilistic analysis of historical databases (re and meteorological), or by deterministic user-input parameters.
This data structure contains data relative to: re starting
points, forest fuel moisture contents, wind speed and
direction, and availability of existing re ghting
infrastructures and resources, considering possible
changes along the dened time period.
5. The Socio-Economic Risk characterisation module
(SRM): It is responsible for the analysis and characterisation of the socio-economic risk in the area under
study. This characterisation is the result of a rigorous
analysis of the different socio-economic elements that
affect the re risk in each location. The module
considers the variables indicating the behaviour of the
socio-economic re risk (local permanent population,
tourists, domestic animals, houses, type and height of
vegetation) and constructs the transformation functions
that will determine the nal value of re risk. The
transformation functions are calculated each time the
module is accessed. With these re risk values a socioeconomic re risk map is developed, which is the nal
output of this module, and is also essential for the
Probabilistic Planning module (see below).
6. The Probabilistic Planning (PP) module: This is the core
of the system. It is designed to accomplish two different
but related tasks: (i) to assist the user on resources and
infrastructure allocation planning, following specic
criteria based on actual forest service operations; and
(ii) to perform efciency analysis of resources and
infrastructures operation over the generated scenario,
along the considered simulation time period. In this
way, the system allows the user to optimise the
respective planning strategy and to comply with budget
and re ghting campaign objectives.
7. The Valuation (VAL) module: It is composed of two
sub-modules: (i) a function computing an overall re
risk value for a specic area taking into account three
aspects: economic, environmental and social; (ii) a

Please cite this article as: Marc Bonazountas, et al., A decision support system for managing forest re casualties, Journal of Environmental
Management (2006), doi:10.1016/j.jenvman.2006.06.016

ARTICLE IN PRESS
M. Bonazountas et al. / Journal of Environmental Management ] (]]]]) ]]]]]]

Non geo-referenced
DBs

Fire Behaviour
Simulator

Fuel Mapping
Integral Risk
Analysis

D
A
S

Socioeconomic
Risk
Characterisation

Probabilistic Planning
Cartography
&
geograhic DBs

Scenarios Generation

VALUATION

USER INTERFACE
Fig. 1. Flow chart of the system modules.

report generation sub-module employing the output

data of the PP module, in order to present the results
of the former analysis in a structured form according
to forest re service ofcial templates. These reports
contain the simulation parameters data, the infrastructures and resources used, and the resulting outcome
of the simulation: cost, losses, number of res, burned
area surface and associated attribute values of these
areas.
8. The User Interface (UI) module: It constitutes the shell
for user front-end system operations. Although it can be
considered as a module itself, UI participates in each
module providing support through the following four
main groups of functionalities:
 Serves as a common and standard data input platform.
 Data inputs according to pre-determined valid value
ranges.
 Assists user input by means of proposed default values
and pre-dened data tables.
 Presents, in a comprehensive manner, the existing raw
data as well as the data resulting from the processes
carried out by the different system modules. For this
purpose, tables, charts and maps are designed and used
in the form of reports and cartographic representations.

3. Integrated risk estimation

Integrated risk maps are generated from forest fuel
distribution and weather patterns. Weather patterns
include information on the moisture content of the soil,
atmospheric stability and wind speed/direction provided by
tables. Integrating risk involves taking into account the re
potential due to the fuel and the conditions together with
the intrinsic monetary value of the terrain where re could
occur. An approach to geographically characterise this

