Abstract

Wildfires are a problem of great interest in the world because of the damage they cause every year to forest, wild fauna and flora, an also threatening human lives, among others. There exist some computational models for numerical simulations to address this phenomena, and a few of them have an open-source implementation. The goal of this work is to extend a CPU Python’s implementation of wildfire open-source framework developed in NumPy, to a GPU improved version using CuPy. The algorithm used is based on a numerical discretization of a wildfire spreading mathematical model described by a system of partial differential equations. Computational and mathematical components, numerical simulations and applications are described in details in the document. In addition, this work includes a brief performance comparison between both implementations, pointing out that we can achieve lower execution times using the CuPy GPU implementation without spending enough programming effort.