Evaluation performances of different forest fire spread models using cellular automata (case study: The forests of Lakan district in Rasht)

Abstract

Accurate prediction of forest fire spread is crucial in minimizing its destructive effects. Forest fire depends on various factors e.g. topography, vegetation and climate. One of the challenges in modeling forest fire concerns the way it interacts with static and dynamic spatiotemporal trajectories affecting its spread such as slope, wind speed and wind direction. In this study, three previously developed approaches Karafyllidis, Berjak and Progias for modeling those parameters were analyzed, followed by investigating the effects of pixel size and time steps in a cellular automata. The study was conducted in the Lakan forest district in the vicinity of Rasht in Guilan province. The available topographic, vegetation, wind speed and wind direction data were initially analyzed in GIS. Then the three modeling approaches were implemented, followed by a consequent sensitivity analysis on the pixel size and time steps of switching in cellular automata, The effectiveness of the approaches was compared by means of Kappa coefficient .The results indicate that the Berjak method with a 3-7 m pixel size is more appropriate for modeling the spread of fire across the study site.

Keywords

• forest fire
• pixel size
• Cellular Automata
• time steps
• dynamic spatiotemporal parameters