integrated risk is proposed in this work, according to which

territorial information is stored as raster layers. Uniform
conditions (forest fuel, slope, aspect, wind and other
weather parameters) are considered in each cell of a raster
layer. This allows the determination of local spread
conditions hence enabling the characterisation of re
destruction potential.
In order to quantify this concept, it is proposed to
introduce the time required for the re to burn the area
equivalent to a raster cell (teq). For this purpose, an elliptic
re spread is considered and teq corresponds to the time
when the ellipse area equals the raster cell area, which is
xed by the user. The process is repeated for every cell.
This allows considering an instantaneous destruction
rate expressed in surface units per time for every cell.
Combined with land value per surface unit, which is
estimated from economic, social and ecological components, it yields a loss rate in monetary units per time unit.
Finally, taking into account the prediction of the number
of res that are expected to occur in each cell, and
assuming that all the re outbreaks will propagate the same
way within the same raster cell, it is possible to estimate an
integrated value loss rate by simply multiplying the loss
rate by the number of expected res.
These risk indices are calculated for every cell in the
raster maps, thus synthesising the destruction potential and
value loss for the whole area. The resulting maps are
extremely useful to planners helping them to identify the
areas to which attention should be paid according to the
combination of re potential and land value.
4. Forest fuel mapping
In order to map the forest fuel several options
are available. In this work, we used LANDSAT and
SPOT satellite images as a valid, accurate, timely and

Please cite this article as: Marc Bonazountas, et al., A decision support system for managing forest re casualties, Journal of Environmental
Management (2006), doi:10.1016/j.jenvman.2006.06.016

ARTICLE IN PRESS
4

M. Bonazountas et al. / Journal of Environmental Management ] (]]]]) ]]]]]]

cost-effective means to obtain forest fuel maps. CORINE

land cover, introduced by Cornaert et al. (1992) and
modied by Sifakis et al. (2004), is used as the basis of the
classication scheme of the vegetation types we nally
used. Since the classication scheme of Sifakis et al. (2004)
is quite sophisticated for the needs of this work, two new
classication approaches for the vegetation identied by
image analysis, were introduced.
In order to obtain geographical distributions of fuel
availability it is required to introduce a relationship table
between vegetation classes and parameterised, fuel availability models.
Image classication has been used to identify vegetation
patches, grouped by classes. A divergent approach is
proposed based on sequential desegregation of a few initial,
easily distinguishable from satellite, coarse vegetation
classes into ner ones (Caballero, 2001). On the other
hand, a convergent approach starts with a very ne
division into a relatively large number of vegetation
classes. Aggregation into coarser vegetation types leads
to fuel models based on other criteria. The classication
system was evaluated during an in situ inspection and
found that it performs well, within the limits that the
specic application requires.
Fuel load is a critical parameter in forest fuel management. It is closely associated to vegetation type, state, and
age and is a determinant factor of the re spread and re
line intensity. Nevertheless, this parameter is difcult to
deduce from satellite observations. The rst approach
examined was the study of correlation between vegetation
indices (such as NDVI obtained from AVHRR observations) and fuel load. Results have been somewhat
discouraging due to the low correlation observed and
additional sources of information are considered, such as
the vegetation closure and average height. Therefore, any
forest fuel characterisation is subject to canopy cover,
which depends on the remote sensing of surface fuels.

5. Socio-economic risk prediction

A relationship between socio-economic factors and the
number and typology of forest res has been proven to
exist (Bonazountas et al., 2005). Most of the Mediterranean regions suffer from extensive forest res and in some
cases socio-economic factors are key determinants in
wildre causality classication. So the design and implementation of a Socio-Economic Risk Model (SER) that
provides the expected number of res by unit area, utilising
the relationship between socio-economic factors and forest
res is of great interest.
The approach adopted proposes to substitute the
traditional model of the most probable attributed cause
that attributes most forest res to arson and offenders, by a
SER based on an objectively dened risk situation.
Appropriate analytical categories are established and the
respective risk factors are identied.

