Abstract

Forest fires are recognized as a serious problem. Despite its importance, research into modelling of forest fire occurrence is lacking for the southern hemisphere, in particular for Chile. We investigated how landscape heterogeneity affects the probability of the occurrence of forest fires in Central Chile. We fitted a logistic regression model which included climatic, topographic, human-related and land-cover variables. Estimated probabilities of forest fire occurrence increased positively with the minimum temperature and precipitation seasonality, and decreased with distance to cities. The model correctly classified 65% of the validation data set. Our predictions suggest that 38% of the study area has a very high probability of forest fire occurrence. Under the assumed future scenario we predict an important increase in the probability of forest fire occurrence. The prediction model should be seen as a tool for addressing urgent protection and prevention issues, which should be discussed and evaluated further.