The objective of this work is to develop an empirical model that estimates spatially distributed probabilities of ignition of wildland fires in central Portugal. Upon validation, this model can be used as a decision support tool for management and help delineate fire prevention actions. In addition, the model can be included as a component of a fire risk index, providing information on both environmental and human factors.
The model is based on variables such as biomass, cover type, elevation, slope, distance to roads, distance to human settlings and to landuse interfaces, among others, and is derived by applying logistic regression techniques in a raster based geographical information system (GIS). The GIS integrates themes derived from satellite images, digital maps and field data. The field data used was collected by the Portuguese Forest Service and corresponds to the years of 1992 and 1993, and consists of records on fire ignition. A layer of ignition points is created and the relationship between these cells and the values of the other landscape variables can be analysed and maybe a pattern can be identified. For the analysis three quarters of the ignition cells are used, so that the model can be validated with the remainder. Additionally, univariate statistical tests are used to assess the validity of individual model variables and to obtain a better understanding of the contribution of each variable before developing the global model.
In this paper we present the model resulting from the described study and its application for generating a fire risk map in central Portugal. The map presented corresponds to the situation in the summer of 1993. The model is suitable for generating updated versions of this map every year.