The socio-economic risk module characterises and

analyses the socio-economic risk in the area under study.
This characterisation is the result of a rigorous analysis of
various socio-economic elements that affect the risk of re
in each location.
In the socio-economic risk module primary data (raw
socio-economic data from different sources), variables and
indicators are used (Bonazountas et al., 2005).
Variables are considered as qualitative elements (nondimensional) that characterise the socio-economic setting
of the forestry environment in a specic temporal or
geographic unit. A set of 156 variables has been selected
that inuence the re risk. This set was classied into 10
groups (Forestry, Livestock, Agriculture, Land uses, Sociodemographic, etc.), and 48 subgroups according to the area
under study. Indicators are functions that quantify the
variables from the available primary data. Each variable
can be quantied by more than one indicator, but among
the available indicators, those that offer greater accuracy
and reliability in the representation of the variable have
been selected.
Socio-economic factors are those components encountered within the forestry environment that contribute to the
explanation of both the intensity and spread of the wildre.
The module incorporates and uses ve factors representing
forestry protability, demographic pressure, social tension,
forestry culture and organisational logic. These factors are
sufcient to explain the increase/decrease of re risk.
The variables that indicate the socio-economic re risk and
the associated transformation functions determine the
nal value of re risk and yield a socio-economic risk
map (Fig. 2).

6. Fire defence service planning

Forest re defence efciency could be conceptually
simplied as the minimisation of time involved in the
operations, such as re detection, communication to base,
forces preparation and start-up, air and ground transport,
deployment of resources in the re area, re extinction and
control. To pre-determine the efciency in such simple
terms it has been proposed to apply spatial analysis

Fig. 2. Socioeconomic Risk Module (SRM) results: The separate risk

maps are integrated through SRM sub-module to produce the nal risk
map. The different colours present the different degrees of risk for a forest
re ignition.

Please cite this article as: Marc Bonazountas, et al., A decision support system for managing forest re casualties, Journal of Environmental
Management (2006), doi:10.1016/j.jenvman.2006.06.016

ARTICLE IN PRESS
M. Bonazountas et al. / Journal of Environmental Management ] (]]]]) ]]]]]]

techniques to the existing resources in a region. Planners

are allowed to select different distributions of bases, water
points, lookouts and roads and see the effect of their new
locations in terms of efciency (Fig. 3).
The access time of ground vehicles was calculated in a
traditional manner, using GIS network analysis tools. It is,
herein, proposed to use employ spatial analysis instead,
using raster maps vs. vector coverage. Thus, roads are
converted to raster cells into which a transport-resistance
value is stored, depending on the type and state of road and
terrain slope. This is interpreted as time, namely that
invested by vehicles in crossing each cell. Cross-country
access is considered also, and for this type of transportation, type of vegetation and slope are also considered in the
calculation.
From the ground Impedance Map an accumulated time
access map is produced using iterative automata that
follow the minimum-time path (Caballero, 2001). The
resulting map stores the time invested in travelling from a
given point to every cell. Thus, ground coverage, meaning
the percentage of cells with access time equal to or less than
a prescribed threshold value, can be determined. Road
network improvement and forest accessibility works,
evidently, have a direct inuence on ground coverage.
Aerial coverage of each raster cell is estimated by directly
considering the distance to airbases and the average speed
of aircrafts. The resulting map is combined with ground
access to estimate the overall coverage in terms of re
ghting resources. Apart from the estimation of access time
to the re area, the aforementioned maps are used to

calculate the efciency of re ghting vehicles in their water

re-charging cycles to the nearest water point.
Fire ghting efciency involves the knowledge of re
ghting personnel and available vehicles for every fuel
model, under circumstances that are largely variable.
Cross-country ground accessibility depends on vegetation
coverage and slope. Fire characteristics reduce or even
inhibit re ghting in direct attack mode, whilst the
construction of a re line is mainly restricted by vegetation
and slope. Air attack efciency could also be diminished
due to many factors, such as wind speed, smoke density,
topography or available daylight. All these considerations
are taken into account in the simulation of re ghting
operations, grouped into direct attack on ames, indirect
attack, and aircraft water discharges. All the calculations
are based on real data and observations, providing average
productivity values.
New infrastructure works, re-location or improvement
of existing infrastructures and forces deployment in a
region, have a cost. To optimise the use of an assigned
budget, it is proposed to effect a plan in accordance with
the predicted risks, understood in the way presented above.
This ability to combine risk forecasting, efciency assessment, and cost management is a real added value for a
decision-making supportive system.
6.1. Fire simulation
It is very useful for the forest managers to be able to
simulate the temporal evolution of a re and to estimate

Fig. 3. Forest re defence calculations.

