Forward Model


FARSITE – an operational fire area simulator from the US Forest Service to simulate advancement of a fire front.

A forward model is used in our approach to propagate advancement of a moving fire front. We use two different models to accomplish this, a research model, firefly, based on a level-set approach, and FARSITE, an operational model based on Huygens principle, used to respond to actively burning fires. By investigating both approaches, both detailed scientific knowledge to improve data assimilation as well as a pathway for immediate use in the field can be approached.

FARSITE is an operational fire area simulator developed by the US Forest Service to simulate the growth and speed of a propagating wildland fire. As an operational model, FARSITE is a simulator used to respond to actively burning fires, utilizing simplified analytic models to predict the propagation of a fire as a function of time. Underlying this tool, simplified mathematical models must be relied upon to solve for fire propagation faster than real time. The Rothermel model, a one-dimensional semi-empirical fire spread model based on energy considerations is the basis for simulations, followed by elliptical assumptions of fire front shape correlated to real fire behavior. Additional submodels for crown fires, spotting, etc. are included that make the model ideal for large-scale operational use. Unfortunately, FARSITE requires all data to be input before running a simulation, even fuel moisture, wind speed and direction, which are known to change dynamically. Because of this, it is unable to adjust to changes in the environment or correct for deviations between actual fire behavior and user inputs (fuel, weather, etc.) which often have enormous errors. Our work will be the first to try to integrate this tool into a data assimilation approach.


firefly, a fire spread model using a level-set approach developed by Trouve, Ricci and Rochoux.

Firefly is a regional-scale data-driven wildfire spread model, designed from the ground up to support data assimilation research. It was developed as a joint effort between Melanie Rochoux and Sophie Ricci at CERFACS in France and Arnaud Trouve at the University of Maryland. Similar to FARSITE, Firefly described the fire as a two-dimensional front that self-propagates normal to itself into unburnt vegetation, basing the forward rate-of-spread on the Rothermel model. Unlike FARSITE, Firefly uses a level-set approach to propagate the fire front forward, not including empirically-measured elliptic assumptions. An Ensemble Kalman Filter (EnKF)-based data assimilation algorithm that features a choice between a parameter estimation (PE) approach in which the estimation targets (the control variables) are the input parameters of the ROS model and a state estimation (SE) approach in which the estimation targets are the spatial coordinates of the discretized fire front are included in this firefly-EnKF version.