Instantiating a Diffusion Network Model to Support Wildfire Management

Abstract

Wildfires require effective responses considering multiple constraints and conflicting goals. We provide a methodology and a tool enabling stakeholders to compute risk maps and use them in practical and realistic scenarios. The territory is modeled as a network where nodes are land patches subject to fire and links model the probability of fire spread from one patch to another. We discuss a risk function and show how to compute it effectively. We show how to instantiate the model on a real landscape. The methodology describes how to compute each patch's borders and probabilities of ignition and how to estimate the probability of fire spreading from one patch to a neighboring one. We embed the methodology into an ad-hoc modular tool-chain using geographical data, a fire simulator and geospatial tools. As a proof-of-concept, the tool-chain is applied in three different experiments on a region of Corsica, France, aiming at simulating a realistic scenario and measuring the sensitivity of the methodology with increasing wind speed or variable wind directions. We finally introduce the web application that incorporates the tool-chain and enables users to manipulate the model intuitively through an interactive map, evaluate what-if scenarios, and simulate the effects of different fire preventive measures.