P3 WRF-HiFire: development of a high-order accurate fire-spread model in WRF using the level-set method and its evaluation

Munoz-Esparza, Domingo, Pedro Jimenez, and Branko Kosovic, National Center for Atmospheric Research

WRF includes a coupled model to simulate wildland fires and its impact on weather conditions, which uses the level-set method to track and propagate the fire front. In order to improve the accuracy of the level-set method for coupled wildfire/atmospheric modeling, we have carried out major developments of the existing level-set algorithm in WRF as described below. First, we have rewritten the level-set algorithm to be fully parallelized, since the released version of the code only allows for a specific set of low-order numerical schemes and needs to perform additional calculations out of tile dimensions. We have also implemented higher-order numerical discretization schemes for the advection of the level-set method: third- and fifth-order weighted essentially non-oscillatory (WENO) and a third-order explicit Runge–Kutta (ERK) scheme for temporal integration. Finally, we have implemented a reinitialization method that allows the level-set function to remain a signed distance function throughout the entire simulation, improving the stability of the simulation.

Developments are evaluated against both idealized and real wildland fires. Using idealized cases consisting of a fire line propagating in a uniform velocity field and in a turbulent-flow environment, we show that WRF-HiFire requires a grid eight times less refined to obtain the same results as the released WRF-Fire version. In addition, the rate-of-spread error is reduced by a factor of three for the same grid size. Moreover, in the case of turbulent flow, larger discretization errors in WRF-Fire combined with non-linearities of the forcing velocity field result in different fire perimeter shapes. On the contrary, WRF-HiFire always exhibits similar fire perimeter shapes regardless of the spatial discretization. Finally, we will show the improvements of the new algorithm in simulating the large grass fire of Last Chance that occurred in Colorado in 2012. The simulations were validated using satellite data for the burned area and illustrate the improved simulation of the fire evolution with WRF-HiFire.