A GPU-Accelerated Hydrodynamics Solver For Atmosphere-Fire Interactions

Abstract

A fundamental process to understand fire spread is the atmospheric flow. Building computational tools to simulate this complex flow has several challenges including boundary layer effects, resolving vegetation and the forest canopies, conserving fluid mass, and incorporating fire-induced flows. We develop a two-dimensional hydrodynamic solver that models fire-induced flow as a convective sink that converts the two-dimensional horizontal flow into a vertical flow through the buoyant plume. The resulting equations are the two-dimensional Navier-Stokes equations, but with point source delta functions appearing in the conservation of mass equation. We develop a projection method to solve these equations and implement them on a GPU architecture. The ultimate goalis to simulate wildfire spread faster than real-time, and with the ability for users to introduce real-time updates in an augmented reality sandbox.