Developing a geospatial data-driven solution for rapid natural wildfire risk assessment

Abstract

Computational natural wildfire simulation is a computing-intensive process. The process is also challenging because of the need to integrate data with wide spatial and temporal variability. Our study sought to simulate rapidly spreading natural wildfire with fidelity and quality through computational realization. We developed a novel probabilistic wildfire risk assessment tool whose operation was driven by real-time wildfire observations. A Gaussian transformation incorporating present and historical geographical data to the wildfire model was adopted to accommodate scale differences in the datasets. The model outputs, therefore, depict possible spread pathways using Monte Carlo simulations. We created a computational solution for deploying wildfire simulations to a highly scalable, distributed and parallel computing framework, which facilitated a fairly linear increase in the simulation run time as the computation load increased exponentially. Our computational solution synthesized and fully automated the various stages of the process, from data preparation to analysis and visualization. The platform can potentially provide real-time decision-making support to wildfire hazard management.