Please cite this article as: Marc Bonazountas, et al., A decision support system for managing forest re casualties, Journal of Environmental
Management (2006), doi:10.1016/j.jenvman.2006.06.016

ARTICLE IN PRESS
M. Bonazountas et al. / Journal of Environmental Management ] (]]]]) ]]]]]]

the costs related to the operations needed for re defence,

as well as to losses due to wildre destruction (Jenkins,
2002; Vakalis et al., 2004). Simulations are based on a very
detailed forest re spread engine that calculates re
propagation and re characteristics for every cell. Flame
length, heat per unit area and re line intensity are also
estimated and used to size the re ghting resources
required to achieve a successful containment and extinction. An indicator of re importance for every re outbreak
in the region is introduced. This index evolves with time as
the re perimeter changes in size, location and intensity.
Instantly, the re importance index is used to adjust re
ghting resources and to allow efcient dispatching. Other
criteria are also considered such as efcient management or
re ghting units availability.
6.2. Evaluation and reporting
The system simulates a re area during a simulation time
period (13 days). As re scenario components such as
wind and other weather parameters are obtained from
probabilistic calculations they could differ from the real
world values in any single simulation. Thus, it is proposed
to run the same simulation several times using stochastic
perturbations of the key parameters. The nal output is the
tendencies of the re area evolution in term of probabilities
rather than deterministic predictions.
The simulation results (Fig. 3) include weather and wind
pattern evolution, re outbreaks distribution, and average
re growth values such as size, re line intensities, and re
importance. Along with these values several reports are
also obtained, regarding the resources usage, dispatching
and efciency.
The cost of re defence operations carried out during the
simulation period and the infrastructures maintenance cost
can also be computed.
The nal evaluation allows planners to identify the best
strategies by comparing costs, efciencies and losses.

forest fuel classication scheme. Two images were used: a

LANDSAT TM image from August 1997, and a SPOT
image from September 1997. A classication scheme was
adopted for the detailed evaluation of forest fuel availability in 25 sites during a visit in northern Evoia. The rst
satellite image and the classication scheme were used to
produce Fig. 3. Fig. 3 shows the forest re ghting plans
while the possible cost of the operation is calculated. The
classication scheme was applied to the second satellite
image to produce Fig. 4 which presents the fuel load
distribution over a part of the district of northern Evoia.
The dark area in the centre represents a burned region.
To examine if the results of the simulation made by the
DSS system are consistent with a case of real wild re local
forest managers applied the software tool in the application
area under real time conditions (e.g. during a forest re
ignited in the area).
The results of the simulation (Fig. 3) were compared
with the results of the re (Fig. 4) and it was found that the
software tool performed adequately. Specically, the extent
of the burned area was 119% of the simulated one; the time
calculated for the ground vehicles and aerial forces to reach
the re place was adequate in the rst stage (105% of the
time calculated) while in the second stage, it was slightly
higher (121%). The simulated cost of the re ghting
operation was 17% lower than the real one. From the
above it is obvious that the simulation was in general in
very good agreement with the real event and the software
tool could be used in real time conditions with remarkable
accuracy.

7. System verication
The system was applied in the island of Evoia located in
Central Greece and vulnerable to forest res. Evoia is the
second largest island of the Aegean Sea with an area of
3897 km2. It is very close to the mainland of Greece just 1 h
drive from the Athens metropolitan area, having direct
road connection through a bridge and several ferry-boat
lines along the coast, thus receiving weekend tourists.
Evoia has a complex mountainous and wooded landscape,
especially in its northern part. Vegetation is largely
xerophytic with extended pine forests that make the area
vulnerable to forest res. The summer season is very dry
with high temperatures and winds blowing strong from
NE-directions (etesian winds) leading to high re-propagation speed and intensity.
For the application area, fuel maps were constructed
combining satellite imagery data and an on-site developed

Fig. 4. Fuel loads in the application area. The dark area in the centre
represents the burned area.

