Dionysus: a Randomized Fire Growth Simulator
The Prometheus Fire Growth Model is a deterministic wildfire simulator, developed for use in Canada. The spread of a simulated fire is governed by a set of differential equations which are solved numerically. Given weather, topographical and fuel information, the simulated fire front
is plotted at equally spaced times. Unpredictability of fire behaviour makes deterministic predictions inaccurate, motivating the need for stochastic alternatives. We argue that the ensemble approaches that are gaining popularity do not adequately model the uncertainty of the fire growth process. Furthermore, ensemble methods invariably require repeated simulation rendering them useless for real-time wildfire prediction.
By statistically modelling the data upon which the Prometheus growth model is based, it is possible to obtain a distribution of fire front predictions. If the weather forecast is reliable, repeatestochastic simulation is not required, so probability contours require no more computing time than what is required for a single deterministic contour.
This is joint work with L. Han.