Authors: Jeffrey Katan*, Universite De Montreal, Liliana Perez, Universite de Montreal
Topics: Spatial Analysis & Modeling, Natural Resources
Keywords: agent-based modelling, complex systems, forest fires, geographic information systems (GIS), hybrid resolution
Session Type: Paper
Start / End Time: 12:40 PM / 2:20 PM
Room: Taylor, Marriott, Mezzanine Level
Presentation File: No File Uploaded
An important ecological, cultural, and commercial natural resource, forests ecosystems are shaped by important disturbance processes such as wildfires. Understanding and modelling the spatial behaviour of wildfires is important when designing adaptive fire management strategies in order to aid firefighting activities and protect human life and property. Forest fire behaviour is commonly modelled using cellular automata (CA). The oft-cited reason for choosing CA over an agent-based approach is the high computational demand of the latter. With the increasing availability of great processing power, computational capacity should no longer be an intrinsic limitation in model design. This study presents a novel approach to modelling forest fires using agent based modelling (ABM). In CA approaches, neighborhood types can cause distortions in fire spread patterns and cellular automata have a fixed resolution. In addition, data are not always available in as fine a resolution as desired. In this ABM approach, mobile agents travel across a simulated environment based on raster datasets. The agents represent the fire front and operate below the lower bound of resolution. Furthermore, this approach can be easily adjusted to data of different resolutions. The model is applied to a realistic landscape and the results are compared to real fire behaviour and cellular automata models.