Developed by Dr. Chris Lautenberger at Cloudfire, ELMFIRE is an open-source, scientific wildland fire spread model licensed under the Eclipse Public License 2.0. Implemented as part of the PyreCast near-term fire forecasting system, ELMFIRE is engineered to operate on a high-resolution gridded landscape, integrating local fire behavior equations with wind, fuel, topography, and moisture data to predict fire growth in both time and space.

At its core, ELMFIRE represents the evolving fire perimeter as a continuous surface within a 2D or 3D grid. This mathematical surface tracks the arrival time of fire across every cell in the domain. By solving the Hamilton-Jacobi partial differential equation using high-order numerical solvers, ELMFIRE allows for the realistic simulation of fire spread in any direction, overcoming limitations of traditional raster or vector-based models.

ELMFIRE is designed to capture complex wildfire behavior, including non-convex fire shapes, rapid shifts in spread direction, and detailed interactions between fire fronts, spotting, and terrain. Its architecture accommodates probabilistic and ensemble inputs, enabling robust representation of uncertainty and supporting advanced applications such as real-time forecasting through platforms like PyreCast, utility risk mitigation, and dynamic hazard mapping. The model’s modular design operates efficiently on both CPU and GPU environments, allowing scalable deployment across research, operational, and planning contexts. As an open-source tool built on a rigorous mathematical foundation, ELMFIRE is highly adaptable and well-suited for integration into modern decision-support systems, making it a key component of next-generation wildfire science.