A probability model for long-term forest fire occurrence in the Karst forest management area of Slovenia

Abstract

The aim of this study is to develop a long-term forest fire occurrence probability model in the Karst forest management area of Slovenia. The target area has the greatest forest fire occurrence rates and the largest burned areas in the country. To discover how the forest stand characteristics influence forest fire occurrence, we developed a long-term linear regression model. The geographically weighted regression method was applied to build the model, using forest management plans and land-based datasets as explanatory variables and a past forest fire activity dataset as a predicted variable. The land-based dataset was used to represent human activity as a key component in fire occurrence. Variables representing the natural and the anthropogenic environment used in the model explained 39% of past forest fire occurrences and predicted areas with the highest likelihood of forest fire occurrence. The results show that forest fire occurrence probability in a stand increases with lower wood stock, lower species diversity and lower thickness diversity, and in stands dominated by conifer trees under normal canopy closure. These forests stand characteristics are planned to be used in forest management and silviculture planning to reduce fire damage in Slovenian forests.