Computational Modeling of Large Wildfires: A Roadmap

Abstract—Wildland f ire behavior, par ticular ly that of large, uncontrol led wild fires, has not been wel l understood or predic ted. Our methodology to simula te this phenomenon uses h igh-reso lution dynamic models made o f numerical weather prediction (NWP) mode ls coupled to fire behavior models to s imulate f ire behavior. NWP mode ls are capable of modeling very high re solu tion (< 100 m) atmospher ic f lows. The wild land f ire component is based upon semi-empirical formulas for firel ine rate of spread, post-frontal heat re lease, and a canopy fire. The f ire behavior is coupled to the a tmospheric model such that low level winds drive the spread of the surface fire, which in turn releases sensible heat, latent heat, and smoke f luxes in to the lower atmosphere, feeding back to affect the winds directing the fire. These coupled dynamic models capture the rapid spread downwind, flank runs up canyons, b ifurcations of the f ire into two heads, and rough agreement in area, shape, and d irection of spread at periods for which f ire location data is available. Ye t, in triguing computa tiona l science questions arise in apply ing such mode ls in a predictive manner, inc luding physical processe s that span a vast range of scales, processes such as spotting tha t cannot be modeled deterministical ly, est imating the consequences of uncertain ty, the efforts to steer s imulations with f ield data ("data assimila tion"), l ingering issues with short term forecast ing of weather tha t may show ski ll only on the order of a few hours, and the diff iculty o f gathering pertinent da ta for verification and initia lization in a dangerous environment. Keywords-Numerical model; weather prediction, wildland fire model, wildfire, forest fire, fire behavior I