Parallel Simulation of Wildfire Spread Using Python and MPI

As we all know, wildfires are becoming even more frequent and destructive, demanding faster and more accurate predictive models to assist emergency response. This project proposes a parallel wildfire spread simulation built with Python and exposed through a REST API, designed to leverage high-performance computing (HPC) environments for large-scale analysis.

The simulation engine models fire dynamics across heterogeneous landscapes using a cellular automaton or Rothermel-based physical spread model, accounting for fuel type, moisture, wind, and terrain slope. The system is parallelized with Dask and Numba, enabling distributed computation across multiple nodes and GPUs. This parallelization strategy drastically reduces runtime, allowing ensemble forecasting and uncertainty quantification at high spatial resolution.

A FastAPIinterface orchestrates simulations, allowing external systems to submit parameterized jobs, monitor progress, and retrieve results. This modular design enables integration with GIS datasets and real-time environmental inputs, making it suitable for decision support in wildfire management and research.

The project aims to demonstrate how open-source Python ecosystems can deliver scientifically meaningful, scalable wildfire simulations in HPC environments.