This multi-level agent-based model was created to assess forest fire management in southern Switzerland. It simulates the daily movements of individuals to investigate the impact on fire occurrences. The model employs an open-source platform called GAMA to build spatially explicit agent-based simulations allowing multiple levels of modeling. Preliminary results show the model can perform a visual correlation between fire ignitions and human presence by tracking people moving on the road network and capturing population distribution over time. Investigation of firefighters' response efficiency is still ongoing.
1. A multi-level agent-based model to assess the forest fire
management in the southern swiss alpine region
MINELLI Annalisa (1), TONINI Marj (2)
(1) Indipendent Researcher in Open GIS and Agent Based Modelling, Perugia (Italy)
(2) Institute of Earth Surface Dynamics, Faculty of Geosciences and Environment, University of Lausanne (Switzerland)
Contact: annalisa.minelli@gmail.com
Alpine forest fires
Forest fire are assuming an increasing importance especially in relation to urban sprawl. The wildland urban
interface (WUI) is a central feature related to this phenomena.
In the Swiss Alpine environment WUI lies between the urban area and the steep forests, intermingled with
agricultural lands, vineyards and unproductive surfaces.
Objectives of the study
- Create a simulation model acting at local scale to assess the forest fire dynamic in a pilot area located in the
southern Swiss alpine region (Canton Ticino).
- This allows performing detailed analysis, as for example: simulating the daily movement of each single
active person in order to investigate the influence of the mobility on fires occurrences; investigate the
efficiency of different fires fight strategies, by simulating the displacement of the firefighters.
Open Source Geospatial Tools
We employed GAMA as modeling platform for building spatially explicit agent-based simulations allowing
multilevel modelling.
Data pre and post treatment was performed with QGIS.
Forest fires in the Swiss Alpine environment
Institute of Earth Surface Dynamics
Preliminary results
Method: Multi-Agent System (MAS)
A multi-agent system (MAS) is a computerized system characterized by
multiple intelligent agents interacting within a specific environment.
All the georeferenced objects, put into a Multi Agent platform, acquire a
sort of “intelligence” which allows them to interact with other elements
(agents) and with the surrounding environment.
A multi-agent system (MAS) is a computerized system characterized by
multiple intelligent agents interacting within a specific environment.
All the georeferenced objects, put into a Multi Agent platform, acquire a
sort of “intelligence” which allows them to interact with other elements
(agents) and with the surrounding environment.
population
census
(hectometric grid)
dwelling
census
(hectometric grid)
enterprise
census
(hectometric grid)
active people
moving on the
road network
road network
(with speed limits)
is a fire
ignited?
call the
firefighters
save a vector file
of positions of
people andfires
record the
intervention
time
save a vector
file of traffic with
a time granularity
choose a
fire fighting
technique
YES
proceed to
the next
cycle
NO
forest fires
geodatabase
Our implemented model (figure above) allows performing a visual
correlation between ignition of fires and human presence. Investigation of
the efficiency of firefighters intervention is a work in progress.
Starting from the input data, the simulation begins with people moving on the road network (e.g. from home to work, to schools, to recreational spaces);
each segment of the road network is characterized by different speed limits and the movement of the agent responds to the shortest path route logic.
The frequentation of each zone is captured periodically by specifying a temporal granularity, which could vary in reason of the time extension of the
simulation. At different time step, a caption of the current situation is automatically created and given in output as:
graph of the evolution of nubrer of people at a given distance (here 50, 100 and 200 m) from the forest fire
pie chart of the percentage of people moving (working) or staying home (resting peole)
shapefile with the location and number of people moving ( working) or staying home ( resting people) and the fire event ( )
Dataset sources
Forest firess (1990-2015): from the forest fire database of Switzerland (http://www.wsl.ch/swissfire/).
Features in the landscape (rail way, highway, streets, building and forest): from the Topographic Landscape Model
(TLM 2010) elaborated by the Federal Office of Topography (Swisstopo).
Census of population, dwelling and enterprises (2000): from by the Federal Statistical Office (FSO).
Resting
10%
Working
90%
Time = 12 am
Resting
100%
Working
0%
Time = 10 pm
Resting
86%
Working
14%
Time = 6.30 am
Resting
47%Working
53%
Time = 9 am (day 2)
Resting
10%
Working
90%
Time = 1 pm (day 2)
Resting
100%
Working
0%
Time = 9 pm (day 1)
First example: fire event #1
Start time: 8.30 pm (day 1)
End time: 12 am (day 2)
Second example: fire event #2
Start time: 6.30 am
End time: 12 pm
Time (cycle #) Time (cycle #)