Simulating wildfires from single branches to local weather systems

Amplified by climate change, wildfires are causing devastating damages and loss of life, and are indeed becoming an existential threat in some regions of the U.S. and worldwide. Accurate simulations of wildfires can be used in wildfire prevention, in training, and for operational planning during an ongoing wildfire. Current fast methods use concepts of reaction–diffusion systems, wave proagation, or percolation, whereas methods that model individual trees typically take significant computational time on supercomputers. We present a new numerical multiscale framework that models tree structures down to branches, includes fluid dynamical modeling of local weather systems over realistic terrain, and runs at interactive rates. Our method allows tree representation with a previously unavailable fidelity, while simultaneously capturing atmospheric fluid dynamics including flammagenitus cloud formation and precipitation caused by the fire. The method captures the impact of tree topology, terrain characteristics, and ecosystem composition on fire spread. The interactive computational speed allows for an intuitive interaction with the simulation, where the impact of fire retardants, firebreaks, or rain can be experienced and studied.