Wildfire Simulation with FDS: From GIS Data to High-Performance Computing
The prediction of wildfire behavior is a critical challenge for fire safety engineers and environmental scientists. While the Fire Dynamics Simulator (FDS) is traditionally known for indoor fire modeling, its capabilities for large-scale wildland scenarios have expanded significantly. By combining geospatial data, advanced numerical methods, and the power of cloud computing, it is now possible to simulate complex wildfire spreads with unprecedented detail.
Bridging the Gap: Using qgis2fds for Geometry
One of the most time-consuming aspects of wildfire modeling is accurately representing the terrain and vegetation. FDS requires a specific input format, and manually coding thousands of grid cells to match a hilly landscape is nearly impossible.
This is where qgis2fds becomes a game-changer. It is an open-source plugin for QGIS (Quantum GIS) designed to bridge the gap between Geographic Information Systems and FDS.
- Terrain Extraction: It processes Digital Elevation Models (DEM) to generate realistic 3D terrain for FDS domains using
&OBSTor&GEOMlines. - Land Use Mapping: It translates vegetation layers into FDS “surfaces” with specific fuel properties.
- Seamless Export: Instead of manual preprocessing, you can export simulation-ready
.fdsinput files directly from your GIS environment, ensuring your model is georeferenced and accurate.
Tracking the Fire Front: The LEVELSET Method
To simulate how a fire moves across a surface (like a grass field or a forest floor) without modeling every single leaf and twig as a 3D object, FDS utilizes the Level Set method. In this approach, the fire front is treated as a thin interface that propagates at a specified Rate of Spread (ROS).
Choosing your LEVEL_SET_MODE
The behavior of the simulation depends heavily on how you define the physics interaction via the LEVEL_SET_MODE parameter in the &MISC namelist. Selecting the right mode is crucial for balancing accuracy and computational cost:
| Mode | Description | Use Case |
| 1 | Stand-alone Level Set: Only the front tracking is performed. The wind ignores terrain and there is no actual “fire” (heat/smoke). | Quick geometric verification of spread patterns. |
| 2 | Frozen Wind: The wind field is established over the terrain first, but “freezes” once the fire ignites. | Large-scale spreads where fire-induced winds are secondary to ambient wind. |
| 3 | Terrain-Only: Wind follows the terrain contours, but there is no heat release. | Studying wind flow over complex topography without ignition. |
| 4 | Fully Coupled: The gold standard. Wind, terrain, and fire interact. The fire burns for a set duration and releases heat, which in turn affects the local wind. | High-fidelity research and complex fire-atmosphere interaction studies. |
Implementing it in the .fds file
To set up a fully-coupled wildfire simulation (Mode 4), your code structure would look like this:
1. Global Configuration:
&MISC LEVEL_SET_MODE=4 /
2. Surface & Fuel Definition:
You must define a &SURF that links to a fuel model (like the Rothermel-based ‘BEHAVE’ model):
&SURF ID='FOREST_FUEL',
LEVEL_SET_FUEL_MODEL='BEHAVE',
MOISTURE_FRACTION=0.10,
HEAT_RELEASE_RATE_PER_AREA=250.0 /
3. Ignition:
Define the starting point of the fire front using an &INIT line:
&INIT XB=10.0, 12.0, 20.0, 22.0, 0.0, 2.0, STRAT_ID='FIRE', SURF_ID='FOREST_FUEL' /
Scaling Up: Simulation on CloudHPC.cloud
Wildfire simulations are computationally expensive. Modeling a few square kilometers of terrain with a fine enough grid to capture wind-fire interactions (especially in Mode 4) can take days on a standard workstation.
To overcome these hardware limitations, our simulations are performed on cloudhpc.cloud. By leveraging a high-performance computing (HPC) environment, we can:
- Parallelize the workload: Split the domain into multiple meshes and run them across dozens of vCPUs using MPI.
- Reduce Lead Time: What takes 48 hours locally can often be completed in a fraction of the time on the cloud.
- Visualize Remotely: Use Smokeview directly on the cloud via remote desktop to check results without downloading massive output files.
By integrating qgis2fds for setup, the Level Set method for physics, and CloudHPC for execution, you can move from simple sketches to complex, data-driven wildfire predictions with ease.
CloudHPC is a HPC provider to run engineering simulations on the cloud. CloudHPC provides from 1 to 224 vCPUs for each process in several configuration of HPC infrastructure - both multi-thread and multi-core. Current software ranges includes several CAE, CFD, FEA, FEM software among which OpenFOAM, FDS, Blender and several others.
New users benefit of a FREE trial of 300 vCPU/Hours to be used on the platform in order to test the platform, all each features and verify if it is suitable for their needs
