Volcanic fire hazard modeling

Abstract

Numerical models have been extensively explored for simulating forest fire propagation, employing a spectrum of approaches ranging from intricate physical models to simpler graph or cellular automata-based models. These models rely on various input parameters such as humidity, radiant capacity, vegetation flammability, tree types, wind and topography to accurately replicate fire spread. When integrated with Geographic Information Systems (GIS), these models facilitate classifying soil and mapping burnt areas. Despite the progress in forest fire models, a dedicated model for volcanic-induced fires is notably absent. Leveraging our expertise in lava flow hazard modeling, we have introduced a novel numerical model to address the current lack of a specific model for volcanic-induced fires. This model is based on cellular automata to tackle issues related to accessibility, usability, and computational costs inherent in existing applications. Indeed, the increased availability of low-cost, high-performance computing hardware has enhanced the accessibility and cost-effectiveness of cellular automata-based simulations, enabling the simulation of fire propagation under multiple environmental conditions. Our new model is able to generate fire hazard scenarios, providing insights into the likelihood of combustion. We present test cases using Stromboli Island as a case study, as it is an area where volcanic eruptions already caused fires that also contributed to human loss. Our results exhibit good spatial accuracy, with a Brier score of 0.188±0.002 and 0.073±0.001 for the fire spread on July 3, 2019, and May 25, 2022, respectively.