decisions have to be made in limite

decisions have to be made in limited time. Visualizing simulated
fire spread predictions and how they evolve in space and time on
top of the forest’s terrain can help crisis managers, firefighter
commanders, but also volunteers and affected citizens in taking
proper actions and making correct decisions before it is too late.
However, while accurate modeling of wildfire’s behavior is
certainly a necessity, it remains a challenging problem due to the
large number of time varying parameters involved that are
difficult to estimate in real-time. The most popular models for
surface fires employed in practice are based on Rothermel’s
classical semi-empirical approach (Rothermel, 1972). It has
formed the basis for developing “fire spread” simulators, since it
balances successfully the input data complexity vs. the output
accuracy tradeoff. Among the established wildfire simulation
tools available we mention BehavePlus, FlamMap, FARSITE, FSPro.
A thorough review of the wildfire spread models and tools can be
found in (Sullivan, 2009).
BehavePlus (Andrews et al., 2003) has been developed mainly
for educational purposes. It attempts to avoid simulation setup
complexity, e.g. using spatial data, and cannot be used to describe a
real fire event. FlamMap (Finney, 1999) adds the spatial component,
allowing conditions to vary in different areas of the forest. It takes
as input detailed spatial information for the forest area: slope,
aspect, fuel models and canopy cover, and produces static maps of
fire line characteristics, e.g. fire line intensity. FARSITE (Fire Area
Simulator) (Finney, 1998) adds the temporal component, allowing
conditions to vary during the simulation period. It requires the
same spatial information as FlamMap but also needs temporal
weather data layers. WFDSS (Wildland Fire Decision Support System)
(Noonan-Wright et al., 2011) uses the FSPro (Fire Spread
Probability) simulator, which introduces probabilistic fire spread
from a known perimeter or point based on multiple FARSITE simulations
and historical weather sequences derived from Remote
Automated Weather Stations (RAWS). FSPro also offers the ability to
modify the spatial characteristics of the forest area (i.e. its fuel
models), by allowing the user to assign rules such as e.g. “alter fuel
model 10 to fuel model 11 if the elevation is higher than 1000 m”.
Although the wildfire simulation tools described above are
complete and mature, they are also quite complicated for untrained
users since they require setting up a large number of difficult to
obtain GIS (Geographical Information System) input files. These GIS
files have also to be co-registered, with identical resolution, extent,
projection and datum. Their scope is mainly long-term strategic
decision support, and they are therefore difficult for a non-fire
behavior specialist to setup and use. Moreover, they do not provide
ways for the user to introduce, in an interactive and graphical
manner, possible human interventions to the forest’s spatial characteristics,
e.g. perturbations to the fuel models, in order to create
“what-if” simulation scenarios, making again the parallel use of GIS
platforms a necessity.
Recent work has emphasized the interaction between the user
and the wildfire simulation tools. In Yun et al. (2011) the authors
use advanced graphics and virtual reality techniques to provide
sophisticated renderings of the wildfire. VFire (Hoang et al., 2010)
has created an immersive simulation system, which renders the
virtual world in physical scale. The wildfire forecasting system
developed to aid the Canary Islands authorities (Castrillón et al.,
2011) is focusing mainly on visualization methods for monitoring
a real wildfire event and in using dynamic 3D objects to represent
humans and material resources mobilized to contain it, a very
useful feature from a fire management perspective. It provides to
the authorities all the input data layers that a wildfire simulation
requires, although this is limited to a specific area and does not
reach the large-scale coverage currently provided only by WFDSS
for the United States. Moreover, the use of FARSITE as the
simulation engine imposes serious time constraints and cannot be
used to perform in real-time “what-if” parallel simulation runs
without relying on a High Performance Computing (HPC) resource.
Considering the European continent, to the best of our knowledge,
every currently available wildfire behavior monitoring and simulation
tool relies on data layers (or other tools) that are not available
in the public domain.
Motivated from the above-mentioned limitations, our main goal
has been to design and build a user-friendly Web-based simulation
tool. Based exclusively on publicly available application programming
interfaces (APIs) and Web services, we developed methods
around the simulation core which streamline the simulation procedure
into a workflow with focus on two important areas: (i) hide
completely from the user the simulation set