Please cite this article as: Marc Bonazountas, et al., A decision support system for managing forest re casualties, Journal of Environmental
Management (2006), doi:10.1016/j.jenvman.2006.06.016

ARTICLE IN PRESS
M. Bonazountas et al. / Journal of Environmental Management ] (]]]]) ]]]]]]

8. Conclusions
This publication presents a tool for prevention and
examination of the spread of a wild re. The DSS
established is based on state of the art technologies in
identication of forest characteristics from satellite imagery, simulation of re behaviour, re risk assessment, and
geographic information systems. The volume of user
requirements addressed and the amount of meteorological
information included in the scheme make the system
efcient and effective. The information module of the
DSS represents a signicant part of the development effort,
since it reports details, to forest managers and to the
public, on expected re events and, in case of a real re, on
combating. The system has an open architecture and
permits expandability. It has been applied, veried and
validated in the area of Evoia, Central Greece, during a
real forest re event and delivered results close to the
expectations as these were dened by the relevant
authorities.
Acknowledgments
This research was funded by the European Commission,
under contract EU-FOMFIS project ENV-4CT96-0335.
References
Bonazountas, M., Kallidromitou, D., Kassomenos, P.A., Passas, N., 2005.
Forest re risk analysis. Human and Ecological Risk Assessment 11
(3), 617626.
Caballero, D., 2001. Algorithms for semi-automated dispatching in
FOMFIS system. In: Proceedings of the International Workshop on

Improving Dispatching for Forest Fire Control, Mediterranean

Agronomic Institute of Chania, Crete, Greece, December 68, 2001,
p. 26.
Cornaert, M.-H., Heymann, Y., O Brian, C., Peter, D., Sifakis, N., 1992.
CORINEan information system on the state of the environment of
the European Community. Geodetical Info Magazine 6, 5557.
Gonzales, A.F., Cuevas, J.M., Casanova, J.L., Calle, A., Illera, P., 1997.
Forest re risk assessment using NOAA AVHR images in the Valencia
area, Eastern Spain. International Journal of Remote Sensing 18 (10),
22012207.
Illera, P., Fernandez, A., Delgado, J.A., 1996. Temporal evolution of the
NDVI as an indicator of forest re danger. International Journal of
Remote Sensing 17, 10931105.
Illiadis, A., Papastavrou, A., Lefakis, P., 2002. A computer-based system
that classies the prefectures of Greece in forest res risk zones using
fuzzy sets. Forest Policy and Economics 4, 4354.
Jenkins, M.A., 2002. An examination of numerically simulated wildres to
low-level atmospheric stability and moisture and the consequences
for the Haines index. International Journal of Wildland Fire 11,
213232.
Kalabokidis, K., Gatzogiannis, S., Galatsidas, S., 2002. Introducing
wildre into forest management planning: towards a conceptual
approach. Forest Ecology and Management 158, 4150.
Nunez, L.R., Rodriguez, J.A., Proupin-Castineiras, J., 2000. Design
of risk as a tool to prevent forest res in the humid Atlantic
zone of Galicia (NW Spain). Thermochimica Acta 349 (12),
1169.
Sifakis, N., Paronis, D., Keramitsoglou, I., 2004. Combining AVHRR
imagery with CORINE land cover data to observe forest res and to
assess their consequences. International Journal of Applied Earth
Observation and Geoinformation 5 (4), 263274.
Thomson, W., Vertinsky, I., Schreir, H., Blackwell, B., 2000. Using forest
re hazard modelling in multiple use forest management planning.
Forest Ecology and Management 134, 163L176L.
Vakalis, D., Sarimveis, H., Kiranoudis, C.T., Alexandridis, A., Bafas, G.,
2004. A GIS based operational system for wildland re crisis
management. II. System architecture and case studies. Applied
Mathematical Modelling 28, 411425.

Please cite this article as: Marc Bonazountas, et al., A decision support system for managing forest re casualties, Journal of Environmental
Management (2006), doi:10.1016/j.jenvman.2006.06.016