The Spark Wildfire Prediction System

Abstract

Spark is a computational platform for modelling wildfire behavior, developed for research, risk assessment and operational use. The platform has been designed for flexibility, allowing users to define wildfire models, process weather streams and spatial data, and visualize and analyze the resulting simulations. The core fire propagation solver is a raster-based level set method that calculates the progression of a fire perimeter over given terrain layers and allows the creation, merging and interaction of any number of individual fire perimeters. Propagation in different fuel types is implemented using text-based scripts that define the fire behavior and rate-of-spread, rather than having pre-determined models built into the system. This gives the system flexibility to implement any available fire behavior models. The system is currently being used to research and implement many aspects of fire behavior including the development of new rate-of-spread models, integration with meteorological and remote sensing data and the modelling of firebrand dynamics and fire-atmosphere interaction. The system is freely available to researchers, practitioners and managers in the wildfire community.