Abstract

Forest fire occurrence probability model
Track: Wildland Fire Management
Authors: Tomaz Sturm

We created the forest fire occurrence probability model based on forest stands data and past fire activity. To create a model, each forest stand was attributed a dependent variable, i.e. information on the development phase,canopy closure,main tree species regarding to growing stock,growing stock per hectare,the share of deciduous trees with regard to growing stock,Shannon-Wiener index tree species composition,Shannon-Wiener index size class distribution, elevation of forest stand. As an independent variable we used information on the past fire activity in the forest stands. An ordinary least squares regression was carried out for developing a model with the most significant variables. Five of the original variables have been chosen by the model which explains 1,3%. Geographically weighted regression model showed significant improvement which explains 20% of the variation of the dependent variables. We classified forest stands into four classes of forest fire probability on the basis of the model results.