Wildfire is the most severe threat to Portuguese forests. It is widely accepted that the most cost-effective means for reducing wildfire incidence is by prevention. However, no wildfire probability models are available linking biometric data (that may be controllable by forest managers) with fire risk. This study presents a model that could contribute to overcome this gap. Emphasis was placed in developing a model based on easily measurable data, so that it might be useful to forest managers. Data from the 3rd Portuguese National Forest Inventory and from wildfire perimeters in the years from 1998 to 2004 were used for modelling purposes. A binary logistic regression model was developed to predict the probability of wildfire propagation in pure and mixed stands in Portugal. The models were constructed by exploring relationships between fire propagation and explanatory biometric variables (e.g. number of trees, understory shrub biomass load, basal area) as well as non-biometric variables (climatic, topographic and population related predictors). Results showed that probability of wildfire propagation in a stand increases with understory shrub biomass load and decreases with tree size and presence of hardwoods. These results are instrumental for assessing the impact of forest management on wildfire risk levels, helping forest managers reduce the risk of wildfires.