Using state-of-the-art simulations for effective and sustainable wildfire management.
The problem
Wildfires can be disastrous events that bring devastation to local ecosystems and communities around the world. Anthropogenic activity, to a large extent, is the reason behind the upsurge of wildfires in the latest decades. In the development of FireDynamo, a fast and accurate software for real-time fire propagation prediction and fire risk assessment, we aim to offer tools to help land managers prevent disasters such as the Australian black summer and to help maintain our natural resources in a sustainable way worldwide.
Our solution
We plan to offer a wildfire pathway prediction and fire risk assessment tool. Our aim is to provide a tool which land managers, private landowners and risk assessors can apply to their own regions of interest. The advantage of our solution lays in capability to hyper-localize vegetation data (i.e. assuming individual properties of particular tree species), high spatial resolution (~5-10 m) and speed of large-scale simulation (millions of agents in minutes).
We aim to provide a tool for:
● Effective decision making and efficient use of resources in real time during ongoing fire
● Supporting preventative measures and fire management
● Informing risk evaluations and long-term strategies
How do we do this?
The key to our approach is by using innovative technology developed at CERN. This technology allows very efficient modelling of single agents (i.e. one tree) within a landscape by using satellite, wind, topographical and most importantly local vegetation data for accurate and fast prediction of fire propagation.
Data and Machine Learning
We use satellite data to gather all necessary information - such as soil moisture, topology or infrastructure and using Machine Learning methods on the past fire data we can predict points with highest probability of ignition.
Vegetation and environment
We model the propagation of fire in the forest using single plants as agents. We can assign specific properties and behavior of the vegetation, such as moisture content, susceptibility to fire, burning time or size, that can depend on the species of the tree and the age. The environment of the agents will also depend on the climate factors and the topology of the terrain.
Simulate fire propagation
Using agent-based simulations we can perform very accurate estimations of the progression of fire in the given area. Our goal is to develop sustainable plans for forest management that minimise the risk of fire and provide efficient strategy of fire exhaustion.