Predicting crown fire behavior to support forest fire managementdecision-making

摘要

Fire behavior models are an important component of decision support systems for fire management. This paper describes the modeling of two fundamental crown fire behavior features: the onset of crowning and the spread rate of crown fires. The present study is based largely on the database used in the development of the Canadian Forest Fire Behavior Prediction System. The dataset used in the study consisted of 73 experimental fires in various coniferous forest fuel complexes, 40 of which were classified as crown fires. These fires cover a wide spectrum of fire environment conditions and fire behavior characteristics, with rates of spread ranging from 0.5 - 49.4 m/min, and fireline intensity from 62 – 45,200 kW/m. Crown fire initiation was modeled through a logistic regression approach using 10-m open wind speed, fuel strata gap or height to live crown base, a surface fuel consumption class, and an index of fine dead fuel moisture content as independent variables. Spread rates for active and passive crown fires were modeled through multiple non-linear regression analysis following physical reasoning. Independent variables used in the crown fire spread models were 10-m open wind speed, canopy bulk density and again the index of fine dead fuel moisture content. The crown fire initiation model correctly predicted 85 % of the cases in the dataset used for its construction. The active crown fire spread model yield a R2 of 0.61. The wide variation in fuel complex structure and fire behavior in datasets used to build the crown fire initiation and rate of spread models gives confidence that the models might work well in fuel complexes different from the original ones, given an adequate description of the physical characteristics of the fuel complex.