SILVANUS Fire Spread Model

The Fire Spread Model (FSM), developed by EXUS, is a machine learning (ML) tool aiming to predict the development and spread of wildfires over the next 24-hour period. The FSM uses several parameters as inputs including: the terrain (elevation, slope, aspect), meteorological forecasts (temperature, wind speed, wind direction, etc.), fuel parameters (fuel type, moisture, canopy characteristics, etc.), barriers (firefighter efforts, roads, bodies of water, etc.), and the current location of the fire front. Given this information, the tool will generate a series of images depicting the fire-front location at 28 indicative times over the upcoming 24-hour period.

The main added value of the tool is its seamless integration in the Silvanus platform, enabling automatic execution of the prediction each time updated information (e.g. about fire-front or meteorological data) is available. Further, the tool is significantly faster than other state-of-the-art software solutions. No expert knowledge by the command center operator is required. The results of the simulation are sent to the User Interface for visualization in the command center.

The tool requirements, such as spatial resolution and domain of prediction, have been co-designed with the end-user partners of the project, ensuring that the tool is useful for them and addresses their needs. The speed of prediction and ease of use are the main features the stakeholders found attractive.

The figure below shows several sample predictions, with each row corresponding to a different scenario, e.g. different meteorological conditions, terrain, etc. The column on the left shows the input, represented here as the current fire front location. The prediction for the fire location in one hour (from the launch of the simulation) is shown in the middle column. The new predicted burned area is shown in the column on the right, arising from subtracting the input from the prediction.