THEME ALIGNMENT

INDUSTRY PROBLEM

The rapid expansion of LiDAR, BIM, photogrammetry, and UAV mapping has made high-resolution 3D environmental data widely available for urban planning and hazard modelling. However, current operational wildfire models remain 2D surface-based such as SPARK), unable to ingest 3D grids or capture vertical structure, accurate estimates of fuel distribution, or mixed urban–bush interactions. Our early physics-based voxel fire prototype quickly hit memory limits beyond ~300 × 300 × 150 voxels due to the storage demands of multiple physical fields and solver buffers. It also struggles to incorporate dynamic drivers, such as wind, fuel, moisture, and atmospheric conditions, because the process of simulation cannot be interrupted. As a result, there is no efficient 3D framework for modelling wildfire dynamics in large, mixed voxel environments.