Wildland fire modeling with an Eulerian level set method and automated calibration

• Framework for wildfire simulation using Eulerian Level Set method was presented and verified.
• Automated calibration successfully demonstrated for actual wildland fire.
• Could be combined with data assimilation and Monte Carlo simulation for ensemble predictions.

This paper presents the mathematical development of a geospatial model for simulating wildland fire spread. The Eulerian level set method (LSM), a mathematical technique that tracks interfaces between separate regions on a regular grid, is applied here to track the interface between burned (or burning) areas and green areas. Model physics include surface fire spread rate and direction, transition from surface fire to passive or active crown fire, ember lofting, trajectory tracking, and spot fire formation, acceleration from point ignitions, and modifications to fuel strata attributed to suppression activities. A novel aspect of this work involves application of a stochastic optimization algorithm to automatically calibrate baseline model inputs by comparing calculated fire perimeters to observed (target) fire perimeters. The wildland fire model and associated automated calibration technique are assessed by simulating the first 22 h of progression of the 2007 Moonlight Fire in Northern California. Fuels and topography inputs are obtained from the Landfire project while wind and weather inputs are obtained from high resolution numerical weather prediction. Fire areas simulated with the calibrated model agree well with target perimeters.