Developing a FARSITE wildfire simulation framework for Romania: fuel model mapping and evaluation using historical wildfires

Abstract

Wildfire simulation can support fire-risk assessment and management, but its application depends on the availability of consistent spatial information on fuels, topography, weather, and fuel moisture. This study developed and evaluated a FARSITE-based wildfire simulation framework for Romania, with particular emphasis on generating the spatial input datasets required for its application. A national fuel-model map was produced by integrating land-cover and canopy information and assigning standard fire-behaviour fuel models according to vegetation characteristics. The framework was evaluated through retrospective simulations of four historical wildfires in western and southwestern Romania. Simulated and observed fire perimeters were compared using complementary spatial agreement metrics, while rate of spread, fireline intensity, and flame length were analysed to characterize simulated fire behaviour. Spatial agreement varied among the four case studies, with Sørensen–Dice coefficients ranging from 0.660 to 0.879 and Cohen’s Kappa values from 0.436 to 0.687. The highest agreement was obtained for the Șvinița wildfire, whereas Ilia showed the lowest agreement and the greatest overprediction. Simulated fire behaviour also varied substantially among events and fuel types, with the highest spread rates, fireline intensities, and flame lengths generally associated with grass- and shrub-dominated fuels. Overall, the results indicate that the proposed framework provides a suitable basis for further FARSITE application and development in Romania under the fuel and environmental conditions represented by the evaluated fires. Further improvements should focus on local fuel characterization, more representative meteorological and topographic inputs, incorporation of information on firefighting operations, and evaluation using a broader range of historical wildfires.